test(e2e): synchronize disk restore and heal log assertions

ci: retain full e2e recovery diagnostics
fix(build): refresh cached git provenance for server builds
2026-09-09 21:56:03 +00:00 · 2026-09-09 13:27:18 +08:00 · 2026-09-09 12:31:04 +08:00 · 2026-09-09 12:31:03 +08:00 · 2026-09-09 11:12:59 +08:00 · 2026-09-09 10:23:59 +08:00
115 changed files with 8220 additions and 4225 deletions
@@ -1,2 +1,2 @@
-sha256-darwin=f0c78fdb93471575d9a64c5c46eae6c806bdd0bc10a6e33d7fb574aabd8db5a3
+sha256-darwin=f0d15f2d1183be319d9d977d20b48bb05a29a424c266eed910c67aaa6f1ff955
-sha256-linux=03ed7016cab672de9320e31375a0358eceacb4408b0e79cf063614fa7c878b87
+sha256-linux=340aa702576ebed5266b7c47e591f267a11178fa04bf3828d64f53fe92eb0907
@@ -1,2 +1,2 @@
-sha256-darwin=364f2329a7b72eb9f1608dbe1a3af37af4095354014f3cbe23ca448492d89961
+sha256-darwin=12d30fff5ed48fe95bbfb310dd507048f81782b2954e00eb72435708b1133f9c
-sha256-linux=60983f1ebe7068cf660d473c5f76c76a650410ccc99d71934ddca7fd67607987
+sha256-linux=e917b2fdb303d01e6008ac4f1836c7b268698afe8383bd2aa4e697061df85cf6
@@ -89,7 +89,6 @@ offline-enrollment-e2e-check: core-deps ## Build and exercise the dedicated offl
 test-wiring-check: ## Check tests stay registered and selected by their intended runners
 	@echo "🧪 Checking test wiring..."
 	$(RUSTFS_PYTHON_BIN) ./scripts/check_test_wiring.py
 	$(RUSTFS_PYTHON_BIN) ./scripts/ci_gate.py --check-workflow
 .PHONY: log-analyzer-rules-check
 log-analyzer-rules-check: core-deps ## Check log-analyzer rule anchors still exist verbatim in source
@@ -39,7 +39,6 @@ script-tests: ## Run shell script tests
 	./scripts/test_python_bin.sh
 	./scripts/check_embedded_secrets.sh --self-test
 	$(RUSTFS_PYTHON_BIN) ./scripts/check_test_wiring.py --self-test
 	$(RUSTFS_PYTHON_BIN) ./scripts/ci_gate.py --self-test
 	$(RUSTFS_PYTHON_BIN) ./scripts/check_security_coverage.py --self-test
 	$(RUSTFS_PYTHON_BIN) ./scripts/check_scheduled_validation_freshness.py --self-test
 	$(RUSTFS_PYTHON_BIN) ./scripts/test_security_workflow.py
@@ -197,12 +197,6 @@ test-group = 'e2e-cluster-nightly'
 filter = 'package(e2e_test) & (test(/^kms::kms_vault_test::/) | test(/^kms::kms_rekey_sweep_test::/) | test(/^kms::configured_roundtrip_test::test_configured_vault_kms_admin_and_versioned_cleanup$/))'
 test-group = 'e2e-vault'
 # This four-disk, 65-member rollback probe already drives up to 32 concurrent
 # durable deletions. Reserve this nextest run's capacity for its progress oracle.
 [[profile.default.overrides]]
 filter = 'package(rustfs-ecstore) & test(=store::init::tests::dispatch_manifest_rollback_bounded_concurrency_reaches_tail_behind_slow_member)'
 threads-required = "num-test-threads"
 # ---------------------------------------------------------------------------
 # ci profile — the strict CI gate (ci.yml `cargo nextest run --profile ci`)
 # ---------------------------------------------------------------------------
@@ -1,5 +1,5 @@
 {
-  "schema": 1,
+  "schema": 2,
  "cases": {
    "background-target-restart": {
      "gate": "G14",
@@ -13,6 +13,7 @@
      "min_objects": 9,
      "max_objects": 65,
      "topology": {"nodes": 4, "drives_per_node": 1},
      "erasure": {"data_blocks": 2, "parity_blocks": 2},
      "scope": "Target process restart, exact unversioned S3 bodies and replacement-disk shards; not power loss or EC8+4."
    },
    "background-target-crash": {
@@ -27,30 +28,284 @@
      "min_objects": 9,
      "max_objects": 65,
      "topology": {"nodes": 4, "drives_per_node": 1},
      "erasure": {"data_blocks": 2, "parity_blocks": 2},
      "scope": "Target process killed during partial background rebuild, real unclean-shutdown marker, exact unversioned S3 bodies and replacement-disk shards; not power loss or EC8+4."
    },
    "ec84-target-drive-restart": {
      "gate": "G14",
      "task": "W20/W21",
      "lane": "e2e-distributed",
      "suite": "e2e_test",
      "name": "distributed::heal_test::three_node_four_drive_ec8_4_root_heal_rebuilds_replaced_drive_after_restart",
      "oracle": "ec84-target-drive-restart.json",
      "evidence": "process-restart",
      "unclean_shutdown_marker": false,
      "min_objects": 5,
      "max_objects": 5,
      "topology": {"nodes": 3, "drives_per_node": 4},
      "erasure": {"data_blocks": 8, "parity_blocks": 4},
      "erasure_set_drive_count": 12,
      "scope": "3-node x 4-drive single-set EC8+4, graceful target restart, preformatted replacement drive, exact unversioned S3 bodies and physical target shards; not mixed-version, multi-pool or long-window ABBA."
    },
    "background-target-restart-ec8-4": {
      "gate": "G14",
      "task": "W21",
      "lane": "e2e-nightly",
      "suite": "e2e_test",
      "name": "heal_erasure_disk_rebuild_test::tests::test_cluster_root_heal_recovers_ec84_shards_after_background_target_restart",
      "oracle": "background-target-restart-ec8-4.json",
      "evidence": "process-restart",
      "unclean_shutdown_marker": false,
      "min_objects": 9,
      "max_objects": 65,
      "topology": {"nodes": 3, "drives_per_node": 4},
      "erasure": {"data_blocks": 8, "parity_blocks": 4},
      "erasure_set_drive_count": 12,
      "scope": "Target process restart during partial background rebuild on a single 3x4 EC8+4 set; exact unversioned S3 bodies and replacement-drive shards; not power loss, multi-set, or multi-pool."
    },
    "background-target-crash-ec8-4": {
      "gate": "G14",
      "task": "W21",
      "lane": "e2e-nightly",
      "suite": "e2e_test",
      "name": "heal_erasure_disk_rebuild_test::tests::test_cluster_root_heal_recovers_ec84_shards_after_background_target_crash",
      "oracle": "background-target-crash-ec8-4.json",
      "evidence": "process-crash-restart",
      "unclean_shutdown_marker": true,
      "min_objects": 9,
      "max_objects": 65,
      "topology": {"nodes": 3, "drives_per_node": 4},
      "erasure": {"data_blocks": 8, "parity_blocks": 4},
      "erasure_set_drive_count": 12,
      "scope": "Target process killed during partial background rebuild on a single 3x4 EC8+4 set; real unclean-shutdown marker, exact unversioned S3 bodies and replacement-drive shards; not power loss, multi-set, or multi-pool."
    }
  },
-  "release_pending": {
+  "release_lanes": {
-    "G01": "W02/W04 complete root and quota authority coverage",
+    "single-set-restart": {
-    "G02": "W03 bounded checkpoint progress and independent version inventory",
+      "status": "implemented",
-    "G03": "W17/W18 exact scoped ACK with durable publication and mixed peers",
+      "cases": ["background-target-restart", "background-target-crash"],
-    "G04": "W03/W15/W16 crash at every cache/root/floor/intent boundary",
+      "covers": ["four-node one-drive topology", "unversioned objects", "target restart/crash"]
-    "G05": "W06/W07 per-object outcomes and bounded terminal retention",
+    },
-    "G06": "W06/W08/W23 concurrent status, legacy clients and truncation",
+    "authority-coverage": {
-    "G07": "W12/W13/W14 durable MRF responsibility at every commit boundary",
+      "status": "pending",
-    "G08": "W12/W13/W14 MRF capacity, disk-full and replica-loss matrix",
+      "gates": ["G01", "G12"],
-    "G09": "W13/W18/W23 actual mixed-version reader/writer and rollback payloads",
+      "requires": ["root authority coverage", "quota authority coverage"]
-    "G10": "W05/W09/W10/W11 bounded scheduling and pressure recovery",
+    },
-    "G11": "W04/W19/W24 maintenance and complete producer coverage",
+    "checkpoint-and-crash": {
-    "G12": "W02/W15/W16 both quota paths during reset and settlement",
+      "status": "pending",
-    "G13": "W07/W14 quorum-minus-one, unknown disks, remount, Object Lock, dry-run, grace and commit tail",
+      "gates": ["G02", "G04", "R-E"],
-    "G14": "W20/W21 same-window field evidence; 3x4 EC8+4 and multi-set/pool coverage",
+      "requires": ["bounded checkpoint progress", "boundary crash matrix", "fixed-budget restart evidence"]
-    "P1": "W20 measured cold-walk share and foreground latency/throughput",
+    },
-    "P2": "W20/W24 measured post-stop convergence and cold segment reuse",
+    "status-and-outcome": {
-    "P3": "W20 measured two-hour pressure/heal capacity and recovery window",
+      "status": "pending",
-    "P4": "W20 measured MRF scale and replay cost with retained responsibility",
+      "gates": ["G05", "G06", "R-D"],
-    "R-E": "W03/W05 fixed-budget real process restart through enumeration and classification",
+      "requires": ["per-object outcomes", "legacy status clients", "manager/event/ledger disposition"]
-    "R-D": "W07/W14 manager-to-event-to-ledger exact disposition, including grace",
+    },
-    "R-L": "W13/W14 legacy source conflicts, migration gaps and crash-safe source retirement"
+    "mrf-responsibility": {
-  }
+      "status": "pending",
      "gates": ["G07", "G08", "P4"],
      "requires": ["durable MRF responsibility", "disk-full and replica-loss matrix", "MRF replay cost"]
    },
    "mixed-version-rollback": {
      "status": "pending",
      "gates": ["G03", "G09", "R-L"],
      "requires": ["mixed-version peers", "rollback payloads", "crash-safe source retirement"]
    },
    "scheduler-pressure": {
      "status": "pending",
      "gates": ["G10", "P1", "P2", "P3"],
      "requires": ["bounded scheduling", "foreground latency and throughput", "two-hour pressure evidence"]
    },
    "maintenance-producers": {
      "status": "pending",
      "gates": ["G11", "G13"],
      "requires": ["complete producer coverage", "quorum-minus-one and remount matrix"]
    },
    "ec8-4-multiset": {
      "status": "pending",
      "gates": ["G14"],
      "requires": ["3x4 EC8+4 topology", "multi-set coverage", "multi-pool coverage"]
    }
  },
  "release_requirements": [
    {
      "gate": "G01",
      "task": "W02/W04",
      "lane": "authority-coverage",
      "status": "pending",
      "description": "Complete root and quota authority coverage",
      "requires": ["root authority evidence", "quota authority evidence"]
    },
    {
      "gate": "G02",
      "task": "W03",
      "lane": "checkpoint-and-crash",
      "status": "pending",
      "description": "Bounded checkpoint progress and independent version inventory",
      "requires": ["bounded checkpoint oracle", "independent version inventory"]
    },
    {
      "gate": "G03",
      "task": "W17/W18",
      "lane": "mixed-version-rollback",
      "status": "pending",
      "description": "Exact scoped ACK with durable publication and mixed peers",
      "requires": ["durable scoped ACK publication", "mixed-peer evidence"],
      "evidence_fields": [
        "durable_root_publication_proof",
        "scoped_ack_request_identity",
        "participating_peer_capability_snapshot",
        "mixed_peer_ack_fallback_oracle"
      ]
    },
    {
      "gate": "G04",
      "task": "W03/W15/W16",
      "lane": "checkpoint-and-crash",
      "status": "pending",
      "description": "Crash at every cache, root, floor and intent boundary",
      "requires": ["cache boundary crash evidence", "root/floor/intent crash evidence"]
    },
    {
      "gate": "G05",
      "task": "W06/W07",
      "lane": "status-and-outcome",
      "status": "pending",
      "description": "Per-object outcomes and bounded terminal retention",
      "requires": ["per-object outcome oracle", "terminal retention bounds"]
    },
    {
      "gate": "G06",
      "task": "W06/W08/W23",
      "lane": "status-and-outcome",
      "status": "pending",
      "description": "Concurrent status, legacy clients and truncation",
      "requires": ["concurrent status evidence", "legacy client compatibility", "truncation behavior"]
    },
    {
      "gate": "G07",
      "task": "W12/W13/W14",
      "lane": "mrf-responsibility",
      "status": "pending",
      "description": "Durable MRF responsibility at every commit boundary",
      "requires": ["MRF responsibility oracle", "commit-boundary crash matrix"]
    },
    {
      "gate": "G08",
      "task": "W12/W13/W14",
      "lane": "mrf-responsibility",
      "status": "pending",
      "description": "MRF capacity, disk-full and replica-loss matrix",
      "requires": ["MRF capacity evidence", "disk-full matrix", "replica-loss matrix"]
    },
    {
      "gate": "G09",
      "task": "W13/W18/W23",
      "lane": "mixed-version-rollback",
      "status": "pending",
      "description": "Actual mixed-version reader/writer and rollback payloads",
      "requires": ["mixed-version reader evidence", "mixed-version writer evidence", "rollback payload evidence"],
      "evidence_fields": [
        "mixed_version_reader_evidence",
        "mixed_version_writer_evidence",
        "rollback_payload_evidence"
      ]
    },
    {
      "gate": "G10",
      "task": "W05/W09/W10/W11",
      "lane": "scheduler-pressure",
      "status": "pending",
      "description": "Bounded scheduling and pressure recovery",
      "requires": ["scheduler bound evidence", "pressure recovery evidence"]
    },
    {
      "gate": "G11",
      "task": "W04/W19/W24",
      "lane": "maintenance-producers",
      "status": "pending",
      "description": "Maintenance and complete producer coverage",
      "requires": ["maintenance producer matrix", "complete producer inventory"]
    },
    {
      "gate": "G12",
      "task": "W02/W15/W16",
      "lane": "authority-coverage",
      "status": "pending",
      "description": "Both quota paths during reset and settlement",
      "requires": ["reset quota-path evidence", "settlement quota-path evidence"]
    },
    {
      "gate": "G13",
      "task": "W07/W14",
      "lane": "maintenance-producers",
      "status": "pending",
      "description": "Quorum-minus-one, unknown disks, remount, Object Lock, dry-run, grace and commit tail",
      "requires": ["quorum-minus-one matrix", "unknown-disk/remount matrix", "Object Lock dry-run grace evidence"]
    },
    {
      "gate": "G14",
      "task": "W20/W21",
      "lane": "ec8-4-multiset",
      "status": "pending",
      "description": "Same-window field evidence with 3x4 EC8+4 and multi-set/pool coverage",
      "requires": ["same-window field evidence", "3x4 EC8+4 evidence", "multi-set evidence", "multi-pool evidence"]
    },
    {
      "gate": "P1",
      "task": "W20",
      "lane": "scheduler-pressure",
      "status": "pending",
      "description": "Measured cold-walk share and foreground latency/throughput",
      "requires": ["cold-walk share measurement", "foreground latency/throughput measurement"]
    },
    {
      "gate": "P2",
      "task": "W20/W24",
      "lane": "scheduler-pressure",
      "status": "pending",
      "description": "Measured post-stop convergence and cold segment reuse",
      "requires": ["post-stop convergence measurement", "cold segment reuse measurement"]
    },
    {
      "gate": "P3",
      "task": "W20",
      "lane": "scheduler-pressure",
      "status": "pending",
      "description": "Measured two-hour pressure/heal capacity and recovery window",
      "requires": ["two-hour pressure measurement", "heal capacity measurement", "recovery-window measurement"]
    },
    {
      "gate": "P4",
      "task": "W20",
      "lane": "mrf-responsibility",
      "status": "pending",
      "description": "Measured MRF scale and replay cost with retained responsibility",
      "requires": ["MRF scale measurement", "MRF replay-cost measurement", "retained responsibility evidence"]
    },
    {
      "gate": "R-E",
      "task": "W03/W05",
      "lane": "checkpoint-and-crash",
      "status": "pending",
      "description": "Fixed-budget real process restart through enumeration and classification",
      "requires": ["fixed-budget restart evidence", "enumeration evidence", "classification evidence"]
    },
    {
      "gate": "R-D",
      "task": "W07/W14",
      "lane": "status-and-outcome",
      "status": "pending",
      "description": "Manager-to-event-to-ledger exact disposition, including grace",
      "requires": ["manager disposition evidence", "event disposition evidence", "ledger disposition evidence", "grace handling"]
    },
    {
      "gate": "R-L",
      "task": "W13/W14",
      "lane": "mixed-version-rollback",
      "status": "pending",
      "description": "Legacy source conflicts, migration gaps and crash-safe source retirement",
      "requires": ["legacy source-conflict evidence", "migration-gap evidence", "crash-safe source retirement evidence"]
    }
  ]
 }
@@ -111,6 +111,10 @@ runs:
      shell: bash
      run: ./scripts/check_no_planning_docs.sh
    - name: Check CI paths stay in sync
      shell: bash
      run: ./scripts/check_ci_paths_sync.sh
    - name: Check io_uring lane --lib precondition
      shell: bash
      run: ./scripts/check_uring_lane_lib_only.sh
@@ -16,7 +16,7 @@ name: Security Audit
 on:
  push:
-    branches: [ main ]
+    branches: [ main, release ]
    paths:
      - '**/Cargo.toml'
      - '**/Cargo.lock'
@@ -32,7 +32,7 @@ on:
      - 'scripts/security/check_workflow_pins.sh'
  pull_request:
    types: [ opened, synchronize, reopened, closed ]
-    branches: [ main ]
+    branches: [ main, release ]
    paths:
      - '**/Cargo.toml'
      - '**/Cargo.lock'
@@ -0,0 +1,79 @@
 # Copyright 2026 RustFS Team
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at
 #
 #     http://www.apache.org/licenses/LICENSE-2.0
 #
 # Unless required by applicable law or agreed to in writing, software
 # distributed under the License is distributed on an "AS IS" BASIS,
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # Reports the existing required checks for paths excluded by ci.yml.
 # Mixed PRs can trigger both workflows; their Quick Checks jobs use one shared
 # action to keep validation coverage aligned. Keep this paths list in sync with
 # ci.yml's pull_request.paths-ignore via scripts/check_ci_paths_sync.sh.
 name: Continuous Integration (docs only)
 on:
  pull_request:
    types: [ opened, synchronize, reopened ]
    branches: [ main, release ]
    paths:
      - "**.md"
      - "docs/**"
      - "deploy/**"
      - "scripts/dev_*.sh"
      - "scripts/probe.sh"
      - "LICENSE*"
      - ".gitignore"
      - ".dockerignore"
      - "README*"
      - "**/*.png"
      - "**/*.jpg"
      - "**/*.svg"
      - ".github/workflows/build.yml"
      - ".github/workflows/docker.yml"
      - ".github/workflows/audit.yml"
      - "flake.lock"
 permissions:
  contents: read
 jobs:
  quick-checks:
    name: Quick Checks
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - name: Checkout repository
        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
        with:
          persist-credentials: false
      - name: Run shared quick checks
        uses: ./.github/actions/quick-checks
  test-and-lint:
    name: Test and Lint
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - name: Checkout repository
        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
        with:
          persist-credentials: false
      # Docs-only PRs skip the full code CI, but they are exactly where a
      # planning-type document could be slipped in (git add -f bypasses
      # .gitignore). Run the guard here so the required "Test and Lint" check
      # stays meaningful for docs-only changes.
      - name: Check no planning docs committed
        run: ./scripts/check_no_planning_docs.sh
      - name: Satisfy required check for docs-only changes
        run: echo "Docs-only change — code CI is skipped by paths-ignore; planning-docs guard passed, reporting success for the required 'Test and Lint' check."
@@ -16,7 +16,7 @@ name: Continuous Integration
 on:
  push:
-    branches: [ main ]
+    branches: [ main, release ]
    paths-ignore:
      - "**.md"
      - "docs/**"
@@ -36,7 +36,26 @@ on:
      - "flake.lock"
  pull_request:
    types: [ opened, synchronize, reopened, closed ]
-    branches: [ main ]
+    branches: [ main, release ]
    # Keep this list in sync with the `paths` list in ci-docs-only.yml, which
    # reports the required "Test and Lint" check for PRs skipped here.
    paths-ignore:
      - "**.md"
      - "docs/**"
      - "deploy/**"
      - "scripts/dev_*.sh"
      - "scripts/probe.sh"
      - "LICENSE*"
      - ".gitignore"
      - ".dockerignore"
      - "README*"
      - "**/*.png"
      - "**/*.jpg"
      - "**/*.svg"
      - ".github/workflows/build.yml"
      - ".github/workflows/docker.yml"
      - ".github/workflows/audit.yml"
      - "flake.lock"
  merge_group:
    types: [ checks_requested ]
  schedule:
@@ -69,32 +88,6 @@ jobs:
      - name: Explain cancellation run
        run: echo "PR closed; this run only cancels older runs in the same concurrency group."
  classify-changes:
    name: Select CI scope
    if: github.event_name != 'pull_request' || github.event.action != 'closed'
    runs-on: ubuntu-latest
    timeout-minutes: 10
    outputs:
      mode: ${{ steps.scope.outputs.mode }}
    steps:
      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
        with:
          fetch-depth: 2
          persist-credentials: false
      - name: Select scope using the base revision's policy
        id: scope
        env:
          CI_BASE_SHA: ${{ github.event.pull_request.base.sha }}
        run: |
          if [[ "$GITHUB_EVENT_NAME" != "pull_request" ]]; then
            printf '%s\n' 'mode=full' >> "$GITHUB_OUTPUT"
          elif [[ "$CI_BASE_SHA" =~ ^[0-9a-f]{40}$ ]] && git show "$CI_BASE_SHA:scripts/ci_gate.py" > "$RUNNER_TEMP/ci-gate-base.py"; then
            python3 -I "$RUNNER_TEMP/ci-gate-base.py" select
          else
            printf '%s\n' 'mode=full' >> "$GITHUB_OUTPUT"
            echo "Base CI policy unavailable; running the full matrix."
          fi
  typos:
    name: Typos
    if: github.event_name != 'pull_request' || github.event.action != 'closed'
@@ -107,7 +100,7 @@ jobs:
      - name: Typos check with custom config file
        uses: crate-ci/typos@37bb98842b0d8c4ffebdb75301a13db0267cef89 # master
-  # Fail early with compile-free checks for every pull request.
+  # Fail early with compile-free checks shared with docs-only CI.
  quick-checks:
    name: Quick Checks
    if: github.event_name != 'pull_request' || github.event.action != 'closed'
@@ -123,9 +116,9 @@ jobs:
        uses: ./.github/actions/quick-checks
  test-and-lint:
-    name: Workspace Test and Lint
+    name: Test and Lint
-    if: needs.classify-changes.outputs.mode == 'full' && (github.event_name != 'pull_request' || github.event.action != 'closed')
+    if: github.event_name != 'pull_request' || github.event.action != 'closed'
-    needs: [ quick-checks, classify-changes ]
+    needs: [ quick-checks ]
    runs-on: sm-standard-4
    timeout-minutes: 90
    env:
@@ -296,6 +289,45 @@ jobs:
      - name: Run rebalance/decommission migration proofs
        run: ./scripts/check_migration_gate_count.sh
      # Record the reason before this job completes as FAILURE. A separate
      # dependent job cancels sibling lanes only after GitHub has preserved this
      # required check's failure verdict.
      - name: Annotate early-stop reason
        if: >-
          failure() && github.event_name == 'pull_request'
          && github.event.pull_request.head.repo.full_name == github.repository
        run: |
          {
            echo "## CI early-stop"
            echo "Job \`${GITHUB_JOB}\` (Test and Lint) failed; a follow-up job will cancel sibling lanes to free runners."
            echo "Sibling jobs showing **cancelled** were stopped by the early-stop follow-up, not by their own failure."
          } >> "$GITHUB_STEP_SUMMARY"
  # Preserve the required Test and Lint FAILURE verdict before stopping sibling
  # lanes. Cancelling from inside test-and-lint changed its own conclusion to
  # CANCELLED and hid the actionable failure in the PR checks UI.
  cancel-after-test-and-lint-failure:
    name: Cancel siblings after Test and Lint failure
    if: >-
      failure() && needs.test-and-lint.result == 'failure'
      && github.event_name == 'pull_request'
      && github.event.pull_request.head.repo.full_name == github.repository
    needs: [ test-and-lint ]
    runs-on: ubuntu-latest
    timeout-minutes: 5
    permissions:
      actions: write
    steps:
      - name: Cancel remaining jobs
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
        run: |
          curl -fsS -X POST \
            -H "Authorization: Bearer ${GH_TOKEN}" \
            -H "Accept: application/vnd.github+json" \
            -H "X-GitHub-Api-Version: 2022-11-28" \
            "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/actions/runs/${GITHUB_RUN_ID}/cancel"
  # Dedicated serial lane for the ILM / lifecycle integration tests. These tests
  # drive the object layer through process-global singletons (the GLOBAL_ENV
  # ECStore, the global tier-config manager, background-expiry workers) and bind
@@ -308,8 +340,8 @@ jobs:
  # See rustfs/backlog#1148 (ilm-1) and #1155.
  test-ilm-integration-serial:
    name: ILM Integration (serial)
-    if: needs.classify-changes.outputs.mode == 'full' && (github.event_name != 'pull_request' || github.event.action != 'closed')
+    if: github.event_name != 'pull_request' || github.event.action != 'closed'
-    needs: [ quick-checks, classify-changes ]
+    needs: [ quick-checks ]
    runs-on: sm-standard-4
    timeout-minutes: 90
    env:
@@ -376,8 +408,8 @@ jobs:
  test-and-lint-rio-v2:
    name: Test and Lint (rio-v2)
-    if: needs.classify-changes.outputs.mode == 'full' && (github.event_name != 'pull_request' || github.event.action != 'closed')
+    if: github.event_name != 'pull_request' || github.event.action != 'closed'
-    needs: [ quick-checks, classify-changes ]
+    needs: [ quick-checks ]
    runs-on: sm-standard-4
    timeout-minutes: 90
    env:
@@ -417,8 +449,8 @@ jobs:
  connect-short-credential-boundary:
    name: Connect Short Credential Boundary
-    if: needs.classify-changes.outputs.mode == 'full' && (github.event_name != 'pull_request' || github.event.action != 'closed')
+    if: github.event_name != 'pull_request' || github.event.action != 'closed'
-    needs: [ quick-checks, classify-changes ]
+    needs: [ quick-checks ]
    runs-on: sm-standard-4
    timeout-minutes: 60
    env:
@@ -475,8 +507,8 @@ jobs:
  test-and-lint-protocols:
    name: "Test and Lint (${{ matrix.features.name }})"
-    if: needs.classify-changes.outputs.mode == 'full' && (github.event_name != 'pull_request' || github.event.action != 'closed')
+    if: github.event_name != 'pull_request' || github.event.action != 'closed'
-    needs: [ quick-checks, classify-changes ]
+    needs: [ quick-checks ]
    runs-on: sm-standard-4
    timeout-minutes: 90
    strategy:
@@ -529,8 +561,8 @@ jobs:
  build-rustfs-debug-binary:
    name: Build RustFS Debug Binary
-    if: needs.classify-changes.outputs.mode == 'full' && (github.event_name != 'pull_request' || github.event.action != 'closed')
+    if: github.event_name != 'pull_request' || github.event.action != 'closed'
-    needs: [ quick-checks, classify-changes ]
+    needs: [ quick-checks ]
    runs-on: sm-standard-4
    timeout-minutes: 30
    env:
@@ -652,8 +684,8 @@ jobs:
    # job had neither, so each closed/merged PR really ran the whole io_uring
    # suite (measured 4m17s / 7m19s / 7m31s on runs 30678272341 / 30678117601 /
    # 30662728539) and kept the cancellation run in progress for minutes.
-    if: needs.classify-changes.outputs.mode == 'full' && (github.event_name != 'pull_request' || github.event.action != 'closed')
+    if: github.event_name != 'pull_request' || github.event.action != 'closed'
-    needs: [ quick-checks, classify-changes ]
+    needs: [ quick-checks ]
    # GitHub-hosted ubuntu-latest runs a recent kernel with io_uring and, unlike
    # a container, applies no seccomp filter that would block io_uring_setup — so
    # the probe succeeds and the tests exercise the real UringBackend/FdCache/
@@ -840,13 +872,14 @@ jobs:
    # Merge gate only (backlog#1149 ci-5): the never-automated user-visible
    # suites — KMS, object_lock, multipart_auth, quota, checksum, encryption,
    # security-boundary, ... — via the e2e-full nextest profile. Too heavy for
-    # every PR, so it is gated to main pushes, the merge queue, and manual
+    # every PR, so it is gated to main/release pushes, the merge queue, and manual
    # dispatch. protocols / the 7 cluster suites / replication / #[ignore] are
    # owned by other lanes (see .config/nextest.toml profile.e2e-full).
    if: >-
      github.event_name == 'workflow_dispatch' ||
      github.event_name == 'merge_group' ||
-      (github.event_name == 'push' && github.ref == 'refs/heads/main')
+      (github.event_name == 'push' &&
        (github.ref == 'refs/heads/main' || github.ref == 'refs/heads/release'))
    needs: [ build-rustfs-debug-binary ]
    runs-on: sm-standard-2
    timeout-minutes: 55
@@ -969,11 +1002,22 @@ jobs:
      # debug binary; each test spawns its own rustfs server on a random port.
      - name: Run e2e full suite
        env:
          RUSTFS_E2E_LOG_DIR: ${{ runner.temp }}/rustfs-e2e-full-logs
          RUSTFS_HEAL_CHAOS_SERVER_RUST_LOG: rustfs::heal=debug,rustfs_heal=debug,rustfs_ecstore::disk=debug,rustfs_lock=debug,rustfs=error
          RUSTFS_E2E_STARTUP_CAS_BINARY: ${{ runner.temp }}/rustfs-startup-cas-input/rustfs
          RUSTFS_E2E_STARTUP_CAS_BUILD_MANIFEST: ${{ runner.temp }}/rustfs-startup-cas-input/rustfs.e2e-startup-cas-build.json
          RUSTFS_E2E_STARTUP_CAS_ARTIFACT_DIR: ${{ runner.temp }}/rustfs-startup-cas-evidence
        run: cargo nextest run --profile e2e-full -p e2e_test
      - name: Upload e2e full diagnostics
        if: failure()
        uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
        with:
          name: e2e-full-diagnostics-${{ github.run_number }}
          path: ${{ runner.temp }}/rustfs-e2e-full-logs/
          if-no-files-found: warn
          retention-days: 7
      - name: Upload junit
        if: always()
        uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
@@ -1180,44 +1224,9 @@ jobs:
          if-no-files-found: ignore
          retention-days: 3
  required-checks:
    name: Test and Lint
    if: always() && (github.event_name != 'pull_request' || github.event.action != 'closed')
    needs:
      - classify-changes
      - typos
      - quick-checks
      - test-and-lint
      - test-ilm-integration-serial
      - test-and-lint-rio-v2
      - connect-short-credential-boundary
      - test-and-lint-protocols
      - build-rustfs-debug-binary
      - uring-integration
      - e2e-tests
      - s3-implemented-tests
      - s3-lifecycle-behavior-tests
      - build-rustfs-debug-binary-rio-v2
      - e2e-tests-rio-v2
      - e2e-full
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
        with:
          persist-credentials: false
      - name: Require the expected result of every CI lane
        env:
          CI_NEEDS: ${{ toJSON(needs) }}
        shell: bash
        run: python3 scripts/ci_gate.py verify
  alert-on-failure:
    name: Alert on scheduled failure
    needs:
      - classify-changes
      - connect-short-credential-boundary
      - required-checks
      - typos
      - quick-checks
      - test-and-lint
@@ -19,7 +19,7 @@
 # case is a two-site 4-node 1-drive pair or a 4-node upgrade). Membership is
 # `[profile.e2e-distributed]` in `.config/nextest.toml`. Storage-sensitive PRs,
 # nightly runs, and manual dispatches all execute the same fail-closed suite.
-# Upgrade cases download the same pinned previous release as e2e-upgrade.yml.
+# Upgrade cases use an independent 1.0.0-rc.2 pin defined below.
 #
 # Isolated pool filesystems: expand/decommission/rebalance cases require
 # independent `statfs` capacity. This job runs on GitHub-hosted
@@ -87,7 +87,7 @@ jobs:
      NO_PROXY: 127.0.0.1,localhost
      HTTP_PROXY: ""
      HTTPS_PROXY: ""
-      # Pinned previous release used by distributed::upgrade_test (same pin as e2e-upgrade.yml).
+      # Independent 1.0.0-rc.2 source pin for distributed::upgrade_test.
      UPGRADE_SOURCE_VERSION: 1.0.0-rc.2
      UPGRADE_SOURCE_ASSET: rustfs-linux-x86_64-gnu-v1.0.0-rc.2.zip
      UPGRADE_SOURCE_SHA256: 7c789386bf85278f865b8e0d359bf4edb84d5aa408cc3fa54a18c25ca74cd6e7
@@ -19,27 +19,17 @@ on:
    - cron: "7 0 * * *"
      timezone: "Asia/Shanghai"
  workflow_dispatch:
    inputs:
      branch:
        description: 'Branch/ref to build and publish as the nightly (empty = scheduled source, see NIGHTLY_BUILD_REF)'
        required: false
        default: ''
 permissions:
  contents: read
 # Scheduled builds follow the NIGHTLY_BRANCH repo variable so the channel can
 # be pointed at e.g. `release` for the GA cycle and back to `main` afterwards
 # without touching this file. Manual runs take the `branch` input, falling
 # back to the branch the run was dispatched from.
 concurrency:
-  group: nightly-gnu-build-${{ github.event_name }}-${{ github.event_name == 'schedule' && (vars.NIGHTLY_BRANCH || 'main') || (inputs.branch || github.ref_name) }}
+  group: nightly-gnu-build-main-${{ github.event_name }}
  cancel-in-progress: ${{ github.event_name == 'workflow_dispatch' }}
 env:
  CARGO_TERM_COLOR: always
  RUST_BACKTRACE: 1
  NIGHTLY_BUILD_REF: ${{ github.event_name == 'schedule' && (vars.NIGHTLY_BRANCH || 'main') || (inputs.branch || github.ref_name) }}
 jobs:
  build:
@@ -53,7 +43,6 @@ jobs:
        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
        with:
          persist-credentials: false
          ref: ${{ env.NIGHTLY_BUILD_REF }}
      - name: Setup Rust environment
        uses: ./.github/actions/setup
@@ -163,104 +152,13 @@ jobs:
          fakeroot dpkg-deb --build "${PKG_DIR}"
          ls -lh "${DEB_FILE}"
          echo "deb_date=${DEB_DATE}" >> "${GITHUB_OUTPUT}"
          echo "deb_file=${DEB_FILE}" >> "${GITHUB_OUTPUT}"
      # Same packaging scheme as .github/workflows/package.yml (fpm), but from
      # the locally built nightly binary instead of a release artifact, with a
      # date-based version that mirrors the DEB.
      - name: Build RPM package
        id: rpm
        shell: bash
        env:
          DEB_DATE: ${{ steps.deb.outputs.deb_date }}
        run: |
          set -euo pipefail
          if ! command -v fpm >/dev/null 2>&1; then
            SUDO=""; [ "$(id -u)" -ne 0 ] && SUDO="sudo -n"
            ${SUDO} apt-get update -qq && ${SUDO} apt-get install -y -qq ruby ruby-dev build-essential rpm >/dev/null
            ${SUDO} gem install fpm --no-document >/dev/null
          fi
          RPM_FILE="rustfs-nightly-${DEB_DATE}.rpm"
          RPM_VERSION="0"
          RPM_RELEASE="0.nightly.${DEB_DATE//-/.}"
          echo "Building RPM: ${RPM_FILE} (version ${RPM_VERSION}-${RPM_RELEASE})"
          # fpm wants the config file to exist before packaging.
          mkdir -p ./tmp-pkg/etc/default
          cat > ./tmp-pkg/etc/default/rustfs << 'ENVEOF'
          # RustFS Environment Configuration
          # See https://rustfs.com/docs/ for more information
          # RUSTFS_VOLUMES=""
          # RUSTFS_ROOT_USER=""
          # RUSTFS_ROOT_PASSWORD=""
          ENVEOF
          fpm -s dir -t rpm \
            --name rustfs \
            --version "$RPM_VERSION" \
            --iteration "$RPM_RELEASE" \
            --architecture x86_64 \
            --package "$RPM_FILE" \
            --depends "glibc >= 2.31" \
            --maintainer "RustFS Team <support@rustfs.com>" \
            --description "High-performance distributed object storage" \
            --url "https://rustfs.com" \
            --license "Apache-2.0" \
            --after-install <(cat << 'POSTINST'
          #!/bin/bash
          set -e
          if ! getent passwd rustfs > /dev/null 2>&1; then
              useradd -r -s /bin/false -d /opt/rustfs rustfs
          fi
          mkdir -p /opt/rustfs /data/rustfs /var/log/rustfs
          chown rustfs:rustfs /opt/rustfs /data/rustfs /var/log/rustfs
          if [ -d /run/systemd/system ]; then
              systemctl daemon-reload
          fi
          POSTINST
            ) \
            --before-remove <(cat << 'PRERM'
          #!/bin/bash
          set -e
          if [ -d /run/systemd/system ] && systemctl is-active --quiet rustfs; then
              systemctl stop rustfs
          fi
          PRERM
            ) \
            --after-remove <(cat << 'POSTRM'
          #!/bin/bash
          set -e
          if [ -d /run/systemd/system ]; then
              systemctl daemon-reload
          fi
          POSTRM
            ) \
            --config-files /etc/default/rustfs \
            "rustfs-nightly-${DEB_DATE}/usr/bin/rustfs=/usr/bin/rustfs" \
            ./tmp-pkg/etc/default/rustfs=/etc/default/rustfs \
            deploy/build/rustfs.service=/lib/systemd/system/rustfs.service \
            LICENSE=/usr/share/doc/rustfs/LICENSE \
            README.md=/usr/share/doc/rustfs/README.md
          [[ -f "$RPM_FILE" ]] || { echo "RPM build failed"; exit 1; }
          rpm -qpl "$RPM_FILE" | grep -Fx '/usr/bin/rustfs' >/dev/null
          stat --printf='%n %s bytes\n' "$RPM_FILE"
          echo "rpm_file=$RPM_FILE" >> "$GITHUB_OUTPUT"
      - name: Upload DEB artifact
        uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
        with:
          name: ${{ steps.deb.outputs.deb_file }}
          path: ${{ steps.deb.outputs.deb_file }}
      - name: Upload RPM artifact
        uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
        with:
          name: ${{ steps.rpm.outputs.rpm_file }}
          path: ${{ steps.rpm.outputs.rpm_file }}
          if-no-files-found: error
      # Persist the nightly deb on Cloudflare R2 (same channel as package.yml)
@@ -289,10 +187,11 @@ jobs:
          export AWS_SECRET_ACCESS_KEY="$R2_SECRET_ACCESS_KEY"
          export AWS_DEFAULT_REGION="auto"
          # The candidate manifest must describe the tree that was actually
          # built. With a ref override (NIGHTLY_BRANCH / dispatch input) that
          # is not necessarily GITHUB_SHA, so always advertise HEAD.
          SOURCE_SHA="$(git rev-parse HEAD)"
          if [[ "${SOURCE_SHA}" != "${GITHUB_SHA}" ]]; then
            echo "Checkout SHA does not match the nightly build run" >&2
            exit 1
          fi
          DEB_SHA256="$(sha256sum "${DEB_FILE}" | cut -d ' ' -f 1)"
          CANDIDATE_KEY="artifacts/rustfs/packages/nightly/runs/${GITHUB_RUN_ID}/${GITHUB_RUN_ATTEMPT}/${DEB_SHA256}/rustfs.deb"
          CANDIDATE_URL="https://dl.rustfs.com/${CANDIDATE_KEY}"
@@ -348,20 +247,6 @@ jobs:
          path: ${{ steps.publish.outputs.candidate_file }}
          if-no-files-found: error
      # Publish the deb/rpm pair to the auto-testing repo's `assets` branch so
      # engineers can download and install the nightly directly. The branch is
      # a single-commit orphan rewritten on every build, which keeps the repo
      # small while the latest files stay reachable at stable raw URLs.
      # Publish the deb/rpm pair as assets of the rolling `nightly` release on
      # rustfs/auto-testing (see scripts/release/publish_nightly_assets.sh).
      - name: Publish packages to auto-testing release assets
        env:
          ASSETS_TOKEN: ${{ secrets.PF_TESTING_GH_TOKEN }}
          DEB_FILE: ${{ steps.deb.outputs.deb_file }}
          RPM_FILE: ${{ steps.rpm.outputs.rpm_file }}
          DEB_DATE: ${{ steps.deb.outputs.deb_date }}
          BUILD_REF: ${{ env.NIGHTLY_BUILD_REF }}
        run: bash scripts/release/publish_nightly_assets.sh
  # Live-Vault lane for the rustfs-kms suite (rustfs/backlog#1774).
  #
  # RUSTFS_KMS_VAULT_TOKEN is the single switch that adds the Vault KV2 and
@@ -399,7 +284,6 @@ jobs:
        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
        with:
          persist-credentials: false
          ref: ${{ env.NIGHTLY_BUILD_REF }}
      - name: Setup Rust environment
        uses: ./.github/actions/setup
@@ -488,7 +372,6 @@ jobs:
        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
        with:
          persist-credentials: false
          ref: ${{ env.NIGHTLY_BUILD_REF }}
      - name: Setup Rust environment
        uses: ./.github/actions/setup
@@ -82,12 +82,6 @@ jobs:
  performance-test:
    runs-on: pf-testing
    timeout-minutes: 900
    env:
      RUSTFS_BENCH_SCRIPT: ${{ github.workspace }}/auto-testing/rustfs_performance_testing.sh
      RUSTFS_WARP_METHODS: ${{ inputs.test_method }}
      RUSTFS_WARP_SIZES: ${{ inputs.object_size }}
      RUSTFS_WARP_DURATION: ${{ inputs.warp_duration || '5m' }}
      RUSTFS_WARP_CONCURRENCY: ${{ inputs.warp_concurrency || '64' }}
    # Run on manual dispatch, or when the nightly build completed successfully.
    # Skipped when nightly failed.
    if: ${{ github.event_name == 'workflow_dispatch' || github.event_name == 'repository_dispatch' }}
@@ -164,15 +158,19 @@ jobs:
      - name: Run benchmark (GET/PUT/MIXED)
        id: benchmark
        run: |
          # Empty on automatic (workflow_run) runs -> full 30 rounds.
          # Manual dispatch can restrict method(s)/size(s).
          export WARP_METHODS="${{ inputs.test_method }}"
          export WARP_SIZES="${{ inputs.object_size }}"
          ./auto-testing/rustfs_performance_test.sh \
            --step 5 -y \
            --warp-duration "${{ inputs.warp_duration || '5m' }}" \
            --warp-concurrency "${{ inputs.warp_concurrency || '64' }}" \
            --log-file "${LOG_FILE}"
      - name: Analyze results
        if: ${{ steps.benchmark.conclusion == 'success' }}
        run: |
          export WARP_METHODS="${RUSTFS_WARP_METHODS}" WARP_SIZES="${RUSTFS_WARP_SIZES}"
          export WARP_DURATION="${RUSTFS_WARP_DURATION}" WARP_CONCURRENCY="${RUSTFS_WARP_CONCURRENCY}"
          ./auto-testing/rustfs_performance_test.sh --step 6 -y --log-file "${LOG_FILE:-/dev/null}"
      - name: Collect RustFS version info
@@ -52,7 +52,6 @@ docs
 __pycache__/
 !docs/
 docs/*
 !docs/README.md
 !docs/architecture/
 !docs/architecture/**
 !docs/operations/
@@ -7,16 +7,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [Unreleased]
 ### Replication
 - Object Lock replication PUTs now carry a required integrity header, fixing target rejection introduced by the plain-payload default ([#7097](https://github.com/rustfs/rustfs/pull/7097)). This changes the default outbound request for locked objects but adds no persisted format.
 - Multipart source objects stay on the multipart transport even when their checksum record is a whole-object checksum, so objects above the single-PUT limit remain replicable ([#7047](https://github.com/rustfs/rustfs/pull/7047)).
 - Targets that mint their own version IDs now use a per-target version ledger for tag, retention, legal-hold, and permanent-delete mutations; ambiguous pre-ledger matches fail with backoff instead of guessing ([#7368](https://github.com/rustfs/rustfs/pull/7368)). This adds dual-prefixed internal metadata keys that older readers ignore.
 - Single-part source checksums are forwarded as `x-amz-checksum-*` headers instead of user metadata, so the replica preserves checksum responses ([#7313](https://github.com/rustfs/rustfs/pull/7313)). This changes the default outbound headers for checksummed objects.
 - Site-replication outage recovery now uses a bounded 30-second retry drain plus the 600-second full reconciliation pass, persists destructive liabilities before local deletion, and fences replay settlement and peer edits ([#7148](https://github.com/rustfs/rustfs/pull/7148)). Persisted additions are optional and ignored by older readers.
 - IAM snapshot/deletion replay, target-assigned delete-marker purges, timestamp ordering, and best-effort peer broadcast now close the control-plane gaps found by the R6 review ([#7195](https://github.com/rustfs/rustfs/pull/7195)).
 - Upgrade and rollback: upgrade every node in one site consecutively and verify reconciliation before moving to the next site; do not intentionally run a site mixed-version. Target-version ledger keys are harmless on rollback, although old code cannot use their routing. Before rolling back past [#7307](https://github.com/rustfs/rustfs/pull/7307), drain or repair every pending version purge: older code can free a retained version's data directory before its remote purge is acknowledged. See `docs/operations/site-replication-operations.md`.
 ### Security
 - **Presigned URLs honour only signed headers** (GHSA-g8w9-qw9q-fghr): a SigV4 presigned request that carries an `x-amz-*` request header not listed in `X-Amz-SignedHeaders` is now rejected with `403 AccessDenied` ("There were headers present in the request which were not signed"), matching AWS S3. Previously the holder of a presigned `PutObject` URL could add unsigned `x-amz-tagging`, `x-amz-storage-class`, `x-amz-website-redirect-location`, ACL, metadata, Object Lock or SSE headers and have them applied. Presigners that intend a property must set it before signing so the SDK lists the header in `SignedHeaders`; `x-amz-cf-id` (CloudFront) remains tolerated unsigned. Header-signed SigV4 and SigV2 requests are unchanged.
@@ -211,10 +211,7 @@ For developers who want to build RustFS Docker images from source with multi-arc
 ```bash
 # Build multi-architecture images locally
-./docker-buildx.sh
+./docker-buildx.sh --build-arg RELEASE=latest
 # Build a single-platform image locally
 ./docker-buildx.sh -p linux/amd64
 # Build and push to registry
 ./docker-buildx.sh --push
@@ -150,10 +150,7 @@ docker compose -f docker-compose-simple.yml up -d
 ```bash
 # 在本地构建多架构镜像
-./docker-buildx.sh
+./docker-buildx.sh --build-arg RELEASE=latest
 # 在本地构建单平台镜像
 ./docker-buildx.sh -p linux/amd64
 # 构建并推送到仓库
 ./docker-buildx.sh --push
@@ -145,6 +145,38 @@ pub fn consume_verified_mrf_repair_events(anchors: &mut Vec<MrfDurableRepairAnch
    before.saturating_sub(anchors.len())
 }
 /// Consume recorded verified repairs for one bucket without draining
 /// unrelated or still-unmatched proofs. If the caller crashes before
 /// persisting the retained anchor set, the proof may be replayed by repair
 /// instead of silently deleting the old responsibility.
 pub fn consume_recorded_verified_mrf_repair_events_for(bucket: &str, anchors: &mut Vec<MrfDurableRepairAnchor>) -> usize {
    let Some(registry) = MRF_VERIFIED_REPAIR_EVENTS.get() else {
        return 0;
    };
    let Ok(mut events) = registry.lock() else {
        return 0;
    };
    let before = anchors.len();
    let mut retained = std::collections::VecDeque::with_capacity(events.len());
    while let Some(event) = events.pop_front() {
        if event.bucket.as_ref() != bucket {
            retained.push_back(event);
            continue;
        }
        let mut matched = false;
        anchors.retain(|anchor| {
            let proven = anchor.is_proven_by(&event);
            matched |= proven;
            !proven
        });
        if !matched {
            retained.push_back(event);
        }
    }
    *events = retained;
    before.saturating_sub(anchors.len())
 }
 #[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
 pub struct MrfScope {
    pub pool_index: u32,
@@ -821,6 +853,71 @@ mod tests {
        assert!(retained.is_empty());
    }
    #[test]
    fn recorded_verified_repair_consumer_retains_unmatched_proofs() {
        let bucket = Arc::<str>::from(format!("recorded-proof-{}", Uuid::new_v4()));
        let other_bucket = Arc::<str>::from(format!("recorded-proof-other-{}", Uuid::new_v4()));
        let incarnation = Uuid::new_v4();
        let lease = MrfIngressLease::new(21);
        let retained_anchor = MrfDurableRepairAnchor {
            kind: MrfKind::PartialWrite,
            bucket: bucket.clone(),
            object: Arc::from("retained"),
            version_id: Some([7; 16]),
            scope: Some(MrfScope {
                pool_index: 1,
                set_index: 2,
            }),
            lease,
            bucket_incarnation_id: incarnation,
        };
        let waiting_anchor = MrfDurableRepairAnchor {
            object: Arc::from("waiting"),
            lease: MrfIngressLease::new(22),
            ..retained_anchor.clone()
        };
        let matched_event = MrfVerifiedRepairEvent {
            kind: retained_anchor.kind,
            bucket: bucket.clone(),
            object: retained_anchor.object.clone(),
            version_id: retained_anchor.version_id,
            scope: retained_anchor.scope,
            lease: Some(retained_anchor.lease),
            bucket_incarnation_id: retained_anchor.bucket_incarnation_id,
            disposition: MrfVerifiedRepairDisposition::Repaired,
        };
        let same_bucket_unmatched = MrfVerifiedRepairEvent {
            object: Arc::from("future"),
            lease: Some(MrfIngressLease::new(23)),
            ..matched_event.clone()
        };
        let other_bucket_event = MrfVerifiedRepairEvent {
            bucket: other_bucket.clone(),
            ..matched_event.clone()
        };
        note_mrf_verified_repair(matched_event);
        note_mrf_verified_repair(same_bucket_unmatched.clone());
        note_mrf_verified_repair(other_bucket_event.clone());
        let mut anchors = vec![retained_anchor, waiting_anchor.clone()];
        assert_eq!(consume_recorded_verified_mrf_repair_events_for(&bucket, &mut anchors), 1);
        assert_eq!(anchors, vec![waiting_anchor]);
        let remaining_bucket_events = take_mrf_verified_repair_events_for(&bucket);
        assert_eq!(
            remaining_bucket_events,
            vec![same_bucket_unmatched],
            "same-bucket proofs without a retained anchor must remain available"
        );
        let remaining_other_events = take_mrf_verified_repair_events_for(&other_bucket);
        assert_eq!(
            remaining_other_events,
            vec![other_bucket_event],
            "proofs for other buckets must not be drained by this consumer"
        );
    }
    #[tokio::test]
    async fn try_send_delivers_and_respects_capacity() {
        let mut receiver = init_mrf_channel().expect("first initialization should succeed");
@@ -44,6 +44,7 @@ use sha2::{Digest, Sha256};
 use std::collections::{BTreeMap, BTreeSet};
 use std::error::Error;
 use std::path::{Path, PathBuf};
 use tokio::time::{Duration, Instant, sleep};
 use tracing::info;
 use uuid::Uuid;
 use walkdir::WalkDir;
@@ -374,6 +375,33 @@ pub(crate) fn census_object_version_on_disk(
    })
 }
 /// Wait for the background PUT tail to commit every physical part on one disk.
 /// Invalid metadata remains an immediate error instead of a retryable absence.
 pub(crate) async fn wait_for_complete_physical_shard_on_disk(
    disk: &Path,
    bucket: &str,
    key: &str,
    version_id: Option<&str>,
    timeout: Duration,
 ) -> ChaosResult<VersionShardCensus> {
    let deadline = Instant::now() + timeout;
    loop {
        let census = census_object_version_on_disk(disk, bucket, key, version_id)?;
        if census.is_complete() && !census.expected_part_numbers.is_empty() {
            return Ok(census);
        }
        let remaining = deadline.saturating_duration_since(Instant::now());
        if remaining.is_zero() {
            return Err(format!(
                "physical shard for {bucket}/{key}@{version_id:?} on {} did not become complete within {timeout:?}: {census:?}",
                disk.display()
            )
            .into());
        }
        sleep(remaining.min(Duration::from_millis(50))).await;
    }
 }
 /// `POST` a signed (SigV4, service `s3`) admin request without relying on the
 /// external `awscurl` binary. Mirrors the admin heal calls used by the heal
 /// regression suite.
@@ -451,4 +479,30 @@ mod tests {
        assert!(expected.matches_manifest(&expected));
        assert!(!changed.matches_manifest(&expected));
    }
    #[tokio::test]
    async fn physical_shard_readiness_fails_closed_with_last_census() {
        let disk = tempfile::tempdir().expect("temporary disk");
        let error = wait_for_complete_physical_shard_on_disk(disk.path(), "bucket", "missing", None, Duration::ZERO)
            .await
            .expect_err("missing physical shards must fail the baseline gate");
        assert!(error.to_string().contains("has_xl_meta: false"));
        assert!(error.to_string().contains("bucket/missing"));
    }
    #[tokio::test]
    async fn physical_shard_readiness_does_not_retry_invalid_metadata() {
        let disk = tempfile::tempdir().expect("temporary disk");
        let object = disk.path().join("bucket").join("corrupt");
        std::fs::create_dir_all(&object).expect("object directory");
        std::fs::write(object.join("xl.meta"), b"invalid metadata").expect("corrupt metadata fixture");
        let error = tokio::time::timeout(
            Duration::from_secs(1),
            wait_for_complete_physical_shard_on_disk(disk.path(), "bucket", "corrupt", None, Duration::from_secs(30)),
        )
        .await
        .expect("corrupt metadata must fail immediately")
        .expect_err("invalid metadata must not be accepted as a complete baseline");
        assert!(!error.to_string().contains("did not become complete"));
    }
 }
@@ -34,7 +34,7 @@ use serde_json;
 use std::ffi::OsStr;
 use std::fs as stdfs;
 use std::io::ErrorKind;
-use std::net::SocketAddr;
+use std::net::{Ipv6Addr, SocketAddr};
 use std::path::{Path, PathBuf};
 use std::process::{Child, Command, Stdio};
 use std::sync::Once;
@@ -1468,7 +1468,9 @@ impl RustFSTestClusterEnvironment {
            return Err(format!("a volume proxy is already configured for node {node_idx}").into());
        }
        let target = self.nodes[node_idx].address.parse::<SocketAddr>()?;
-        let proxy = crate::fault_proxy::FaultProxy::start(target).await?;
+        // Endpoint locality requires the server port to match. IPv6 loopback
        // keeps that port while leaving the direct IPv4 S3 listener available.
        let proxy = crate::fault_proxy::FaultProxy::start_on((Ipv6Addr::LOCALHOST, target.port()).into(), target).await?;
        self.volume_proxy_addresses[node_idx] = Some(proxy.local_addr());
        Ok(proxy)
    }
@@ -2373,7 +2375,7 @@ mod tests {
    #[tokio::test]
    async fn volume_proxy_rewrites_cluster_volume_endpoint() {
-        let mut env = RustFSTestClusterEnvironment::new(1)
+        let mut env = RustFSTestClusterEnvironment::with_topology(ClusterTopology::single_pool_multidrive(2, 2))
            .await
            .expect("cluster environment should allocate a node");
        let direct = env.nodes[0].address.clone();
@@ -2388,6 +2390,24 @@ mod tests {
        assert!(!volumes.contains(&direct), "volumes must not retain the direct address: {volumes}");
        proxy.shutdown().await;
        for node in &env.nodes {
            let local_port = node.address.parse::<SocketAddr>().expect("node address").port();
            let local_paths = volumes
                .split_whitespace()
                .filter_map(|volume| {
                    let endpoint = reqwest::Url::parse(volume).expect("volume endpoint");
                    rustfs_utils::is_local_host(
                        endpoint.host().expect("volume endpoint host"),
                        endpoint.port().expect("volume endpoint port"),
                        local_port,
                    )
                    .expect("endpoint locality")
                    .then(|| endpoint.path().to_string())
                })
                .collect::<Vec<_>>();
            assert_eq!(local_paths, node.data_dirs, "the proxy must preserve local disk ownership");
        }
    }
    #[test]
@@ -59,6 +59,9 @@ const POOL_META_V3_ENV: [(&str, &str); 2] = [
 #[derive(Clone, Copy, Debug)]
 pub(crate) enum DistLayout {
    /// 3 nodes × 4 drives, one erasure pool. With `EC:4` this is the
    /// release-evidence EC8+4 geometry.
    ThreeByFourEc84,
    /// 4 nodes × 4 drives, one erasure pool spanning every endpoint.
    FourByFour,
    /// 4 nodes × 1 drive, one erasure pool (minimum 4-node 4-disk layout).
@@ -94,6 +97,7 @@ impl DistCluster {
    pub async fn new_stopped_with_env(layout: DistLayout, extra_env: &[(&str, &str)]) -> TestResult<Self> {
        let topology = match layout {
            DistLayout::ThreeByFourEc84 => ClusterTopology::single_pool_multidrive(3, DRIVES_PER_NODE),
            DistLayout::FourByFour => ClusterTopology::single_pool_multidrive(NODE_COUNT, DRIVES_PER_NODE),
            DistLayout::FourNodeFourDisk => ClusterTopology::single_pool(NODE_COUNT),
            DistLayout::SingleNodeFourDrive => ClusterTopology::per_node_pools(DRIVES_PER_NODE, vec![vec![0]]),
@@ -101,7 +105,7 @@ impl DistCluster {
        let mut cluster = RustFSTestClusterEnvironment::with_topology(topology).await?;
        let pool_storage_roots = match layout {
            DistLayout::SingleNodeFourDrive => Some(configured_pool_storage_roots()?),
-            DistLayout::FourByFour | DistLayout::FourNodeFourDisk => None,
+            DistLayout::ThreeByFourEc84 | DistLayout::FourByFour | DistLayout::FourNodeFourDisk => None,
        };
        let mut owned_pool_dirs = Vec::new();
        if let Some(roots) = pool_storage_roots.as_deref() {
@@ -0,0 +1,250 @@
 // Copyright 2026 RustFS Team
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
 //
 //     http://www.apache.org/licenses/LICENSE-2.0
 //
 // Unless required by applicable law or agreed to in writing, software
 // distributed under the License is distributed on an "AS IS" BASIS,
 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 // See the License for the specific language governing permissions and
 // limitations under the License.
 use super::harness::{DistCluster, DistLayout, TestResult, assert_inventory, payload_for, put_object, unique_bucket, wait_until};
 use crate::chaos::{
    VersionShardCensus, census_object_version_on_disk, sha256_hex, signed_admin_post, wait_for_complete_physical_shard_on_disk,
 };
 use crate::common::{init_logging, rustfs_binary_path};
 use crate::scanner_heal_evidence::{EvidenceTopology, RestartObservation, ScannerHealEvidenceCase, restart_evidence_run};
 use aws_sdk_s3::Client;
 use aws_sdk_s3::primitives::ByteStream;
 use std::collections::{BTreeMap, HashSet};
 use std::path::{Path, PathBuf};
 use std::time::Duration;
 const EC84_NODE_COUNT: usize = 3;
 const EC84_DRIVES_PER_NODE: usize = 4;
 const EC84_DATA_BLOCKS: usize = 8;
 const EC84_PARITY_BLOCKS: usize = 4;
 #[derive(Clone)]
 struct ExpectedShard {
    key: String,
    body: Vec<u8>,
    baseline: VersionShardCensus,
 }
 fn assert_ec84_geometry(census: &VersionShardCensus, key: &str) -> TestResult {
    if census.data_blocks != Some(EC84_DATA_BLOCKS) || census.parity_blocks != Some(EC84_PARITY_BLOCKS) {
        return Err(format!("object {key} did not use EC8+4 geometry: {census:?}").into());
    }
    let erasure_index = census
        .erasure_index
        .ok_or_else(|| format!("object {key} did not record an erasure index: {census:?}"))?;
    if !(1..=EC84_DATA_BLOCKS + EC84_PARITY_BLOCKS).contains(&erasure_index) {
        return Err(format!("object {key} has out-of-range erasure index {erasure_index}: {census:?}").into());
    }
    if !census.is_complete() || census.expected_part_numbers.is_empty() {
        return Err(format!("object {key} does not have complete physical shard evidence: {census:?}").into());
    }
    Ok(())
 }
 fn assert_replaced_drive_empty(drive: &Path, bucket: &str, keys: &[String]) -> TestResult {
    for key in keys {
        let census = census_object_version_on_disk(drive, bucket, key, None)?;
        if census.has_xl_meta {
            return Err(format!("replacement drive unexpectedly retained {bucket}/{key}: {census:?}").into());
        }
    }
    Ok(())
 }
 async fn put_large_inventory(client: &Client, bucket: &str) -> TestResult<Vec<ExpectedShard>> {
    let mut expected = Vec::new();
    for index in 0..4 {
        let key = format!("ec84/prefix-{}/object-{index:04}.bin", index % 2);
        let body = payload_for(&key, 10 * 1024 * 1024);
        put_object(client, bucket, &key, body.clone()).await?;
        expected.push(ExpectedShard {
            key,
            body,
            baseline: VersionShardCensus {
                version_id: None,
                has_xl_meta: false,
                data_dir: None,
                erasure_index: None,
                data_blocks: None,
                parity_blocks: None,
                expected_part_numbers: Default::default(),
                present_part_fingerprints: Default::default(),
                inline_data_fingerprint: None,
            },
        });
    }
    Ok(expected)
 }
 #[tokio::test]
 async fn three_node_four_drive_ec8_4_root_heal_rebuilds_replaced_drive_after_restart() -> TestResult {
    init_logging();
    let server_binary = rustfs_binary_path();
    let evidence_run = restart_evidence_run(
        &server_binary,
        ScannerHealEvidenceCase {
            id: "ec84-target-drive-restart",
            oracle: "ec84-target-drive-restart.json",
            evidence: "process-restart",
            unclean_shutdown_marker: false,
            topology: EvidenceTopology::new(3, 4),
            storage_class_standard: Some("EC:4"),
            erasure_set_drive_count: Some("12"),
        },
    )?;
    let mut dist = DistCluster::start_with_env(
        DistLayout::ThreeByFourEc84,
        &[
            ("RUSTFS_STORAGE_CLASS_STANDARD", "EC:4"),
            ("RUSTFS_HEAL_ENABLED", "true"),
            ("RUSTFS_HEAL_AUTO_HEAL_ENABLE", "false"),
            ("RUSTFS_HEAL_MRF_ENABLE", "false"),
            ("RUSTFS_SCANNER_ENABLED", "false"),
        ],
    )
    .await?;
    assert_eq!(dist.cluster.nodes.len(), EC84_NODE_COUNT);
    assert_eq!(dist.cluster.topology.drives_per_node, EC84_DRIVES_PER_NODE);
    let bucket = unique_bucket("healec84");
    dist.create_bucket(&bucket).await?;
    let writer = dist.client(0)?;
    let mut expected = put_large_inventory(&writer, &bucket).await?;
    let replaced_node = 1;
    let replaced_drive_index = 2;
    let replaced_drive = PathBuf::from(&dist.cluster.nodes[replaced_node].data_dirs[replaced_drive_index]);
    for item in &mut expected {
        item.baseline =
            wait_for_complete_physical_shard_on_disk(&replaced_drive, &bucket, &item.key, None, Duration::from_secs(10)).await?;
        assert_ec84_geometry(&item.baseline, &item.key)?;
    }
    let format_path = replaced_drive.join(".rustfs.sys").join("format.json");
    let format_json = std::fs::read(&format_path)?;
    let pid_before = dist.cluster.nodes[replaced_node]
        .process
        .as_ref()
        .ok_or("target process is absent")?
        .id();
    dist.cluster.stop_node_gracefully(replaced_node).await?;
    let unclean_shutdown_marker = Path::new(&dist.cluster.nodes[replaced_node].data_dir)
        .join(".rustfs.sys")
        .join("unclean-shutdown")
        .is_file();
    assert!(!unclean_shutdown_marker, "graceful target shutdown must remove its unclean marker");
    let retired_drive = PathBuf::from(format!("{}.retired", replaced_drive.display()));
    std::fs::rename(&replaced_drive, &retired_drive)?;
    std::fs::create_dir_all(format_path.parent().ok_or("replacement format path has no parent")?)?;
    std::fs::write(&format_path, format_json)?;
    assert_replaced_drive_empty(
        &replaced_drive,
        &bucket,
        &expected.iter().map(|item| item.key.clone()).collect::<Vec<_>>(),
    )?;
    let outage_key = "ec84/written-while-node-restarting.bin";
    let outage_body = payload_for(outage_key, 10 * 1024 * 1024);
    writer
        .put_object()
        .bucket(&bucket)
        .key(outage_key)
        .body(ByteStream::from(outage_body.clone()))
        .send()
        .await?;
    dist.cluster.start_node(replaced_node).await?;
    let heal_body =
        r#"{"recursive":true,"dryRun":false,"remove":false,"recreate":true,"scanMode":2,"updateParity":false,"nolock":false}"#;
    let heal_url = format!("{}/rustfs/admin/v3/heal/{bucket}?forceStart=true", dist.cluster.nodes[0].url);
    signed_admin_post(&heal_url, Some(heal_body), &dist.cluster.access_key, &dist.cluster.secret_key).await?;
    wait_until(
        Duration::from_secs(120),
        || async {
            for item in &expected {
                let current = census_object_version_on_disk(&replaced_drive, &bucket, &item.key, None)?;
                if !current.matches_manifest(&item.baseline) {
                    return Ok(false);
                }
            }
            let outage = census_object_version_on_disk(&replaced_drive, &bucket, outage_key, None)?;
            Ok(outage.is_complete()
                && outage.data_blocks == Some(EC84_DATA_BLOCKS)
                && outage.parity_blocks == Some(EC84_PARITY_BLOCKS))
        },
        "EC8+4 replacement drive rebuilt baseline and outage shards",
    )
    .await?;
    let inventory = expected
        .iter()
        .map(|item| (item.key.clone(), item.body.clone()))
        .chain(std::iter::once((outage_key.to_string(), outage_body.clone())))
        .collect::<BTreeMap<_, _>>();
    let expected_keys = inventory.keys().cloned().collect::<HashSet<_>>();
    let mut node_listings = Vec::new();
    for node_index in 0..dist.cluster.nodes.len() {
        let client = dist.client(node_index)?;
        assert_inventory(&client, &bucket, &inventory).await?;
        let listing = client.list_objects_v2().bucket(&bucket).send().await?;
        let observed = listing
            .contents()
            .iter()
            .filter_map(|object| object.key().map(str::to_owned))
            .collect::<HashSet<_>>();
        assert_eq!(observed, expected_keys, "node {node_index} listing diverged after EC8+4 heal");
        let mut keys = observed.into_iter().collect::<Vec<_>>();
        keys.sort();
        node_listings.push(keys);
    }
    if let Some(evidence_run) = evidence_run {
        let target_client = dist.client(replaced_node)?;
        let mut objects = Vec::with_capacity(inventory.len());
        for (key, body) in &inventory {
            let response = target_client.get_object().bucket(&bucket).key(key).send().await?;
            let actual = response.body.collect().await?.into_bytes();
            assert_eq!(actual.as_ref(), body.as_slice(), "object body changed for {key}");
            let physical = census_object_version_on_disk(&replaced_drive, &bucket, key, None)?;
            assert_ec84_geometry(&physical, key)?;
            let baseline = expected.iter().find(|item| item.key == *key).map(|item| &item.baseline);
            objects.push(serde_json::json!({
                "key": key, "version_id": null,
                "expected_bytes": body.len(), "actual_bytes": actual.len(),
                "expected_sha256": sha256_hex(body), "actual_sha256": sha256_hex(&actual),
                "expected_physical": baseline, "physical": physical,
            }));
        }
        let pid_after = dist.cluster.nodes[replaced_node]
            .process
            .as_ref()
            .ok_or("restarted target is absent")?
            .id();
        evidence_run.write(
            &server_binary,
            RestartObservation {
                nodes: dist.cluster.nodes.len(),
                drives_per_node: dist.cluster.topology.drives_per_node,
                pid_before,
                pid_after,
                unclean_shutdown_marker,
                objects,
                node_listings,
            },
        )?;
    }
    Ok(())
 }
@@ -25,6 +25,7 @@ mod data_integrity_movement_test;
 mod expand_decommission_rebalance_test;
 mod extra_test;
 mod harness;
 mod heal_test;
 mod object_lock_test;
 mod observability_test;
 mod replication_quota_test;
@@ -498,7 +498,9 @@ mod tests {
                .stderr(log)
                .spawn()?,
        );
-        let status = tokio::time::timeout(Duration::from_secs(10), async {
+        // macOS evaluates each fresh binary copy before its capability hook can run.
        let probe_timeout = if cfg!(target_os = "macos") { 60 } else { 10 };
        let status = tokio::time::timeout(Duration::from_secs(probe_timeout), async {
            loop {
                if let Some(status) = child.0.try_wait()? {
                    return Ok::<_, std::io::Error>(status);
@@ -114,7 +114,12 @@ impl FaultProxy {
    /// Bind a listener on `127.0.0.1:0` and start forwarding accepted
    /// connections to `target`. Starts in [`FaultMode::Pass`].
    pub async fn start(target: SocketAddr) -> io::Result<Self> {
-        let listener = TcpListener::bind((Ipv4Addr::LOCALHOST, 0)).await?;
+        Self::start_on((Ipv4Addr::LOCALHOST, 0).into(), target).await
    }
    /// Bind the selected address and forward accepted connections to `target`.
    pub(crate) async fn start_on(address: SocketAddr, target: SocketAddr) -> io::Result<Self> {
        let listener = TcpListener::bind(address).await?;
        let listen_addr = listener.local_addr()?;
        let (mode_tx, mode_rx) = watch::channel(FaultMode::Pass);
@@ -401,11 +401,11 @@ mod tests {
        let mp_view = get_full(&client, multipart_key).await?;
        assert_eq!(mp_view.sha256, sha256_hex(&multipart_body), "degraded baseline multipart body mismatch");
        baseline_degraded.insert(multipart_key.to_string(), mp_view);
-        // Restore the disk so Phase B restarts from a clean, complete disk set.
+        // Stop disk writers before restoring the complete layout reused by Phase B.
        harness.kill_server();
        harness.bring_disk_online(0)?;
        // ---- Phase B: codec streaming (gates opened) ----
        harness.kill_server();
        for (k, v) in codec_env() {
            harness.set_env(k, v);
        }
@@ -497,6 +497,8 @@ mod tests {
        let mp_view = get_full(&client, multipart_key).await?;
        assert_eq!(mp_view.sha256, sha256_hex(&multipart_body), "degraded codec multipart body mismatch");
        codec_degraded.insert(multipart_key.to_string(), mp_view);
        // All server reads are complete; stop disk writers before restoring disk0.
        harness.kill_server();
        harness.bring_disk_online(0)?;
        // A/B under parity reconstruction: codec == legacy, byte-for-byte and
@@ -16,18 +16,20 @@
 #[cfg(test)]
 mod tests {
-    use crate::chaos::{VersionShardCensus, census_object_version_on_disk, sha256_hex, signed_admin_post};
+    use crate::chaos::{
        VersionShardCensus, census_object_version_on_disk, sha256_hex, signed_admin_post,
        wait_for_complete_physical_shard_on_disk,
    };
    use crate::common::{
        FAST_DATA_USAGE_SCANNER_ENV, RustFSTestClusterEnvironment, RustFSTestEnvironment, admin_request, init_logging,
        rustfs_binary_path,
    };
    use crate::scanner_heal_evidence::{EvidenceTopology, RestartObservation, ScannerHealEvidenceCase, restart_evidence_run};
    use crate::storage_api::RUSTFS_META_BUCKET;
    use aws_sdk_s3::primitives::ByteStream;
    use http::Method;
    use sha2::{Digest, Sha256};
    use std::collections::HashSet;
    use std::error::Error;
    use std::io::{Read, Write};
    use std::net::SocketAddr;
    use std::path::{Path, PathBuf};
    use std::process::Command;
@@ -39,34 +41,14 @@ mod tests {
    const POOL_METADATA_OBJECT: &str = "pool.bin";
    #[derive(serde::Deserialize)]
    struct EvidenceBuild {
        sha256: String,
    }
    #[derive(serde::Deserialize)]
    struct RestartEvidenceRun {
        schema: u32,
        run_id: String,
        source_revision: String,
        test_build: serde_json::Value,
        binary: EvidenceBuild,
        test_binary: EvidenceBuild,
    }
    #[derive(Clone, Copy)]
    struct ScannerHealEvidenceCase {
        id: &'static str,
        oracle: &'static str,
        evidence: &'static str,
        unclean_shutdown_marker: bool,
    }
    const BACKGROUND_TARGET_RESTART_EVIDENCE: ScannerHealEvidenceCase = ScannerHealEvidenceCase {
        id: "background-target-restart",
        oracle: "background-target-restart.json",
        evidence: "process-restart",
        unclean_shutdown_marker: false,
        topology: EvidenceTopology::new(4, 1),
        storage_class_standard: None,
        erasure_set_drive_count: None,
    };
    const BACKGROUND_TARGET_CRASH_EVIDENCE: ScannerHealEvidenceCase = ScannerHealEvidenceCase {
@@ -74,82 +56,30 @@ mod tests {
        oracle: "background-target-crash.json",
        evidence: "process-crash-restart",
        unclean_shutdown_marker: true,
        topology: EvidenceTopology::new(4, 1),
        storage_class_standard: None,
        erasure_set_drive_count: None,
    };
-    struct RestartEvidenceContext {
+    const BACKGROUND_TARGET_RESTART_EC84_EVIDENCE: ScannerHealEvidenceCase = ScannerHealEvidenceCase {
-        directory: PathBuf,
+        id: "background-target-restart-ec8-4",
-        run: RestartEvidenceRun,
+        oracle: "background-target-restart-ec8-4.json",
-        case: ScannerHealEvidenceCase,
+        evidence: "process-restart",
-    }
+        unclean_shutdown_marker: false,
        topology: EvidenceTopology::new(3, 4),
        storage_class_standard: Some("EC:4"),
        erasure_set_drive_count: Some("12"),
    };
-    fn file_sha256(path: &Path) -> Result<String, Box<dyn Error + Send + Sync>> {
+    const BACKGROUND_TARGET_CRASH_EC84_EVIDENCE: ScannerHealEvidenceCase = ScannerHealEvidenceCase {
-        let mut file = std::fs::File::open(path)?;
+        id: "background-target-crash-ec8-4",
-        let mut digest = Sha256::new();
+        oracle: "background-target-crash-ec8-4.json",
-        let mut buffer = [0_u8; 64 * 1024];
+        evidence: "process-crash-restart",
-        loop {
+        unclean_shutdown_marker: true,
-            let read = file.read(&mut buffer)?;
+        topology: EvidenceTopology::new(3, 4),
-            if read == 0 {
+        storage_class_standard: Some("EC:4"),
-                break;
+        erasure_set_drive_count: Some("12"),
-            }
+    };
            digest.update(&buffer[..read]);
        }
        Ok(digest.finalize().iter().map(|byte| format!("{byte:02x}")).collect())
    }
    fn restart_evidence_run(
        binary: &Path,
        case: ScannerHealEvidenceCase,
    ) -> Result<Option<RestartEvidenceContext>, Box<dyn Error + Send + Sync>> {
        let Some(directory) = std::env::var_os("RUSTFS_SCANNER_HEAL_RUN_DIR") else {
            return Ok(None);
        };
        if case.id.is_empty()
            || case.oracle.is_empty()
            || !case.oracle.ends_with(".json")
            || case.oracle.contains('/')
            || case.oracle.contains('\\')
            || case.oracle.contains("..")
            || !matches!(case.evidence, "process-restart" | "process-crash-restart")
            || (case.evidence == "process-crash-restart") != case.unclean_shutdown_marker
        {
            return Err("invalid scanner/heal evidence case".into());
        }
        let directory = PathBuf::from(directory);
        let receipt = directory.join("run.json");
        if receipt.metadata()?.len() > 1024 * 1024 {
            return Err("oversized scanner/heal execution receipt".into());
        }
        let run: RestartEvidenceRun = serde_json::from_slice(&std::fs::read(receipt)?)?;
        if run.schema != 1 || run.run_id.len() != 32 || run.source_revision.len() != 40 {
            return Err("invalid scanner/heal execution identity".into());
        }
        let built = compiled_test_identity();
        for key in ["source_revision", "dirty", "lock_blob", "features"] {
            assert_eq!(built[key], run.test_build[key], "compiled test identity differs for {key}");
        }
        assert_eq!(file_sha256(binary)?, run.binary.sha256, "server binary must match the run receipt");
        assert_eq!(
            file_sha256(&std::env::current_exe()?)?,
            run.test_binary.sha256,
            "test executable must match the run receipt"
        );
        if directory.join(case.oracle).exists() {
            return Err("scanner/heal oracle already exists; create a new execution receipt".into());
        }
        Ok(Some(RestartEvidenceContext { directory, run, case }))
    }
    fn compiled_test_identity() -> serde_json::Value {
        serde_json::json!({
            "source_revision": env!("RUSTFS_E2E_BUILD_COMMIT"),
            "dirty": env!("RUSTFS_E2E_BUILD_DIRTY") != "false",
            "lock_blob": env!("RUSTFS_E2E_BUILD_LOCK"),
            "features": env!("RUSTFS_E2E_BUILD_FEATURES"),
            "target": env!("RUSTFS_E2E_BUILD_TARGET"),
            "profile": env!("RUSTFS_E2E_BUILD_PROFILE"),
            "rustflags_hex": env!("RUSTFS_E2E_BUILD_RUSTFLAGS_HEX"),
        })
    }
    struct TcpPortBlackhole {
        port: u16,
@@ -332,11 +262,10 @@ mod tests {
    // Healing may rewrite non-identity bookkeeping in xl.meta. The census
    // therefore compares the canonical selected metadata fields plus every
-    // physical shard, while the payload seed makes object mix-ups observable.
+    // physical shard.
    #[derive(Debug)]
    struct PhysicalObjectManifest {
        key: String,
        payload_seed: u8,
        shard_census: VersionShardCensus,
    }
@@ -456,6 +385,79 @@ mod tests {
        )
    }
    async fn wait_for_admin_cluster_start_log(
        log_path: &Path,
        client_token: &str,
        deadline: Instant,
    ) -> Result<(), Box<dyn Error + Send + Sync>> {
        loop {
            let coordinator_log = std::fs::read_to_string(log_path)?;
            if coordinator_log
                .lines()
                .filter_map(|line| serde_json::from_str::<serde_json::Value>(line).ok())
                .any(|event| {
                    event["event"] == "heal_task_state"
                        && event["task_id"] == client_token
                        && event["heal_type"] == "cluster"
                        && event["state"] == "started"
                })
            {
                return Ok(());
            }
            if Instant::now() >= deadline {
                return Err(
                    format!("node 0 must have started the exact admin task before interruption: task_id={client_token}").into(),
                );
            }
            sleep(Duration::from_millis(10)).await;
        }
    }
    #[tokio::test]
    async fn test_admin_cluster_start_log_waits_for_exact_delayed_event() -> Result<(), Box<dyn Error + Send + Sync>> {
        use std::io::Write;
        let mut log = tempfile::NamedTempFile::new()?;
        for (task_id, heal_type, state) in [
            ("other-task", "cluster", "started"),
            ("admin-task", "object", "started"),
            ("admin-task", "cluster", "completed"),
        ] {
            writeln!(
                log,
                "{}",
                serde_json::json!({"event": "heal_task_state", "task_id": task_id, "heal_type": heal_type, "state": state})
            )?;
        }
        let log_path = log.path().to_path_buf();
        let started = wait_for_admin_cluster_start_log(&log_path, "admin-task", Instant::now() + Duration::from_secs(1));
        tokio::pin!(started);
        // Poll the reader before publishing the start event, without depending
        // on scheduling or a fixed writer delay to reproduce log visibility.
        tokio::select! {
            biased;
            result = &mut started => panic!("unrelated events must leave the exact start pending: {result:?}"),
            _ = std::future::ready(()) => {}
        }
        writeln!(
            log,
            "{}",
            serde_json::json!({"event": "heal_task_state", "task_id": "admin-task", "heal_type": "cluster", "state": "started"})
        )?;
        started.await?;
        Ok(())
    }
    #[tokio::test]
    async fn test_admin_cluster_start_log_respects_existing_deadline() -> Result<(), Box<dyn Error + Send + Sync>> {
        let log = tempfile::NamedTempFile::new()?;
        let error = wait_for_admin_cluster_start_log(log.path(), "admin-task", Instant::now())
            .await
            .expect_err("missing exact start must fail at the supplied deadline");
        assert!(error.to_string().contains("task_id=admin-task"), "{error}");
        Ok(())
    }
    fn cluster_heal_is_idle(status: &serde_json::Value) -> bool {
        let operations = &status["healOperations"];
        status["clusterStatusComplete"] == serde_json::Value::Bool(true)
@@ -892,7 +894,10 @@ mod tests {
        cluster: &RustFSTestClusterEnvironment,
        previous_cycle_end: u64,
    ) -> Result<u64, Box<dyn Error + Send + Sync>> {
-        let deadline = Instant::now() + Duration::from_secs(60);
+        let started = Instant::now();
        let mut deadline = started + Duration::from_secs(60);
        let catch_up_deadline = deadline + Duration::from_secs(300);
        let mut catch_up_wait_observed = false;
        loop {
            let mut latest_cycle_end = 0;
            let mut versions_observed = false;
@@ -920,24 +925,61 @@ mod tests {
                let versions_scanned = metrics["versions_scanned"]
                    .as_u64()
                    .ok_or("scanner status is missing its version-coverage counter")?;
-                latest_cycle_end = latest_cycle_end.max(cycle_end);
+                let cycle_result = metrics["last_cycle_result"]
                    .as_str()
                    .ok_or("scanner status is missing its cycle result")?;
                if cycle_result == "success" {
                    latest_cycle_end = latest_cycle_end.max(cycle_end);
                }
                versions_observed |= versions_scanned > 0;
                let backlog = &status["pause_backlog"];
                if !catch_up_wait_observed
                    && backlog["persistence_state"].as_str() == Some("healthy")
                    && backlog["durable"].as_bool() == Some(true)
                    && backlog["phase"].as_str() == Some("catching_up")
                    && backlog["rate_limited"].as_bool() == Some(true)
                    && backlog["retry_exhausted"].as_bool() == Some(false)
                {
                    let next_attempt = backlog["next_attempt_at_unix_secs"]
                        .as_u64()
                        .ok_or("rate-limited scanner backlog is missing its next attempt")?;
                    let interval = backlog["thresholds"]["catch_up_min_interval_seconds"]
                        .as_u64()
                        .ok_or("rate-limited scanner backlog is missing its catch-up interval")?;
                    let now = std::time::SystemTime::now().duration_since(std::time::UNIX_EPOCH)?.as_secs();
                    let remaining = next_attempt.saturating_sub(now);
                    if remaining > 0 {
                        if interval > 300 || remaining > interval {
                            return Err(
                                format!("scanner catch-up schedule exceeds the bounded recovery budget: {backlog}").into()
                            );
                        }
                        // The durable catch-up interval overrides SCANNER_CYCLE=1.
                        // Honor one observed retry without restarting the deadline on every poll.
                        deadline = deadline
                            .max(Instant::now() + Duration::from_secs(remaining + 60))
                            .min(catch_up_deadline);
                        catch_up_wait_observed = true;
                    }
                }
                observations.push(format!(
-                    "node{node_index}: end={cycle_end}, versions={versions_scanned}, cycle={}, active={}, leader={}, result={}",
+                    "node{node_index}: end={cycle_end}, versions={versions_scanned}, cycle={}, active={}, leader={}, result={}, backlog={}",
                    metrics["current_cycle"],
                    metrics["current_cycle_active"],
                    metrics["leader_lock_state"],
                    metrics["last_cycle_result"],
                    backlog,
                ));
            }
-            // The coordinator records cycle completion, but remote workers
+            // Only a successful coordinator cycle counts as completion; deferred
-            // record scanned versions. Both witnesses need not share a node.
+            // and superseded attempts also advance its end timestamp. Remote
            // workers record version coverage, so the witnesses can span nodes.
            if latest_cycle_end > previous_cycle_end && versions_observed {
                return Ok(latest_cycle_end);
            }
            if Instant::now() >= deadline {
                return Err(format!(
-                    "enabled scanner did not complete an object-scanning cycle after {previous_cycle_end}: {observations:?}"
+                    "enabled scanner did not complete a successful object-scanning cycle after {previous_cycle_end}: {observations:?}"
                )
                .into());
            }
@@ -957,7 +999,7 @@ mod tests {
    async fn test_cluster_root_heal_recovers_remote_shards_after_background_target_restart()
    -> Result<(), Box<dyn Error + Send + Sync>> {
        timeout(
-            Duration::from_secs(420),
+            Duration::from_secs(720),
            run_cluster_root_heal_interruption(InterruptionScenario::BackgroundTargetRestart),
        )
        .await?
@@ -967,17 +1009,37 @@ mod tests {
    async fn test_cluster_root_heal_recovers_remote_shards_after_background_target_crash()
    -> Result<(), Box<dyn Error + Send + Sync>> {
        timeout(
-            Duration::from_secs(420),
+            Duration::from_secs(720),
            run_cluster_root_heal_interruption(InterruptionScenario::BackgroundTargetCrash),
        )
        .await?
    }
    #[tokio::test(flavor = "multi_thread")]
    async fn test_cluster_root_heal_recovers_ec84_shards_after_background_target_restart()
    -> Result<(), Box<dyn Error + Send + Sync>> {
        timeout(
            Duration::from_secs(720),
            run_cluster_root_heal_interruption(InterruptionScenario::BackgroundTargetRestartEc84),
        )
        .await?
    }
    #[tokio::test(flavor = "multi_thread")]
    async fn test_cluster_root_heal_recovers_ec84_shards_after_background_target_crash()
    -> Result<(), Box<dyn Error + Send + Sync>> {
        timeout(
            Duration::from_secs(720),
            run_cluster_root_heal_interruption(InterruptionScenario::BackgroundTargetCrashEc84),
        )
        .await?
    }
    #[tokio::test(flavor = "multi_thread")]
    async fn test_cluster_root_heal_recovers_remote_shards_after_coordinator_restart() -> Result<(), Box<dyn Error + Send + Sync>>
    {
        timeout(
-            Duration::from_secs(420),
+            Duration::from_secs(720),
            run_cluster_root_heal_interruption(InterruptionScenario::BackgroundCoordinatorRestart),
        )
        .await?
@@ -1010,29 +1072,38 @@ mod tests {
        IsolatedTargetRestart,
        BackgroundTargetRestart,
        BackgroundTargetCrash,
        BackgroundTargetRestartEc84,
        BackgroundTargetCrashEc84,
        BackgroundCoordinatorRestart,
        TargetEndpointBlackhole,
    }
    async fn run_cluster_root_heal_interruption(scenario: InterruptionScenario) -> Result<(), Box<dyn Error + Send + Sync>> {
        let server_binary = rustfs_binary_path();
-        let evidence_run = match scenario {
+        let evidence_case = match scenario {
-            InterruptionScenario::BackgroundTargetRestart => {
+            InterruptionScenario::BackgroundTargetRestart => Some(BACKGROUND_TARGET_RESTART_EVIDENCE),
-                restart_evidence_run(&server_binary, BACKGROUND_TARGET_RESTART_EVIDENCE)?
+            InterruptionScenario::BackgroundTargetCrash => Some(BACKGROUND_TARGET_CRASH_EVIDENCE),
-            }
+            InterruptionScenario::BackgroundTargetRestartEc84 => Some(BACKGROUND_TARGET_RESTART_EC84_EVIDENCE),
-            InterruptionScenario::BackgroundTargetCrash => {
+            InterruptionScenario::BackgroundTargetCrashEc84 => Some(BACKGROUND_TARGET_CRASH_EC84_EVIDENCE),
                restart_evidence_run(&server_binary, BACKGROUND_TARGET_CRASH_EVIDENCE)?
            }
            _ => None,
        };
        let evidence_run = match evidence_case {
            Some(case) => restart_evidence_run(&server_binary, case)?,
            None => None,
        };
        let mut evidence_objects = Vec::new();
        let (background_enabled, interruption_node, interruption_kind) = match scenario {
            InterruptionScenario::IsolatedTargetRestart => (false, 1, "target_restart"),
            InterruptionScenario::BackgroundTargetRestart => (true, 1, "background_target_restart"),
            InterruptionScenario::BackgroundTargetCrash => (true, 1, "background_target_crash"),
            InterruptionScenario::BackgroundTargetRestartEc84 => (true, 1, "background_target_restart_ec8_4"),
            InterruptionScenario::BackgroundTargetCrashEc84 => (true, 1, "background_target_crash_ec8_4"),
            InterruptionScenario::BackgroundCoordinatorRestart => (true, 0, "coordinator_restart"),
            InterruptionScenario::TargetEndpointBlackhole => (false, 1, "target_endpoint_blackhole"),
        };
        let topology = evidence_case
            .map(|case| case.topology)
            .unwrap_or_else(|| EvidenceTopology::new(4, 1));
        init_logging();
        info!(
            event = "heal_interruption_started",
@@ -1044,11 +1115,15 @@ mod tests {
            "Starting root-heal interruption test"
        );
-        let mut cluster = RustFSTestClusterEnvironment::new(4).await?;
+        let mut cluster = RustFSTestClusterEnvironment::with_topology(topology.cluster_topology()).await?;
        cluster.set_env("RUSTFS_UNSAFE_BYPASS_DISK_CHECK", "true");
        cluster.set_env("RUSTFS_HEAL_ENABLED", "true");
-        // Capture physical baselines after the PUT rename fanout has drained.
+        if let Some(storage_class) = evidence_case.and_then(|case| case.storage_class_standard) {
-        cluster.set_env("RUSTFS_PUT_RENAME_EARLY_ACK_ENABLE", "false");
+            cluster.set_env("RUSTFS_STORAGE_CLASS_STANDARD", storage_class);
        }
        if let Some(erasure_set_drive_count) = evidence_case.and_then(|case| case.erasure_set_drive_count) {
            cluster.set_env("RUSTFS_ERASURE_SET_DRIVE_COUNT", erasure_set_drive_count);
        }
        // Heal control uses the first lexicographically sorted grid host.
        // Keep that coordinator distinct from the remote target at index 1.
        cluster.nodes.sort_by(|left, right| left.url.cmp(&right.url));
@@ -1071,10 +1146,19 @@ mod tests {
        let server_rust_log = std::env::var("RUSTFS_HEAL_CHAOS_SERVER_RUST_LOG")
            .unwrap_or_else(|_| "rustfs::heal::task=info,rustfs=error".to_string());
        cluster.set_env("RUST_LOG", server_rust_log);
-        let log_dir = std::env::var("RUSTFS_HEAL_CHAOS_LOG_DIR").unwrap_or_else(|_| format!("{}/logs", cluster.temp_dir));
+        let log_dir = if let Some(directory) = std::env::var_os("RUSTFS_HEAL_CHAOS_LOG_DIR") {
            PathBuf::from(directory)
        } else if let Some(directory) = std::env::var_os("RUSTFS_E2E_LOG_DIR") {
            let cluster_name = Path::new(&cluster.temp_dir)
                .file_name()
                .ok_or("cluster directory has no name")?;
            PathBuf::from(directory).join(cluster_name).join("heal")
        } else {
            PathBuf::from(&cluster.temp_dir).join("logs")
        };
        std::fs::create_dir_all(&log_dir)?;
        for node_index in 0..cluster.nodes.len() {
-            cluster.set_node_capture_log_path(node_index, format!("{log_dir}/node{node_index}.log"))?;
+            cluster.set_node_capture_log_path(node_index, log_dir.join(format!("node{node_index}.log")).to_string_lossy())?;
        }
        cluster.start_with_binary(&server_binary).await?;
        let clients = cluster.create_all_clients()?;
@@ -1097,11 +1181,22 @@ mod tests {
            .and_then(|value| value.parse::<usize>().ok())
            .unwrap_or(4 * 1024 * 1024)
            .clamp(1024 * 1024, 16 * 1024 * 1024);
        let mut created_online_objects = Vec::with_capacity(online_object_count);
        let mut expected_manifests = Vec::with_capacity(online_object_count);
        let mut unclean_shutdown_marker_observed = None;
-        for index in 0..online_object_count {
+        let mut attempt_count = 0usize;
-            let key = format!("cluster/online/object-{index:04}.bin");
+        let max_online_attempts = online_object_count.saturating_mul(topology.total_drives().max(1));
-            let payload_seed = u8::try_from(index + 1).expect("clamped object count must fit in u8");
+        while expected_manifests.len() < online_object_count {
            if attempt_count >= max_online_attempts {
                return Err(format!(
                    "target replacement drive held only {}/{} baseline object shards after {attempt_count} writes",
                    expected_manifests.len(),
                    online_object_count
                )
                .into());
            }
            let key = format!("cluster/online/object-{attempt_count:04}.bin");
            let payload_seed = ((attempt_count % 251) + 1) as u8;
            timeout(
                Duration::from_secs(30),
                clients[0]
@@ -1113,6 +1208,13 @@ mod tests {
            )
            .await??;
            let shard_census = census_object_version_on_disk(&replaced_disk, bucket, &key, None)?;
            if !shard_census.has_xl_meta {
                timeout(Duration::from_secs(30), clients[0].delete_object().bucket(bucket).key(&key).send()).await??;
                attempt_count += 1;
                continue;
            }
            let shard_census =
                wait_for_complete_physical_shard_on_disk(&replaced_disk, bucket, &key, None, Duration::from_secs(10)).await?;
            assert!(
                shard_census.is_complete(),
                "node 1 should hold a complete baseline shard for {key}: {shard_census:?}"
@@ -1121,11 +1223,9 @@ mod tests {
                !shard_census.expected_part_numbers.is_empty(),
                "chaos objects must use physical part shards rather than inline data: {shard_census:?}"
            );
-            expected_manifests.push(PhysicalObjectManifest {
+            created_online_objects.push((key.clone(), payload_seed));
-                key,
+            expected_manifests.push(PhysicalObjectManifest { key, shard_census });
-                payload_seed,
+            attempt_count += 1;
                shard_census,
            });
        }
        let expected_pool_metadata = if background_enabled {
@@ -1171,31 +1271,38 @@ mod tests {
            if node_index == 1 {
                continue;
            }
-            let census = census_object_version_on_disk(Path::new(&node.data_dir), bucket, outage_key, None)?;
+            for (drive_index, drive) in node.data_dirs.iter().enumerate() {
-            assert!(
+                let census = census_object_version_on_disk(Path::new(drive), bucket, outage_key, None)?;
-                census.is_complete(),
+                assert!(
-                "online node {node_index} must hold a complete outage-object shard: {census:?}"
+                    census.is_complete(),
-            );
+                    "online node {node_index} drive {drive_index} must hold a complete outage-object shard: {census:?}"
-            let erasure_index = census
+                );
-                .erasure_index
+                let erasure_index = census.erasure_index.ok_or_else(|| {
-                .ok_or_else(|| format!("online node {node_index} outage-object shard has no erasure index: {census:?}"))?;
+                    format!("online node {node_index} drive {drive_index} outage-object shard has no erasure index: {census:?}")
-            assert!(
+                })?;
-                (1..=cluster.nodes.len()).contains(&erasure_index),
+                assert!(
-                "online node {node_index} outage-object erasure index is out of range: {census:?}"
+                    (1..=topology.total_drives()).contains(&erasure_index),
-            );
+                    "online node {node_index} drive {drive_index} outage-object erasure index is out of range: {census:?}"
-            assert!(
+                );
-                outage_peer_erasure_indices.insert(erasure_index),
+                assert!(
-                "outage-object erasure index {erasure_index} is duplicated across online nodes"
+                    outage_peer_erasure_indices.insert(erasure_index),
-            );
+                    "outage-object erasure index {erasure_index} is duplicated across online drives"
                );
            }
        }
        assert_eq!(
            outage_peer_erasure_indices.len(),
-            cluster.nodes.len().saturating_sub(1),
+            topology.total_drives().saturating_sub(cluster.nodes[1].data_dirs.len()),
-            "every online node must contribute one unique outage-object erasure index"
+            "every online drive must contribute one unique outage-object erasure index"
        );
        let missing_outage_erasure_indices = (1..=topology.total_drives())
            .filter(|index| !outage_peer_erasure_indices.contains(index))
            .collect::<HashSet<_>>();
        assert_eq!(
            missing_outage_erasure_indices.len(),
            cluster.nodes[1].data_dirs.len(),
            "the stopped node must account for every missing outage-object erasure index"
        );
        let expected_outage_target_erasure_index = (1..=cluster.nodes.len())
            .find(|index| !outage_peer_erasure_indices.contains(index))
            .ok_or("online outage-object shards leave no erasure index for the replacement target")?;
        let heal_body = r#"{"recursive":true,"dryRun":false,"remove":false,"recreate":true,"scanMode":2,"updateParity":false,"nolock":false}"#;
        if !background_enabled {
@@ -1294,6 +1401,9 @@ mod tests {
            }
            sleep(Duration::from_millis(50)).await;
        }
        // Task execution and its non-blocking log writer advance independently.
        // Observe the exact start before taking the partial-rebuild snapshot.
        wait_for_admin_cluster_start_log(&log_dir.join("node0.log"), client_token, partial_deadline).await?;
        let (partial_count, partial_manifest) = loop {
            // Hash one committed shard to prove progress without letting a
            // full-corpus hash pass consume the interruption window.
@@ -1328,19 +1438,6 @@ mod tests {
        let pre_interrupt_status: serde_json::Value = serde_json::from_str(&pre_interrupt_status_body)
            .map_err(|err| format!("pre-interrupt background heal status is not JSON ({err}): {pre_interrupt_status_body}"))?;
        let pre_interrupt_replacement = replacement_recovery_status(&cluster).await?;
        let coordinator_log = std::fs::read_to_string(format!("{log_dir}/node0.log"))?;
        assert!(
            coordinator_log
                .lines()
                .filter_map(|line| serde_json::from_str::<serde_json::Value>(line).ok())
                .any(|event| {
                    event["event"] == "heal_task_state"
                        && event["task_id"] == client_token
                        && event["heal_type"] == "cluster"
                        && event["state"] == "started"
                }),
            "node 0 must have started the exact admin task before interruption"
        );
        let pre_interrupt_operations = &pre_interrupt_status["healOperations"];
        assert_eq!(
            pre_interrupt_operations["activeBySource"]["admin"].as_u64(),
@@ -1465,7 +1562,11 @@ mod tests {
                "Restored target endpoint forwarding"
            );
        } else {
-            if scenario == InterruptionScenario::BackgroundTargetRestart {
+            let graceful_restart = matches!(
                scenario,
                InterruptionScenario::BackgroundTargetRestart | InterruptionScenario::BackgroundTargetRestartEc84
            );
            if graceful_restart {
                cluster.stop_node_gracefully(interruption_node).await?;
            } else {
                cluster.stop_node(interruption_node)?;
@@ -1487,7 +1588,7 @@ mod tests {
            if background_enabled {
                let marker_exists = unclean_shutdown_marker.is_file();
                unclean_shutdown_marker_observed = Some(marker_exists);
-                let expected_marker = !matches!(scenario, InterruptionScenario::BackgroundTargetRestart);
+                let expected_marker = !graceful_restart;
                assert!(
                    marker_exists == expected_marker,
                    "background restart/crash lane observed unexpected unclean-shutdown marker state"
@@ -1543,7 +1644,19 @@ mod tests {
                }
            }
            if Instant::now() >= heal_deadline {
-                let matching = matching_manifest_count(&replaced_disk, bucket, &expected_manifests)?;
+                let mut mismatched_manifests = Vec::new();
                for expected in &expected_manifests {
                    let actual = census_object_version_on_disk(&replaced_disk, bucket, &expected.key, None)?;
                    if !actual.matches_manifest(&expected.shard_census) {
                        mismatched_manifests.push(serde_json::json!({
                            "key": expected.key,
                            "expected": expected.shard_census,
                            "actual": actual,
                        }));
                    }
                }
                let matching = expected_manifests.len() - mismatched_manifests.len();
                let mismatched_manifests = serde_json::Value::Array(mismatched_manifests);
                let outage_census = census_object_version_on_disk(&replaced_disk, bucket, outage_key, None)?;
                let pool_metadata =
                    census_object_version_on_disk(&replaced_disk, RUSTFS_META_BUCKET, POOL_METADATA_OBJECT, None)?;
@@ -1566,7 +1679,7 @@ mod tests {
                    Err(_) => "replacement status request exceeded 5s diagnostic budget".to_string(),
                };
                return Err(format!(
-                    "root heal did not recover after {interruption_kind} within {heal_timeout_secs}s: baseline={matching}/{}, outage={outage_census:?}, pool_metadata={pool_metadata:?}, status={final_status}, task_status={task_status}, pre_interrupt_status={pre_interrupt_status}, pre_heal_replacement={pre_heal_replacement}, pre_interrupt_replacement={pre_interrupt_replacement}, replacement_status={replacement_status}",
+                    "root heal did not recover after {interruption_kind} within {heal_timeout_secs}s: baseline={matching}/{}, mismatched_manifests={mismatched_manifests}, outage={outage_census:?}, pool_metadata={pool_metadata:?}, status={final_status}, task_status={task_status}, pre_interrupt_status={pre_interrupt_status}, pre_heal_replacement={pre_heal_replacement}, pre_interrupt_replacement={pre_interrupt_replacement}, replacement_status={replacement_status}",
                    expected_manifests.len()
                )
                .into());
@@ -1588,29 +1701,35 @@ mod tests {
            "outage object must have a complete target shard: {outage_census:?}"
        );
        assert_eq!(
            outage_census
                .erasure_index
                .filter(|index| missing_outage_erasure_indices.contains(index)),
            outage_census.erasure_index,
-            Some(expected_outage_target_erasure_index),
+            "the outage object must be rebuilt into one of the stopped node's missing erasure slots"
            "the outage object must be rebuilt into its own missing erasure slot"
        );
        if let Some(cycle_end) = scanner_cycle_floor {
            wait_for_scanner_cycle_after(&cluster, cycle_end).await?;
        }
-        let mut expected_keys = expected_manifests
+        let mut expected_keys = created_online_objects
            .iter()
-            .map(|manifest| manifest.key.clone())
+            .map(|(key, _)| key.clone())
            .collect::<HashSet<_>>();
        assert!(expected_keys.insert(outage_key.to_string()));
        let node_listings = assert_all_nodes_list_exact_keys(&clients, bucket, &expected_keys).await?;
        let target_client = cluster.create_s3_client(1)?;
-        for expected in &expected_manifests {
+        for (key, payload_seed) in &created_online_objects {
-            let response = target_client.get_object().bucket(bucket).key(&expected.key).send().await?;
+            let response = target_client.get_object().bucket(bucket).key(key).send().await?;
            let actual = response.body.collect().await?.into_bytes();
-            let expected_body = deterministic_object_body(object_size_bytes, expected.payload_seed);
+            let expected_body = deterministic_object_body(object_size_bytes, *payload_seed);
-            assert_eq!(actual.as_ref(), expected_body.as_slice(), "object body changed for {}", expected.key);
+            assert_eq!(actual.as_ref(), expected_body.as_slice(), "object body changed for {key}");
-            if evidence_run.is_some() {
+            if evidence_run.is_some()
                && let Some(expected) = expected_manifests
                    .iter()
                    .find(|manifest| manifest.key.as_str() == key.as_str())
            {
                evidence_objects.push(serde_json::json!({
                    "key": expected.key, "version_id": expected.shard_census.version_id,
                    "expected_bytes": expected_body.len(), "actual_bytes": actual.len(),
@@ -1684,32 +1803,18 @@ mod tests {
        if let Some(evidence_context) = evidence_run {
            let restarted_pid = cluster.nodes[1].process.as_ref().ok_or("restarted target is absent")?.id();
            assert_ne!(target_pid, restarted_pid, "target must be a new process");
-            assert_eq!(
+            evidence_context.write(
-                file_sha256(&server_binary)?,
+                &server_binary,
-                evidence_context.run.binary.sha256,
+                RestartObservation {
-                "server build changed during restart"
+                    nodes: cluster.nodes.len(),
-            );
+                    drives_per_node: cluster.nodes[0].data_dirs.len(),
-            let evidence = serde_json::json!({
+                    pid_before: target_pid,
-                "schema": 1, "case": evidence_context.case.id, "evidence": evidence_context.case.evidence,
+                    pid_after: restarted_pid,
-                "run_id": evidence_context.run.run_id, "source_revision": evidence_context.run.source_revision,
+                    unclean_shutdown_marker: unclean_shutdown_marker_observed.ok_or("missing shutdown marker observation")?,
-                "test_build": compiled_test_identity(),
+                    objects: evidence_objects,
-                "binary_sha256": evidence_context.run.binary.sha256,
+                    node_listings,
-                "test_binary_sha256": evidence_context.run.test_binary.sha256,
+                },
-                "topology": {"nodes": cluster.nodes.len(), "drives_per_node": cluster.nodes[0].data_dirs.len()},
+            )?;
                "pid_before": target_pid, "pid_after": restarted_pid,
                "unclean_shutdown_marker": unclean_shutdown_marker_observed.unwrap_or(false),
                "objects": evidence_objects, "node_listings": node_listings,
            });
            let data = serde_json::to_vec(&evidence)?;
            if data.len() > 1024 * 1024 {
                return Err("scanner/heal oracle exceeds the 1 MiB artifact budget".into());
            }
            let mut output = std::fs::OpenOptions::new()
                .write(true)
                .create_new(true)
                .open(evidence_context.directory.join(evidence_context.case.oracle))?;
            output.write_all(&data)?;
            output.sync_all()?;
        }
        Ok(())
@@ -21,7 +21,7 @@
 //! One S3 GET can select readers on multiple EC nodes, so the counter tracks
 //! distributed reader selection rather than HTTP request count.
-use crate::common::{RustFSTestClusterEnvironment, RustFSTestEnvironment, init_logging};
+use crate::common::{RustFSTestClusterEnvironment, RustFSTestEnvironment, init_logging, signal_process};
 use aws_sdk_s3::Client;
 use aws_sdk_s3::primitives::ByteStream;
 use aws_sdk_s3::types::{
@@ -2207,6 +2207,33 @@ async fn four_node_manual_transition_job_status_survives_node_restart() -> TestR
    Ok(())
 }
 struct SuspendedTransitionTarget<'a> {
    // Keep the owned child borrowed until it is resumed so its PID cannot be reused.
    child: &'a std::process::Child,
    suspended: bool,
 }
 impl<'a> SuspendedTransitionTarget<'a> {
    fn suspend(child: &'a std::process::Child) -> TestResult<Self> {
        signal_process(child.id(), "STOP")?;
        Ok(Self { child, suspended: true })
    }
    fn resume(&mut self) -> TestResult {
        signal_process(self.child.id(), "CONT")?;
        self.suspended = false;
        Ok(())
    }
 }
 impl Drop for SuspendedTransitionTarget<'_> {
    fn drop(&mut self) {
        if self.suspended {
            let _ = signal_process(self.child.id(), "CONT");
        }
    }
 }
 #[tokio::test]
 async fn four_node_manual_transition_distributed_admission_conflict_reports_status_and_backpressure() -> TestResult {
    init_logging();
@@ -2244,7 +2271,21 @@ async fn four_node_manual_transition_distributed_admission_conflict_reports_stat
            .send()
            .await?;
    }
-    put_lifecycle_with_transition_retry(&hot_client, &bucket, &tier_name).await?;
+    // Lifecycle PUT starts its own backfill. Keep its first page on a separate
    // node and stop it at queue backpressure before it reaches the tested prefix:
    // one active worker, one queued item, then the first rejected item.
    for index in 0u8..3 {
        hot_client
            .put_object()
            .bucket(&bucket)
            .key(format!("transition/automatic-admission/object-{index:02}.bin"))
            .body(ByteStream::from(payload(KIB, index)))
            .send()
            .await?;
    }
    let mut suspended_cold = SuspendedTransitionTarget::suspend(cold.process.as_ref().ok_or("cold-tier process missing")?)?;
    let lifecycle_client = hot.create_s3_client(2)?;
    put_lifecycle_with_transition_retry(&lifecycle_client, &bucket, &tier_name).await?;
    let (node0, node1) = tokio::join!(
        start_manual_transition_job_on_node(&hot, 0, &bucket, prefix, &tier_name, false, 64),
@@ -2304,6 +2345,31 @@ async fn four_node_manual_transition_distributed_admission_conflict_reports_stat
    assert_eq!(status["job_id"].as_str(), Some(job_id));
    assert_eq!(status["status_endpoint"].as_str(), Some(status_endpoint));
    let deadline = Instant::now() + Duration::from_secs(30);
    loop {
        let status = read_manual_transition_job_status_endpoint(&hot, accepted.0, status_endpoint).await?;
        assert_eq!(
            status["status"].as_str(),
            Some("running"),
            "blocked cold tier must keep the admitted job running: {status}"
        );
        if status["report"]["skipped_queue_full"].as_u64().is_some_and(|count| count > 0) {
            assert!(
                status["report"]["enqueued"].as_u64().is_some_and(|count| count > 0),
                "the job must own pending transitions while the cold tier is suspended: {status}"
            );
            break;
        }
        if Instant::now() >= deadline {
            return Err(format!(
                "manual transition job did not reach queue backpressure while the cold tier was suspended: {status}"
            )
            .into());
        }
        sleep(Duration::from_millis(50)).await;
    }
    suspended_cold.resume()?;
    let terminal = wait_for_manual_transition_job_terminal(&hot, conflict.0, job_id, false).await?;
    assert_eq!(terminal["job_id"].as_str(), Some(job_id));
    assert_eq!(terminal["bucket"].as_str(), Some(bucket.as_str()));
@@ -23,6 +23,9 @@ pub mod common;
 #[cfg(test)]
 pub mod chaos;
 #[cfg(test)]
 mod scanner_heal_evidence;
 // Programmable S3 target for replication failure-path tests (backlog#1147 repl-8)
 // and on-demand-migration source scenarios (backlog#2151).
 #[cfg(test)]
@@ -25,190 +25,6 @@ const KEY: &str = "thumb/79/concurrent-overwrite.jpg";
 type TestResult = Result<(), Box<dyn std::error::Error + Send + Sync>>;
 async fn assert_quorum_object_body(client: &Client, bucket: &str, key: &str, expected: &[u8]) -> TestResult {
    let body = client
        .get_object()
        .bucket(bucket)
        .key(key)
        .send()
        .await?
        .body
        .collect()
        .await?
        .into_bytes();
    assert_eq!(body.as_ref(), expected, "quorum read returned incorrect contents for {key}");
    Ok(())
 }
 async fn wait_for_quorum_read_admission(clients: &[Client], bucket: &str) -> TestResult {
    // SIGKILL can orphan a granted lease. Wait for shared metadata-lock
    // admission before asserting the stable quorum boundary; cold bodies
    // remain unread throughout this readiness probe.
    let deadline =
        tokio::time::Instant::now() + rustfs_lock::fast_lock::DEFAULT_LOCK_TIMEOUT + std::time::Duration::from_secs(15);
    loop {
        let mut ready = true;
        for client in clients {
            for key in ["warm-small", "warm-large"] {
                match client.head_object().bucket(bucket).key(key).send().await {
                    Ok(_) => {}
                    Err(error) if error.raw_response().is_some_and(|response| response.status().as_u16() == 503) => {
                        ready = false;
                        break;
                    }
                    Err(error) => return Err(error.into()),
                }
            }
            if !ready {
                break;
            }
        }
        if ready {
            return Ok(());
        }
        if tokio::time::Instant::now() >= deadline {
            return Err(format!("read quorum did not become available after lease convergence for {bucket}").into());
        }
        tokio::time::sleep(std::time::Duration::from_millis(100)).await;
    }
 }
 #[tokio::test]
 async fn test_degraded_cluster_read_quorum_follows_erasure_layout() -> TestResult {
    crate::common::init_logging();
    for (node_count, parity) in [(4, 2), (6, 3), (6, 2)] {
        let read_quorum = node_count - parity;
        let write_quorum = read_quorum + usize::from(read_quorum == parity);
        let mut cluster = RustFSTestClusterEnvironment::new(node_count).await?;
        cluster.set_env("RUSTFS_STORAGE_CLASS_STANDARD", format!("EC:{parity}"));
        // Wait for every seed fanout before removing any physical shard.
        cluster.set_env("RUSTFS_PUT_RENAME_EARLY_ACK_ENABLE", "false");
        cluster.set_env("RUSTFS_OBS_METRICS_EXPORT_ENABLED", "false");
        cluster.set_env("RUST_LOG", "warn,rustfs_lock=debug");
        cluster.start().await?;
        let clients = cluster
            .create_all_clients()?
            .into_iter()
            .map(|client| {
                Client::from_conf(
                    client
                        .config()
                        .to_builder()
                        .retry_config(aws_sdk_s3::config::retry::RetryConfig::standard().with_max_attempts(1))
                        .build(),
                )
            })
            .collect::<Vec<_>>();
        let bucket = format!("read-quorum-{node_count}-{parity}");
        clients[0].create_bucket().bucket(&bucket).send().await?;
        let small = b"read quorum is derived from the erasure layout".to_vec();
        let large = (0..1_048_576)
            .map(|index| u8::try_from(index % 251).expect("bounded payload byte"))
            .collect::<Vec<_>>();
        for (key, body) in [
            ("warm-small", &small),
            ("warm-large", &large),
            ("cold-small", &small),
            ("cold-large", &large),
            ("below-quorum", &large),
        ] {
            clients[node_count - 1]
                .put_object()
                .bucket(&bucket)
                .key(key)
                .body(Bytes::copy_from_slice(body).into())
                .send()
                .await?;
        }
        for node in &cluster.nodes {
            for key in ["warm-small", "warm-large", "cold-small", "cold-large", "below-quorum"] {
                let census =
                    crate::chaos::census_object_version_on_disk(std::path::Path::new(&node.data_dir), &bucket, key, None)?;
                assert!(census.is_complete(), "seed shard must be complete before fault injection: {census:?}");
                assert_eq!(census.data_blocks, Some(read_quorum));
                assert_eq!(census.parity_blocks, Some(parity));
            }
        }
        for client in &clients {
            assert_quorum_object_body(client, &bucket, "warm-small", &small).await?;
            assert_quorum_object_body(client, &bucket, "warm-large", &large).await?;
        }
        for offline_node in (read_quorum..node_count).rev() {
            cluster.stop_node(offline_node)?;
            wait_for_quorum_read_admission(&clients[..offline_node], &bucket).await?;
            for client in clients.iter().take(offline_node) {
                client.head_bucket().bucket(&bucket).send().await?;
                assert_quorum_object_body(client, &bucket, "warm-large", &large).await?;
            }
        }
        // Exercise more than the five-second positive bucket-validation TTL.
        // Every sample must succeed; polling must not hide a transient failure.
        let validation_deadline = tokio::time::Instant::now() + std::time::Duration::from_secs(6);
        loop {
            for client in clients.iter().take(read_quorum) {
                assert_quorum_object_body(client, &bucket, "warm-small", &small).await?;
                assert_quorum_object_body(client, &bucket, "warm-large", &large).await?;
                let listing = client.list_objects_v2().bucket(&bucket).send().await?;
                for key in ["warm-small", "warm-large", "cold-small", "cold-large", "below-quorum"] {
                    assert!(listing.contents().iter().any(|entry| entry.key() == Some(key)), "listing omitted {key}");
                }
            }
            if tokio::time::Instant::now() >= validation_deadline {
                break;
            }
            tokio::time::sleep(std::time::Duration::from_millis(250)).await;
        }
        for client in clients.iter().take(read_quorum) {
            assert_quorum_object_body(client, &bucket, "cold-small", &small).await?;
            assert_quorum_object_body(client, &bucket, "cold-large", &large).await?;
        }
        let write = clients[0]
            .put_object()
            .bucket(&bucket)
            .key("quorum-write")
            .body(Bytes::copy_from_slice(&small).into())
            .send()
            .await;
        if read_quorum >= write_quorum {
            write?;
        } else {
            let error = write.expect_err("a read quorum must not authorize a write that needs more votes");
            assert_eq!(error.as_service_error().and_then(|error| error.meta().code()), Some("ServiceUnavailable"));
        }
        cluster.stop_node(read_quorum - 1)?;
        for client in clients.iter().take(read_quorum - 1) {
            match client.get_object().bucket(&bucket).key("below-quorum").send().await {
                Ok(response) => assert!(
                    response.body.collect().await.is_err(),
                    "fewer than {read_quorum} valid fragments must not reconstruct an uncached object"
                ),
                Err(error) => assert_eq!(
                    error.as_service_error().and_then(|error| error.meta().code()),
                    Some("ServiceUnavailable"),
                    "a quorum loss must not be mistaken for a missing object"
                ),
            }
        }
        for node in 0..read_quorum - 1 {
            cluster.stop_node(node)?;
        }
        cluster.start().await?;
        for client in &clients {
            assert_quorum_object_body(client, &bucket, "warm-large", &large).await?;
            assert_quorum_object_body(client, &bucket, "below-quorum", &large).await?;
        }
    }
    Ok(())
 }
 async fn put_object(client: Client, payload: Vec<u8>, writer_id: usize) -> Result<(), String> {
    client
        .put_object()
@@ -1137,12 +1137,7 @@ async fn test_odm_admin_config_is_redacted_and_status_counts_match_the_source()
        let miss = env.raw_get(bucket, miss_key).await?;
        assert_eq!(miss.status, 404, "{}", String::from_utf8_lossy(&miss.body));
    }
-    let (listed, _, _) = tokio::try_join!(
+    assert!(env.wait_local_listed(bucket, hit_key, SETTLE).await?);
        env.wait_local_listed(bucket, hit_key, SETTLE),
        env.wait_for_status_counter(bucket, "/counters/pulled_objects_total/inline", 1, SETTLE),
        env.wait_for_status_counter(bucket, "/counters/pulled_bytes_total", body.len() as u64, SETTLE),
    )?;
    assert!(listed);
    let status = env.status_json(bucket).await?;
    assert_eq!(status.pointer("/configured").and_then(Value::as_bool), Some(true), "{status}");
@@ -79,6 +79,11 @@ pub async fn test_sftp_compliance_suite() -> Result<()> {
            .await
            .map_err(|e| anyhow!("{}", e))?;
        // Protocol listeners can accept connections before IAM is initialized.
        // A signed S3 request establishes readiness before the first SFTP login.
        let s3 = build_test_s3_client(&format!("http://{COMPLIANCE_RW_S3_ADDRESS}"));
        wait_for_s3_ready(&s3, 30).await?;
        let (session, sftp) = connect_sftp_to(COMPLIANCE_RW_SFTP_ADDRESS).await?;
        cmptst_01::run_medium_binary_round_trip(&sftp).await?;
@@ -101,8 +106,6 @@ pub async fn test_sftp_compliance_suite() -> Result<()> {
        // reach the finalised object as x-amz-meta-* user metadata
        // through the CreateMultipartUpload input field. The S3 client
        // connects to the same rustfs process this suite already drives.
        let s3 = build_test_s3_client(&format!("http://{COMPLIANCE_RW_S3_ADDRESS}"));
        wait_for_s3_ready(&s3, 30).await?;
        cmptst_34::run_open_attrs_round_trip_multipart(&sftp, &s3).await?;
        drop(sftp);
@@ -168,6 +168,10 @@ pub async fn test_sftp_core_operations() -> Result<()> {
            .await
            .map_err(|e| anyhow!("{}", e))?;
        // Protocol listeners can accept connections before IAM is initialized.
        let s3 = build_test_s3_client(S3_ENDPOINT);
        wait_for_s3_ready(&s3, S3_READY_ATTEMPTS).await?;
        let (session, sftp) = connect_sftp().await?;
        // --- 1. Subsystem canary: SFTP session reachable after password auth ---
@@ -348,16 +352,6 @@ pub async fn test_sftp_core_operations() -> Result<()> {
        let _ = bad_session.disconnect(russh::Disconnect::ByApplication, "", "en").await;
        info!("PASS: bad-password authentication rejected");
        // --- Cross-protocol setup: aws-sdk-s3 client against the same server ---
        // The rustfs binary spawned for this suite serves both SFTP on port
        // 9022 and S3 on port 9000. The S3 stack may need a moment to finish
        // initialising after TCP is listening, so list_buckets is polled
        // until it succeeds before any cross-protocol assertion runs.
        info!("Testing SFTP: prepare aws-sdk-s3 client and wait for S3 readiness");
        let s3 = build_test_s3_client(S3_ENDPOINT);
        wait_for_s3_ready(&s3, S3_READY_ATTEMPTS).await?;
        info!("PASS: S3 endpoint reachable from cross-protocol client");
        // --- SFTP write, S3 read: SHA256 round-trip ---
        // SFTP creates the object, then assert_cross_protocol_sha_match
        // fetches it via both S3 GetObject and SFTP READ and compares
@@ -522,6 +516,9 @@ pub async fn test_sftp_idle_timeout_disconnects() -> Result<()> {
            .await
            .map_err(|e| anyhow!("{}", e))?;
        let s3 = build_test_s3_client(&format!("http://{IDLE_S3_ADDRESS}"));
        wait_for_s3_ready(&s3, S3_READY_ATTEMPTS).await?;
        let (session, sftp) = connect_sftp_to(IDLE_SFTP_ADDRESS).await?;
        // Confirm the session is live before the wait so a failure in the
@@ -4235,16 +4235,6 @@ async fn test_bucket_replication_acceptance_matrix_local_dual_targets() -> TestR
    <ExistingObjectReplication><Status>Enabled</Status></ExistingObjectReplication>
    <Destination><Bucket>{target_b_arn}</Bucket></Destination>
  </Rule>
  <Rule>
    <ID>matrix-and-tags</ID>
    <Priority>135</Priority>
    <Status>Enabled</Status>
    <Filter><And><Prefix>and-tags/</Prefix><Tag><Key>env</Key><Value>prod</Value></Tag><Tag><Key>tier</Key><Value>gold</Value></Tag></And></Filter>
    <DeleteMarkerReplication><Status>Disabled</Status></DeleteMarkerReplication>
    <DeleteReplication><Status>Enabled</Status></DeleteReplication>
    <ExistingObjectReplication><Status>Enabled</Status></ExistingObjectReplication>
    <Destination><Bucket>{target_b_arn}</Bucket></Destination>
  </Rule>
  <Rule>
    <ID>matrix-disabled</ID>
    <Priority>140</Priority>
@@ -4299,7 +4289,6 @@ async fn test_bucket_replication_acceptance_matrix_local_dual_targets() -> TestR
        "matrix-prefix",
        "matrix-tag",
        "matrix-disabled",
        "matrix-and-tags",
        "matrix-priority-high",
        "Priority>200",
        "<Status>Disabled</Status>",
@@ -4420,30 +4409,6 @@ async fn test_bucket_replication_acceptance_matrix_local_dual_targets() -> TestR
    put_single_tag_current(&source_client, source_bucket, "tagged/no-match.txt", "route", "tagged").await?;
    assert_replication_key_absent(&target_client_b, target_bucket_b, "tagged/no-match.txt", Duration::from_secs(3)).await?;
    // S3 and MinIO both read `And.Tags` as AND: an object carrying only one of
    // the required tags is not admitted. Matching any single tag would push
    // data to a destination the rule never selected (backlog#2366 P1-1), and
    // the two-tag rule is the shape `mc replicate add --tags "k1=v1&k2=v2"`
    // writes, so a single-tag rule passing is not evidence for this.
    source_client
        .put_object()
        .bucket(source_bucket)
        .key("and-tags/partial.txt")
        .tagging("env=prod")
        .body(ByteStream::from_static(b"one of two tags"))
        .send()
        .await?;
    assert_replication_key_absent(&target_client_b, target_bucket_b, "and-tags/partial.txt", Duration::from_secs(3)).await?;
    source_client
        .put_object()
        .bucket(source_bucket)
        .key("and-tags/full.txt")
        .tagging("env=prod&tier=gold")
        .body(ByteStream::from_static(b"both tags"))
        .send()
        .await?;
    wait_for_user_get_object(&target_client_b, target_bucket_b, "and-tags/full.txt").await?;
    source_client
        .put_object()
        .bucket(source_bucket)
@@ -0,0 +1,181 @@
 // Copyright 2026 RustFS Team
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
 //
 //     http://www.apache.org/licenses/LICENSE-2.0
 //
 // Unless required by applicable law or agreed to in writing, software
 // distributed under the License is distributed on an "AS IS" BASIS,
 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 // See the License for the specific language governing permissions and
 // limitations under the License.
 //! Build-bound evidence for scanner and heal restart tests.
 use crate::common::ClusterTopology;
 use sha2::{Digest, Sha256};
 use std::error::Error;
 use std::io::{Read, Write};
 use std::path::{Path, PathBuf};
 #[derive(serde::Deserialize)]
 struct EvidenceBuild {
    sha256: String,
 }
 #[derive(serde::Deserialize)]
 struct RestartEvidenceRun {
    schema: u32,
    run_id: String,
    source_revision: String,
    test_build: serde_json::Value,
    binary: EvidenceBuild,
    test_binary: EvidenceBuild,
 }
 #[derive(Clone, Copy)]
 pub(crate) struct ScannerHealEvidenceCase {
    pub(crate) id: &'static str,
    pub(crate) oracle: &'static str,
    pub(crate) evidence: &'static str,
    pub(crate) unclean_shutdown_marker: bool,
    pub(crate) topology: EvidenceTopology,
    pub(crate) storage_class_standard: Option<&'static str>,
    pub(crate) erasure_set_drive_count: Option<&'static str>,
 }
 #[derive(Clone, Copy)]
 pub(crate) struct EvidenceTopology {
    pub(crate) nodes: usize,
    pub(crate) drives_per_node: usize,
 }
 impl EvidenceTopology {
    pub(crate) const fn new(nodes: usize, drives_per_node: usize) -> Self {
        Self { nodes, drives_per_node }
    }
    pub(crate) fn total_drives(self) -> usize {
        self.nodes * self.drives_per_node
    }
    pub(crate) fn cluster_topology(self) -> ClusterTopology {
        ClusterTopology::single_pool_multidrive(self.nodes, self.drives_per_node)
    }
 }
 pub(crate) struct RestartEvidenceContext {
    directory: PathBuf,
    run: RestartEvidenceRun,
    case: ScannerHealEvidenceCase,
 }
 fn file_sha256(path: &Path) -> Result<String, Box<dyn Error + Send + Sync>> {
    let mut file = std::fs::File::open(path)?;
    let mut digest = Sha256::new();
    let mut buffer = [0_u8; 64 * 1024];
    loop {
        let read = file.read(&mut buffer)?;
        if read == 0 {
            break;
        }
        digest.update(&buffer[..read]);
    }
    Ok(digest.finalize().iter().map(|byte| format!("{byte:02x}")).collect())
 }
 pub(crate) fn restart_evidence_run(
    binary: &Path,
    case: ScannerHealEvidenceCase,
 ) -> Result<Option<RestartEvidenceContext>, Box<dyn Error + Send + Sync>> {
    let Some(directory) = std::env::var_os("RUSTFS_SCANNER_HEAL_RUN_DIR") else {
        return Ok(None);
    };
    if case.id.is_empty()
        || case.oracle.is_empty()
        || !case.oracle.ends_with(".json")
        || case.oracle.contains('/')
        || case.oracle.contains('\\')
        || case.oracle.contains("..")
        || !matches!(case.evidence, "process-restart" | "process-crash-restart")
        || (case.evidence == "process-crash-restart") != case.unclean_shutdown_marker
    {
        return Err("invalid scanner/heal evidence case".into());
    }
    let directory = PathBuf::from(directory);
    let receipt = directory.join("run.json");
    if receipt.metadata()?.len() > 1024 * 1024 {
        return Err("oversized scanner/heal execution receipt".into());
    }
    let run: RestartEvidenceRun = serde_json::from_slice(&std::fs::read(receipt)?)?;
    if run.schema != 1 || run.run_id.len() != 32 || run.source_revision.len() != 40 {
        return Err("invalid scanner/heal execution identity".into());
    }
    let built = compiled_test_identity();
    for key in ["source_revision", "dirty", "lock_blob", "features"] {
        assert_eq!(built[key], run.test_build[key], "compiled test identity differs for {key}");
    }
    assert_eq!(file_sha256(binary)?, run.binary.sha256, "server binary must match the run receipt");
    assert_eq!(
        file_sha256(&std::env::current_exe()?)?,
        run.test_binary.sha256,
        "test executable must match the run receipt"
    );
    if directory.join(case.oracle).exists() {
        return Err("scanner/heal oracle already exists; create a new execution receipt".into());
    }
    Ok(Some(RestartEvidenceContext { directory, run, case }))
 }
 fn compiled_test_identity() -> serde_json::Value {
    serde_json::json!({
        "source_revision": env!("RUSTFS_E2E_BUILD_COMMIT"),
        "dirty": env!("RUSTFS_E2E_BUILD_DIRTY") != "false",
        "lock_blob": env!("RUSTFS_E2E_BUILD_LOCK"),
        "features": env!("RUSTFS_E2E_BUILD_FEATURES"),
        "target": env!("RUSTFS_E2E_BUILD_TARGET"),
        "profile": env!("RUSTFS_E2E_BUILD_PROFILE"),
        "rustflags_hex": env!("RUSTFS_E2E_BUILD_RUSTFLAGS_HEX"),
    })
 }
 pub(crate) struct RestartObservation {
    pub(crate) nodes: usize,
    pub(crate) drives_per_node: usize,
    pub(crate) pid_before: u32,
    pub(crate) pid_after: u32,
    pub(crate) unclean_shutdown_marker: bool,
    pub(crate) objects: Vec<serde_json::Value>,
    pub(crate) node_listings: Vec<Vec<String>>,
 }
 impl RestartEvidenceContext {
    pub(crate) fn write(self, binary: &Path, observed: RestartObservation) -> Result<(), Box<dyn Error + Send + Sync>> {
        assert_ne!(observed.pid_before, observed.pid_after, "target must be a new process");
        assert_eq!(file_sha256(binary)?, self.run.binary.sha256, "server build changed during restart");
        let evidence = serde_json::json!({
            "schema": 1, "case": self.case.id, "evidence": self.case.evidence,
            "run_id": self.run.run_id, "source_revision": self.run.source_revision,
            "test_build": compiled_test_identity(),
            "binary_sha256": self.run.binary.sha256,
            "test_binary_sha256": self.run.test_binary.sha256,
            "topology": {"nodes": observed.nodes, "drives_per_node": observed.drives_per_node},
            "pid_before": observed.pid_before, "pid_after": observed.pid_after,
            "unclean_shutdown_marker": observed.unclean_shutdown_marker,
            "objects": observed.objects, "node_listings": observed.node_listings,
        });
        let data = serde_json::to_vec(&evidence)?;
        if data.len() > 1024 * 1024 {
            return Err("scanner/heal oracle exceeds the 1 MiB artifact budget".into());
        }
        let mut output = std::fs::OpenOptions::new()
            .write(true)
            .create_new(true)
            .open(self.directory.join(self.case.oracle))?;
        output.write_all(&data)?;
        output.sync_all()?;
        Ok(())
    }
 }
@@ -563,7 +563,7 @@ pub mod store_list {
 }
 pub mod storage {
-    pub use crate::core::pools::HealLifecycleExpiryContext;
+    pub use crate::core::pools::{HealLifecycleExpiryContext, POOL_META_NAME};
    pub use crate::store::HealWalkVersion;
    pub use crate::store::{
        BootstrapLocalTarget, ECStore, SCANNER_PUBLICATION_LEASE_TTL_MS, ScannerDataMovementPauseStatus, all_local_disk,
@@ -285,6 +285,20 @@ fn peer_replay_state(audience: &str) -> PeerReplayState {
        .unwrap_or_default()
 }
 pub(crate) fn clear_peer_replay_state_for_addr(addr: &str) -> std::io::Result<()> {
    let uri = addr
        .parse::<Uri>()
        .map_err(|_| std::io::Error::other("Invalid gRPC peer URI"))?;
    let audience = uri
        .authority()
        .map(|authority| normalize_tonic_rpc_audience(authority.as_str()))
        .ok_or_else(|| std::io::Error::other("Missing gRPC peer authority"))??;
    if let Ok(mut states) = PEER_REPLAY_STATES.lock() {
        states.remove(&audience);
    }
    Ok(())
 }
 fn apply_peer_replay_response(
    audience: String,
    sent_state: PeerReplayState,
@@ -619,6 +633,13 @@ mod tests {
            .remove(audience);
    }
    fn set_peer_capability(audience: &str, state: PeerReplayState) {
        PEER_REPLAY_STATES
            .lock()
            .expect("peer capability cache lock must not be poisoned")
            .insert(audience.to_string(), state);
    }
    fn rolling_mutation_request(method: &'static str) -> tonic::Request<()> {
        let mut request = tonic::Request::new(rustfs_protos::proto_gen::node_service::GenerallyLockRequest {
            args: "canonical mutation request".to_string(),
@@ -1090,6 +1111,23 @@ mod tests {
        clear_peer_capability(audience);
    }
    #[test]
    fn clear_peer_replay_state_for_addr_removes_normalized_audience() {
        let audience = "clear-peer-replay-state-test:9000";
        let boot_epoch = Uuid::new_v4();
        set_peer_capability(
            audience,
            PeerReplayState {
                boot_epoch: Some(boot_epoch),
                cache_capability: Some(PeerReplayCapability::Capable { boot_epoch }),
            },
        );
        clear_peer_replay_state_for_addr("http://clear-peer-replay-state-test:9000").expect("peer URI should clear replay state");
        assert_eq!(peer_replay_state(audience), PeerReplayState::default());
    }
    #[test]
    fn interceptor_snapshot_prevents_delayed_legacy_response_from_revoking_capability() {
        ensure_test_rpc_secret();
@@ -13,8 +13,9 @@
 // limitations under the License.
 use crate::cluster::rpc::client::{
-    AuthenticatedChannel, TonicInterceptor, embedded_tonic_status, gen_tonic_signature_interceptor, heal_control_time_out_client,
+    AuthenticatedChannel, TonicInterceptor, clear_peer_replay_state_for_addr, embedded_tonic_status,
-    is_network_like_status, message_has_network_needle, node_service_time_out_client, tier_mutation_control_time_out_client,
+    gen_tonic_signature_interceptor, heal_control_time_out_client, is_network_like_status, message_has_network_needle,
    node_service_time_out_client, tier_mutation_control_time_out_client,
 };
 use crate::cluster::rpc::{set_tonic_canonical_body_digest, set_tonic_mutation_body_digest, verify_tonic_rpc_response_proof};
 use crate::error::{Error, Result};
@@ -544,6 +545,16 @@ fn validate_heal_control_response_proof(canonical_response: &[u8], proof: &[u8])
        .map_err(|_| Error::other("peer returned an invalid heal control response proof"))
 }
 fn heal_control_auth_may_need_replay_scope_refresh(err: &Error) -> bool {
    matches!(
        err,
        Error::Io(io_err)
            if embedded_tonic_status(io_err).is_some_and(|status| {
                status.code() == tonic::Code::Unauthenticated && status.message() == "No valid auth token"
            })
    )
 }
 fn decode_remote_version_state_capability(expected_member: &str, result: &[u8]) -> Result<Uuid> {
    let (topology_member, process_epoch) = rustfs_protos::decode_remote_version_state_capability(result).map_err(Error::other)?;
    if topology_member != expected_member {
@@ -1720,45 +1731,72 @@ impl PeerRestClient {
            return Err(Error::other("heal control command exceeds size limit"));
        }
        let capability_probe = rustfs_protos::is_heal_control_capability_probe(&command);
-        self.finalize_result(
+        let result = self
-            async {
+            .heal_control_once(version, &topology_fingerprint, &command, capability_probe)
-                let mut client = self
+            .await;
-                    .get_heal_control_client()
+        if result
-                    .await?
+            .as_ref()
-                    .max_encoding_message_size(rustfs_protos::HEAL_CONTROL_RPC_MAX_MESSAGE_SIZE)
+            .err()
-                    .max_decoding_message_size(rustfs_protos::HEAL_CONTROL_RPC_MAX_MESSAGE_SIZE);
+            .is_some_and(heal_control_auth_may_need_replay_scope_refresh)
-                let canonical_body = rustfs_protos::canonical_heal_control_request_body(version, &topology_fingerprint, &command)
+        {
-                    .map_err(|_| Error::other("heal control request length cannot be represented"))?;
+            self.prepare_heal_control_auth_retry().await;
-                let mut request = Request::new(HealControlRequest {
+            return self
-                    version,
+                .finalize_result(
-                    topology_fingerprint: topology_fingerprint.clone(),
+                    self.heal_control_once(version, &topology_fingerprint, &command, capability_probe)
-                    command: command.clone().into(),
+                        .await,
-                });
+                )
-                request.set_timeout(rustfs_protos::heal_control_execution_timeout());
+                .await;
-                set_tonic_canonical_body_digest(&mut request, &canonical_body)?;
+        }
-                let response = client.heal_control(request).await?.into_inner();
+        self.finalize_result(result).await
-                if !response.success {
+    }
-                    return Err(Error::other(
+
-                        response
+    async fn prepare_heal_control_auth_retry(&self) {
-                            .error_info
+        if let Err(err) = clear_peer_replay_state_for_addr(&self.grid_host) {
-                            .unwrap_or_else(|| "peer heal control failed without an error".to_string()),
+            debug!(
-                    ));
+                peer = %self.grid_host,
-                }
+                error = %err,
-                if !capability_probe {
+                "could not clear heal control replay state before retry"
-                    let canonical_response = rustfs_protos::canonical_heal_control_response_body(
+            );
-                        version,
+        }
-                        &topology_fingerprint,
+        self.evict_connection().await;
-                        &command,
+    }
-                        &response.result,
+
-                    )
+    async fn heal_control_once(
        &self,
        version: u32,
        topology_fingerprint: &str,
        command: &[u8],
        capability_probe: bool,
    ) -> Result<Vec<u8>> {
        let mut client = self
            .get_heal_control_client()
            .await?
            .max_encoding_message_size(rustfs_protos::HEAL_CONTROL_RPC_MAX_MESSAGE_SIZE)
            .max_decoding_message_size(rustfs_protos::HEAL_CONTROL_RPC_MAX_MESSAGE_SIZE);
        let canonical_body = rustfs_protos::canonical_heal_control_request_body(version, topology_fingerprint, command)
            .map_err(|_| Error::other("heal control request length cannot be represented"))?;
        let mut request = Request::new(HealControlRequest {
            version,
            topology_fingerprint: topology_fingerprint.to_string(),
            command: command.to_vec().into(),
        });
        request.set_timeout(rustfs_protos::heal_control_execution_timeout());
        set_tonic_canonical_body_digest(&mut request, &canonical_body)?;
        let response = client.heal_control(request).await?.into_inner();
        if !response.success {
            return Err(Error::other(
                response
                    .error_info
                    .unwrap_or_else(|| "peer heal control failed without an error".to_string()),
            ));
        }
        if !capability_probe {
            let canonical_response =
                rustfs_protos::canonical_heal_control_response_body(version, topology_fingerprint, command, &response.result)
                    .map_err(|_| Error::other("heal control response length cannot be represented"))?;
-                    validate_heal_control_response_proof(&canonical_response, &response.response_proof)?;
+            validate_heal_control_response_proof(&canonical_response, &response.response_proof)?;
-                }
+        }
-                Ok(response.result.to_vec())
+        Ok(response.result.to_vec())
            }
            .await,
        )
        .await
    }
    /// Confirms that a peer supports the current heal-control coordination
@@ -3728,6 +3766,22 @@ mod tests {
        }
    }
    #[test]
    fn heal_control_auth_retry_is_limited_to_transport_auth_rejection() {
        assert!(heal_control_auth_may_need_replay_scope_refresh(&Error::from(
            tonic::Status::unauthenticated("No valid auth token")
        )));
        assert!(!heal_control_auth_may_need_replay_scope_refresh(&Error::from(
            tonic::Status::permission_denied("bad signature")
        )));
        assert!(!heal_control_auth_may_need_replay_scope_refresh(&Error::from(
            tonic::Status::unauthenticated("application rejected heal control")
        )));
        assert!(!heal_control_auth_may_need_replay_scope_refresh(&Error::other(
            "Io error: code: 'Unauthenticated', message: \"No valid auth token\""
        )));
    }
    #[test]
    fn peer_rest_client_network_classifier_keeps_slow_peers_online() {
        // The per-RPC channel deadline (RUSTFS_INTERNODE_RPC_TIMEOUT, 30s)
@@ -3463,11 +3463,6 @@ impl PoolRebalanceActivationFence {
    }
 }
 #[cfg(test)]
 tokio::task_local! {
    pub(crate) static REBALANCE_ACTIVATION_LOCK_ATTEMPT: Arc<tokio::sync::Notify>;
 }
 pub(crate) async fn acquire_pool_rebalance_activation_locks<S>(
    pool: Arc<S>,
    fleet_proof: Option<crate::services::notification_sys::CrossPoolFenceFleetProofToken>,
@@ -3478,21 +3473,17 @@ where
            NamespaceLock = rustfs_lock::NamespaceLockWrapper,
        >,
 {
-    // Match entry admission: rebalance.bin -> pool.bin. An entry retains its
+    // Activation lock order is always pool.bin -> rebalance.bin.
    // run read fence while target mutations acquire the pool metadata fence;
    // activation must not hold pool.bin while waiting for that entry to drain.
    let rebalance_meta_lock = pool.new_ns_lock(RUSTFS_META_BUCKET, REBAL_META_NAME).await?;
    #[cfg(test)]
    let _ = REBALANCE_ACTIVATION_LOCK_ATTEMPT.try_with(|attempted| attempted.notify_one());
    let rebalance_meta_guard = rebalance_meta_lock
        .get_write_lock(get_lock_acquire_timeout())
        .await
        .map_err(activation_rebalance_meta_lock_error)?;
    let pool_meta_lock = pool.new_ns_lock(RUSTFS_META_BUCKET, POOL_META_NAME).await?;
    let pool_meta_guard = pool_meta_lock
        .get_write_lock(get_lock_acquire_timeout())
        .await
        .map_err(activation_pool_meta_lock_error)?;
    let rebalance_meta_lock = pool.new_ns_lock(RUSTFS_META_BUCKET, REBAL_META_NAME).await?;
    let rebalance_meta_guard = rebalance_meta_lock
        .get_write_lock(get_lock_acquire_timeout())
        .await
        .map_err(activation_rebalance_meta_lock_error)?;
    Ok(PoolRebalanceActivationFence {
        pool_meta_guard,
@@ -22103,7 +22094,7 @@ mod pools_tests {
                .resources
                .lock()
                .expect("activation lock recorder should not be poisoned"),
-            vec![REBAL_META_NAME.to_string(), POOL_META_NAME.to_string()]
+            vec![POOL_META_NAME.to_string(), REBAL_META_NAME.to_string()]
        );
        let mut second_acquire = Box::pin(acquire_pool_rebalance_activation_locks(second.clone(), None));
@@ -22119,50 +22110,10 @@ mod pools_tests {
                .resources
                .lock()
                .expect("activation lock recorder should not be poisoned"),
-            vec![REBAL_META_NAME.to_string(), POOL_META_NAME.to_string()]
+            vec![POOL_META_NAME.to_string(), REBAL_META_NAME.to_string()]
        );
    }
    #[tokio::test]
    async fn test_activation_cancellation_releases_rebalance_fence_while_pool_fence_is_contended() {
        use crate::storage_api_contracts::namespace::NamespaceLocking as _;
        let pool = Arc::new(ActivationLockRecorder {
            lock_manager: Arc::new(rustfs_lock::GlobalLockManager::new()),
            owner: "activation-cancellation",
            resources: StdMutex::new(Vec::new()),
        });
        let pool_lock = pool
            .new_ns_lock(crate::disk::RUSTFS_META_BUCKET, POOL_META_NAME)
            .await
            .expect("pool lock should be created");
        let pool_reader = pool_lock
            .get_read_lock(std::time::Duration::from_secs(5))
            .await
            .expect("ordinary mutation should hold the pool read fence");
        pool.resources.lock().expect("recorder should not be poisoned").clear();
        let mut activation = Box::pin(acquire_pool_rebalance_activation_locks(Arc::clone(&pool), None));
        assert!(matches!(futures::poll!(&mut activation), Poll::Pending));
        assert_eq!(
            *pool.resources.lock().expect("recorder should not be poisoned"),
            vec![REBAL_META_NAME.to_string(), POOL_META_NAME.to_string()],
            "activation must hold the run fence before waiting for the pool fence",
        );
        drop(activation);
        let rebalance_lock = pool
            .new_ns_lock(crate::disk::RUSTFS_META_BUCKET, REBAL_META_NAME)
            .await
            .expect("run lock should be created");
        let run_writer = rebalance_lock
            .get_write_lock(std::time::Duration::from_secs(5))
            .await
            .expect("cancelling activation must release its already-acquired run fence");
        assert!(
            !pool_reader.is_released(),
            "cancelling activation must not release another caller's pool fence"
        );
        assert!(!run_writer.is_lock_lost());
    }
    #[test]
    fn decommission_receipt_run_token_changes_with_persisted_start_time() {
        let first = OffsetDateTime::from_unix_timestamp(1_000).expect("first run timestamp should be valid");
@@ -627,13 +627,7 @@ impl From<tokio::task::JoinError> for DiskError {
 impl Clone for DiskError {
    fn clone(&self) -> Self {
        match self {
-            DiskError::Io(io_error) => DiskError::Io(
+            DiskError::Io(io_error) => DiskError::Io(std::io::Error::new(io_error.kind(), io_error.to_string())),
                rustfs_rio::clone_internode_http_io_error(io_error)
                    .and_then(std::io::Error::into_inner)
                    // The helper derives a kind from the source; Clone must retain the original outer kind.
                    .map(|source| std::io::Error::new(io_error.kind(), source))
                    .unwrap_or_else(|| std::io::Error::new(io_error.kind(), io_error.to_string())),
            ),
            DiskError::MaxVersionsExceeded => DiskError::MaxVersionsExceeded,
            DiskError::Unexpected => DiskError::Unexpected,
            DiskError::CorruptedFormat => DiskError::CorruptedFormat,
@@ -1271,49 +1265,6 @@ mod tests {
        assert!(!bad_request.is_retryable_internode_write_failure());
    }
    #[test]
    fn test_internode_http_clone_preserves_retryability_status_and_context() {
        use http::StatusCode;
        use rustfs_rio::InternodeHttpErrorKind::{ConnectionRefused, ConnectionReset, HttpStatus, Unknown};
        for (kind, retryable) in [
            (ConnectionRefused, true),
            (ConnectionReset, true),
            (HttpStatus(StatusCode::TOO_MANY_REQUESTS), true),
            (HttpStatus(StatusCode::SERVICE_UNAVAILABLE), true),
            (HttpStatus(StatusCode::CONFLICT), true),
            (Unknown, false),
            (HttpStatus(StatusCode::BAD_REQUEST), false),
            (HttpStatus(StatusCode::INTERNAL_SERVER_ERROR), false),
        ] {
            let original = DiskError::from(rustfs_rio::new_test_internode_http_io_error(kind));
            assert_eq!(original.internode_http_error_kind(), Some(kind));
            assert_eq!(original.is_retryable_internode_write_failure(), retryable);
            let cloned = original.clone();
            assert_eq!(cloned, original, "clone must preserve the error bucket for {kind:?}");
            assert_eq!(
                cloned.is_retryable_internode_write_failure(),
                retryable,
                "clone changed retryability for {kind:?}"
            );
            assert_eq!(cloned.internode_http_error_kind(), Some(kind));
            if let HttpStatus(status) = kind {
                assert!(cloned.is_internode_http_status(status.as_u16()));
            }
            let DiskError::Io(io_error) = &cloned else {
                panic!("unmarked internode error must remain Io: {cloned:?}");
            };
            let source = io_error
                .get_ref()
                .and_then(|source| source.downcast_ref::<InternodeHttpError>())
                .expect("clone must retain the structured internode error");
            assert_eq!(source.context().method(), "PUT");
            assert_eq!(source.context().target(), "/rustfs/rpc/put_file_stream");
            assert_eq!(source.context().operation(), Some(INTERNODE_OPERATION_PUT_FILE_STREAM));
        }
    }
    #[tokio::test]
    async fn read_stream_conflict_is_not_a_retryable_put_file_failure() {
        use tokio::io::{AsyncReadExt, AsyncWriteExt};
@@ -1358,57 +1309,11 @@ mod tests {
                !error.is_retryable_internode_write_failure(),
                "read-operation 409 must not trigger put-file retry"
            );
            let cloned = error.clone();
            let reduced = crate::disk::error_reduce::reduce_write_quorum_errs(&[Some(error)], &[], 1)
                .expect("the read conflict must remain the dominant error");
            for preserved in [&cloned, &reduced] {
                assert!(
                    !preserved.is_retryable_internode_write_failure(),
                    "cloning or reducing a read conflict must not turn it into a PUT retry"
                );
                assert!(preserved.is_internode_http_status(409));
                let DiskError::Io(io_error) = preserved else {
                    panic!("read conflict must remain Io: {preserved:?}");
                };
                let source = io_error
                    .get_ref()
                    .and_then(|source| source.downcast_ref::<InternodeHttpError>())
                    .expect("read conflict must retain its request context");
                assert_eq!(source.context().method(), "GET");
                assert_eq!(source.context().target(), "/rustfs/rpc/read_file_stream");
                assert_eq!(
                    source.context().operation(),
                    Some(rustfs_io_metrics::internode_metrics::INTERNODE_OPERATION_READ_FILE_STREAM)
                );
            }
        })
        .await
        .expect("isolated read-conflict test must finish within its budget");
    }
    #[test]
    fn test_internode_http_clone_preserves_outer_io_kind_and_message() {
        let source = rustfs_rio::new_test_internode_http_io_error(InternodeHttpErrorKind::ConnectionReset)
            .into_inner()
            .expect("the internode helper must provide a typed source");
        let original_io = io::Error::new(io::ErrorKind::InvalidData, source);
        let message = original_io.to_string();
        let original = DiskError::from(original_io);
        assert_eq!(original.internode_http_error_kind(), Some(InternodeHttpErrorKind::ConnectionReset));
        assert!(original.is_retryable_internode_write_failure());
        let cloned = original.clone();
        let reduced = crate::disk::error_reduce::reduce_write_quorum_errs(&[Some(original)], &[], 1)
            .expect("the wrapped internode error must remain the dominant error");
        for preserved in [&cloned, &reduced] {
            let DiskError::Io(io_error) = preserved else {
                panic!("the wrapped error must remain Io: {preserved:?}");
            };
            assert_eq!(io_error.kind(), io::ErrorKind::InvalidData);
            assert_eq!(io_error.to_string(), message);
        }
    }
    #[test]
    fn test_internode_missing_errors_preserve_disk_error_types() {
        let file_missing = DiskError::from(rustfs_rio::new_test_remote_file_not_found_http_io_error());
@@ -1420,17 +1325,6 @@ mod tests {
        assert_eq!(file_missing, DiskError::FileNotFound);
        assert_eq!(volume_missing, DiskError::VolumeNotFound);
        assert!(matches!(unmarked_server_error, DiskError::Io(_)));
        for missing in [file_missing, volume_missing] {
            assert_eq!(missing.clone(), missing);
            assert_eq!(
                crate::disk::error_reduce::reduce_write_quorum_errs(
                    &[Some(missing.clone()), Some(missing.clone()), None],
                    &[],
                    2
                ),
                Some(missing)
            );
        }
    }
    #[test]
@@ -226,78 +226,6 @@ mod tests {
        assert_eq!(res, Some(quorum_err));
    }
    #[test]
    fn test_write_quorum_reduction_preserves_internode_http_identity() {
        use http::StatusCode;
        use rustfs_rio::InternodeHttpErrorKind::{ConnectionRefused, HttpStatus, Unknown};
        for (kind, retryable) in [
            (ConnectionRefused, true),
            (HttpStatus(StatusCode::SERVICE_UNAVAILABLE), true),
            (HttpStatus(StatusCode::CONFLICT), true),
            (Unknown, false),
            (HttpStatus(StatusCode::BAD_REQUEST), false),
        ] {
            // Construct both producer errors independently: the reducer owns the first clone.
            let first = Error::from(rustfs_rio::new_test_internode_http_io_error(kind));
            let second = Error::from(rustfs_rio::new_test_internode_http_io_error(kind));
            assert_eq!(first.internode_http_error_kind(), Some(kind));
            assert_eq!(second.internode_http_error_kind(), Some(kind));
            assert_eq!(first.is_retryable_internode_write_failure(), retryable);
            let errors = [Some(first), Some(second), None];
            let reduced = reduce_write_quorum_errs(&errors, OBJECT_OP_IGNORED_ERRS, 2)
                .expect("two equal producer errors must dominate one successful write");
            assert_eq!(Some(&reduced), errors[0].as_ref());
            assert_eq!(
                reduced.is_retryable_internode_write_failure(),
                retryable,
                "quorum reduction changed retryability for {kind:?}"
            );
            assert_eq!(reduced.internode_http_error_kind(), Some(kind));
            if let HttpStatus(status) = kind {
                assert!(reduced.is_internode_http_status(status.as_u16()));
            }
            let Error::Io(io_error) = &reduced else {
                panic!("the dominant error must remain Io: {reduced:?}");
            };
            let source = io_error
                .get_ref()
                .and_then(|source| source.downcast_ref::<rustfs_rio::InternodeHttpError>())
                .expect("quorum reduction must retain the structured internode error");
            assert_eq!(source.context().method(), "PUT");
            assert_eq!(source.context().target(), "/rustfs/rpc/put_file_stream");
            assert_eq!(
                source.context().operation(),
                Some(rustfs_io_metrics::internode_metrics::INTERNODE_OPERATION_PUT_FILE_STREAM)
            );
        }
    }
    #[test]
    fn test_clone_and_write_quorum_do_not_promote_non_retryable_errors() {
        use http::StatusCode;
        use rustfs_rio::InternodeHttpErrorKind::{HttpStatus, Unknown};
        for original in [
            Error::from(rustfs_rio::new_test_internode_http_io_error(Unknown)),
            Error::from(rustfs_rio::new_test_internode_http_io_error(HttpStatus(StatusCode::BAD_REQUEST))),
            Error::from(rustfs_rio::new_test_internode_http_io_error(HttpStatus(StatusCode::FORBIDDEN))),
            Error::from(rustfs_rio::new_test_internode_http_io_error(HttpStatus(StatusCode::NOT_FOUND))),
            Error::from(rustfs_rio::new_test_internode_http_io_error(HttpStatus(
                StatusCode::INTERNAL_SERVER_ERROR,
            ))),
            err_io("internode connection reset: PUT /rustfs/rpc/put_file_stream"),
        ] {
            assert!(!original.is_retryable_internode_write_failure());
            let cloned = original.clone();
            let reduced =
                reduce_write_quorum_errs(&[Some(original)], &[], 1).expect("a non-retryable error must remain an error");
            assert!(!cloned.is_retryable_internode_write_failure());
            assert!(!reduced.is_retryable_internode_write_failure());
        }
    }
    #[test]
    fn test_count_errs() {
        let e1 = err_io("a");
@@ -22463,6 +22463,86 @@ mod test {
        );
    }
    #[cfg(unix)]
    #[tokio::test]
    async fn conditional_mrf_manifest_dir_fsync_failure_keeps_recovery_anchors() {
        use tempfile::tempdir;
        const MRF_COMMIT_MANIFEST_SLOT_0: &str = ".heal-mrf-commit.0.bin";
        const MRF_COMMIT_MANIFEST_SLOT_1: &str = ".heal-mrf-commit.1.bin";
        const MRF_SCOPED_JOURNAL_PATH: &str = "buckets/.heal/mrf/journal-scoped.bin";
        let _mode = durability_mode_override::set(DurabilityMode::Relaxed);
        let dir = tempdir().expect("temp dir should be created");
        let endpoint = Endpoint::try_from(dir.path().to_str().expect("temp dir should be utf8")).expect("endpoint should parse");
        let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created");
        let previous_manifest = Bytes::from_static(b"mrf-committed-manifest-v1");
        let successor_manifest = Bytes::from_static(b"mrf-committed-manifest-v2");
        let legacy_journal = Bytes::from_static(b"legacy-mrf-journal-records");
        assert_eq!(
            disk.compare_and_update_file(RUSTFS_META_BUCKET, MRF_COMMIT_MANIFEST_SLOT_0, None, Some(previous_manifest.clone()),)
                .await
                .expect("previous MRF manifest should commit"),
            ConditionalFileUpdate::Updated
        );
        disk.write_all(RUSTFS_META_BUCKET, MRF_SCOPED_JOURNAL_PATH, legacy_journal.clone())
            .await
            .expect("legacy MRF journal should be retained");
        let manifest_path = disk
            .get_object_path(RUSTFS_META_BUCKET, MRF_COMMIT_MANIFEST_SLOT_0)
            .expect("MRF manifest path should resolve");
        let parent = manifest_path.parent().expect("MRF manifest path should have a parent");
        assert!(
            os::fsync_dir_recorder::was_fsynced(parent),
            "system metadata MRF manifest publication must fsync the metadata directory even under relaxed durability"
        );
        os::fsync_dir_recorder::set_failure(parent, ErrorKind::Other);
        let err = disk
            .compare_and_update_file(
                RUSTFS_META_BUCKET,
                MRF_COMMIT_MANIFEST_SLOT_0,
                Some(previous_manifest.clone()),
                Some(successor_manifest),
            )
            .await
            .expect_err("directory fsync failure must fail the MRF manifest successor commit");
        assert!(matches!(err, DiskError::Io(ref err) if err.kind() == ErrorKind::Other));
        assert_eq!(
            disk.read_all(RUSTFS_META_BUCKET, MRF_COMMIT_MANIFEST_SLOT_0)
                .await
                .expect("previous committed MRF manifest should remain readable after rollback"),
            previous_manifest
        );
        os::fsync_dir_recorder::set_failure(parent, ErrorKind::Other);
        let err = disk
            .compare_and_update_file(
                RUSTFS_META_BUCKET,
                MRF_COMMIT_MANIFEST_SLOT_1,
                None,
                Some(Bytes::from_static(b"first-successor-manifest")),
            )
            .await
            .expect_err("directory fsync failure must fail first MRF manifest commit");
        assert!(matches!(err, DiskError::Io(ref err) if err.kind() == ErrorKind::Other));
        assert!(
            matches!(
                disk.read_all(RUSTFS_META_BUCKET, MRF_COMMIT_MANIFEST_SLOT_1).await,
                Err(DiskError::FileNotFound)
            ),
            "uncommitted first MRF manifest must be removed when no committed anchor exists"
        );
        assert_eq!(
            disk.read_all(RUSTFS_META_BUCKET, MRF_SCOPED_JOURNAL_PATH)
                .await
                .expect("legacy MRF journal should remain readable after failed manifest publication"),
            legacy_journal
        );
    }
    #[cfg(unix)]
    #[tokio::test]
    async fn conditional_file_update_dir_fsync_failure_removes_new_file_without_anchor() {
@@ -570,11 +570,25 @@ pub(crate) async fn initialize_local_disk_maps(
    Ok(())
 }
 #[cfg(test)]
 tokio::task_local! {
    pub(crate) static TEST_TIER_CONFIG_MGR: Arc<tokio::sync::RwLock<TierConfigMgr>>;
 }
 pub(crate) async fn init_tier_config_mgr(store: Arc<ECStore>) -> Result<()> {
    #[cfg(not(test))]
    let handle = get_global_tier_config_mgr();
-    TierConfigMgr::reload_handle(&handle, store.clone()).await?;
+    #[cfg(test)]
    let handle = TEST_TIER_CONFIG_MGR
        .try_with(Arc::clone)
        .unwrap_or_else(|_| get_global_tier_config_mgr());
    let initial_reload = TierConfigMgr::reload_handle(&handle, store.clone()).await;
    if initial_reload.is_err() {
        // Keep local recovery active when the initial snapshot cannot be loaded.
        TierConfigMgr::request_committed_mutation_refresh(&handle).await;
    }
    tokio::spawn(TierConfigMgr::refresh_tier_config_handle(handle, store));
-    Ok(())
+    initial_reload.map_err(Error::from)
 }
 #[cfg(test)]
@@ -572,7 +572,7 @@ impl ECStore {
    where
        S: EcstoreObjectIO + StorageNamespaceLocking<Error = Error, NamespaceLock = rustfs_lock::NamespaceLockWrapper>,
    {
-        // Lock order: pool_meta_save_gate -> rebalance.bin -> pool.bin.
+        // Lock order: pool_meta_save_gate -> pool.bin -> rebalance.bin.
        let mut pool_meta_guard = self.pool_meta_save_gate.lock().await;
        pool_meta_guard.ensure_write_safe("rebalance worker activation")?;
        // Classify the durable rebalance record while holding both namespace
@@ -50,11 +50,6 @@ fn ensure_rebalance_entry_active(cancel: &CancellationToken) -> Result<()> {
    Ok(())
 }
 #[cfg(test)]
 tokio::task_local! {
    static REBALANCE_ENTRY_RUN_FENCE_BARRIER: (Arc<tokio::sync::Notify>, Arc<tokio::sync::Notify>);
 }
 #[derive(Debug)]
 struct RebalanceEntryTarget {
    bucket: String,
@@ -261,15 +256,9 @@ impl ECStore {
            .sort_by_key(|v| (v.mod_time.is_none(), std::cmp::Reverse(v.mod_time)));
        // Entry lock order is bucket incarnation -> activation_gate -> rebalance.bin -> movement gate.
        // Target capacity admission can then acquire pool.bin under the run fence.
        // Stop waits for in-flight entries through cleanup, but not for entries admitted later.
        ensure_rebalance_entry_active(&cancel)?;
        let run_guard = self.rebalance_run_guard(rebalance_id.as_ref(), "rebalance entry").await?;
        #[cfg(test)]
        if let Ok((arrived, release)) = REBALANCE_ENTRY_RUN_FENCE_BARRIER.try_with(Clone::clone) {
            arrived.notify_one();
            release.notified().await;
        }
        let lock_lost_signal = run_guard.lock_lost_signal();
        #[cfg(test)]
        let _run_signal_test_fence = lock_lost_signal
@@ -1248,130 +1237,6 @@ mod tests {
        assert_eq!(pool_stats.cleanup_warnings.count, 1, "deferred cleanup must not add a permanent warning");
    }
    #[tokio::test]
    #[serial_test::serial]
    async fn real_rebalance_entry_progresses_while_peer_activation_waits_for_run_fence() {
        const REBALANCE_ID: &str = "rebalance-peer-activation-lock-order";
        let (_temp_dirs, store, peer) = crate::services::rebalance::test_two_pool_stores_with_isolated_node_contexts(Some(
            active_rebalance_meta(REBALANCE_ID),
        ))
        .await;
        assert!(!Arc::ptr_eq(&store.ctx, &peer.ctx), "node-local movement gates must be independent");
        {
            let mut meta = peer.rebalance_meta.write().await;
            let meta = meta.as_mut().expect("peer should know the durable run");
            meta.activation_gate = Arc::default();
            meta.cancel = None;
        }
        let bucket = crate::disk::RUSTFS_META_BUCKET;
        let object = "rebalance-peer-activation-object";
        let version_id = uuid::Uuid::new_v4();
        let payload = b"entry must drain before peer activation takes the pool fence".repeat(1024);
        let source_set = store.pools[0].get_disks_by_key(object);
        let target_set = store.pools[1].get_disks_by_key(object);
        let opts = ObjectOptions {
            versioned: true,
            version_id: Some(version_id.to_string()),
            ..Default::default()
        };
        let mut writer = PutObjReader::from_vec(payload.clone());
        let source_before = source_set
            .put_object(bucket, object, &mut writer, &opts)
            .await
            .expect("source version should be written");
        let entry = metacache_entry_from_source(&source_set, bucket, object).await;
        let arrived = Arc::new(tokio::sync::Notify::new());
        let release = Arc::new(tokio::sync::Notify::new());
        // JoinSet aborts both scoped tasks if an assertion or timeout fails.
        let mut tasks = tokio::task::JoinSet::new();
        let entry_store = Arc::clone(&store);
        tasks.spawn(
            REBALANCE_ENTRY_RUN_FENCE_BARRIER.scope((Arc::clone(&arrived), Arc::clone(&release)), async move {
                entry_store
                    .rebalance_entry(
                        RebalanceEntryTarget {
                            bucket: bucket.to_string(),
                            pool_index: 0,
                        },
                        entry,
                        source_set,
                        Arc::new(RebalanceBucketConfigs::default()),
                        Arc::from(REBALANCE_ID),
                        CancellationToken::new(),
                    )
                    .await
            }),
        );
        tokio::time::timeout(StdDuration::from_secs(30), arrived.notified())
            .await
            .expect("real entry must acquire its persisted run read fence");
        let attempted = Arc::new(tokio::sync::Notify::new());
        let peer_pool = Arc::clone(&peer.pools[0]);
        let (activation_done, activation_result) = tokio::sync::oneshot::channel();
        tasks.spawn(
            crate::core::pools::REBALANCE_ACTIVATION_LOCK_ATTEMPT.scope(Arc::clone(&attempted), async move {
                let result = peer.fence_rebalance_worker_activation(peer_pool, REBALANCE_ID).await;
                let result = result.map(|fence| match fence {
                    super::super::control::RebalanceWorkerActivationFence::Ready(fence) => {
                        fence.ensure_held().expect("peer activation must retain both fences");
                    }
                    super::super::control::RebalanceWorkerActivationFence::NotStartedTerminal => {
                        panic!("the paused entry's run must still require activation");
                    }
                });
                activation_done.send(result).expect("activation receiver should remain alive");
                Ok(RebalanceEntryOutcome::Completed)
            }),
        );
        tokio::time::timeout(StdDuration::from_secs(30), attempted.notified())
            .await
            .expect("peer activation must attempt the persisted rebalance write fence");
        release.notify_one();
        tokio::time::timeout(StdDuration::from_secs(30), async {
            while let Some(result) = tasks.join_next().await {
                assert!(matches!(
                    result
                        .expect("scoped task must not panic")
                        .expect("entry must not fail or defer"),
                    RebalanceEntryOutcome::Completed
                ));
            }
        })
        .await
        .expect("entry and peer activation must both make progress");
        activation_result
            .await
            .expect("peer activation result should be sent")
            .expect("peer activation must not time out behind the entry it blocks");
        let mut reader = target_set
            .get_object_reader(bucket, object, None, HeaderMap::new(), &opts)
            .await
            .expect("the exact target version must be readable");
        let mut actual = Vec::new();
        reader
            .stream
            .read_to_end(&mut actual)
            .await
            .expect("target body should drain completely");
        assert_eq!(actual, payload);
        assert_eq!(reader.object_info.version_id, source_before.version_id);
        assert_eq!(reader.object_info.etag, source_before.etag);
        assert_eq!(reader.object_info.mod_time, source_before.mod_time);
        let source_error = store.pools[0]
            .get_object_info(bucket, object, &opts)
            .await
            .expect_err("completed entry must clean up the source version");
        assert!(crate::error::is_err_object_not_found(&source_error) || crate::error::is_err_version_not_found(&source_error));
        let meta = store.rebalance_meta.read().await;
        let stats = &meta.as_ref().expect("local run must remain installed").pool_stats[0];
        assert_eq!(stats.num_objects, 1);
        assert_eq!(stats.num_versions, 1);
        assert_eq!(stats.cleanup_warnings.count, 0);
    }
    #[tokio::test]
    #[serial_test::serial]
    async fn real_rebalance_run_fence_loss_before_target_commit_preserves_target_and_source() {
@@ -1907,124 +1907,6 @@ fn test_is_transient_rebalance_error_accepts_wrapped_disk_timeout() {
    assert!(is_transient_rebalance_error(&Error::Io(std::io::Error::other(DiskError::Timeout))));
 }
 #[test]
 fn test_rebalance_stage_wrapped_transient_errors_remain_retryable() {
    let cases = [
        Error::Lock(rustfs_lock::LockError::timeout(".rustfs.sys/pool.bin@latest", Duration::from_secs(5))),
        Error::Lock(rustfs_lock::LockError::network(
            "peer unavailable",
            std::io::Error::from(std::io::ErrorKind::ConnectionReset),
        )),
        Error::SlowDown,
        Error::ErasureReadQuorum,
        Error::ErasureWriteQuorum,
        Error::Io(std::io::Error::other(DiskError::Timeout)),
        Error::Io(std::io::Error::from(std::io::ErrorKind::TimedOut)),
    ];
    for mut error in cases {
        for depth in 0..=3 {
            assert!(is_transient_rebalance_error(&error), "transient source lost at depth {depth}: {error:?}");
            assert!(
                should_defer_rebalance_entry_failure(&error),
                "exhausted transient entries must be deferred"
            );
            assert!(should_retry_rebalance_listing(&error, 0, 3));
            assert!(
                !should_retry_rebalance_listing(&error, 2, 3),
                "wrapping must not bypass the attempt limit"
            );
            error = data_movement::data_movement_stage_error_for_test(
                "rebalance_object",
                "put_object",
                "bucket",
                "baseline/00042.bin",
                error,
            );
        }
    }
 }
 #[test]
 fn test_rebalance_stage_wrapped_terminal_errors_remain_terminal() {
    let cases = [
        Error::FileAccessDenied,
        Error::FileCorrupt,
        Error::OperationCanceled,
        Error::DataMovementOverwriteErr("bucket".to_string(), "object".to_string(), "version".to_string()),
        Error::Lock(rustfs_lock::LockError::already_locked("bucket/object", "owner")),
        Error::other("permission denied"),
    ];
    for mut error in cases {
        for depth in 0..=3 {
            assert!(
                !is_transient_rebalance_error(&error),
                "terminal source must survive depth {depth}: {error:?}"
            );
            assert!(!should_defer_rebalance_entry_failure(&error));
            // Object names are untrusted context, not evidence of a transient failure.
            error = data_movement::data_movement_stage_error_for_test(
                "rebalance_object",
                "put_object",
                "bucket",
                "remote lock rpc timed out",
                error,
            );
        }
    }
 }
 #[tokio::test]
 async fn test_rebalance_stage_wrapped_lock_timeout_retries_real_migration_loop() {
    for succeeds_on_retry in [true, false] {
        let backend = MigrationBackendSpy::new(None, None);
        let attempts = AtomicUsize::new(0);
        let waits = AtomicUsize::new(0);
        let mut transfer = |_, _, _| {
            let attempt = attempts.fetch_add(1, Ordering::SeqCst);
            async move {
                if succeeds_on_retry && attempt > 0 {
                    return Ok(());
                }
                Err(data_movement::data_movement_stage_error_for_test(
                    "rebalance_object",
                    "put_object",
                    "bucket",
                    "baseline/00042.bin",
                    Error::Lock(rustfs_lock::LockError::timeout(".rustfs.sys/pool.bin@latest", Duration::from_secs(5))),
                ))
            }
        };
        let version = version_normal();
        let result = migrate_entry_version_with_retry_wait(
            &backend,
            "bucket".to_string(),
            0,
            &version,
            None,
            3,
            false,
            &mut transfer,
            |_: String, _: String, _: ObjectOptions| async { Ok::<_, Error>(ObjectInfo::default()) },
            |_| {
                waits.fetch_add(1, Ordering::SeqCst);
                std::future::ready(())
            },
        )
        .await;
        assert_eq!(result.moved, succeeds_on_retry);
        assert_eq!(result.failed, !succeeds_on_retry);
        assert_eq!(attempts.load(Ordering::SeqCst), if succeeds_on_retry { 2 } else { 3 });
        assert_eq!(backend.get_calls(), attempts.load(Ordering::SeqCst));
        assert_eq!(waits.load(Ordering::SeqCst), attempts.load(Ordering::SeqCst) - 1);
        if !succeeds_on_retry {
            assert_eq!(result.stage, Some("write_target"));
            assert!(should_defer_rebalance_entry_failure(
                result.error.as_ref().expect("exhaustion must retain its source error")
            ));
        }
    }
 }
 #[test]
 fn test_is_transient_rebalance_error_accepts_io_timeout_message() {
    assert!(is_transient_rebalance_error(&Error::Io(std::io::Error::other("timeout"))));
@@ -244,7 +244,6 @@ pub(super) fn resolve_rebalance_bucket_result(
 }
 pub(super) fn is_transient_rebalance_error(err: &Error) -> bool {
    let err = rebalance_error_source(err);
    match err {
        Error::SlowDown
        | Error::ErasureReadQuorum
@@ -257,15 +256,6 @@ pub(super) fn is_transient_rebalance_error(err: &Error) -> bool {
    }
 }
 fn rebalance_error_source(mut err: &Error) -> &Error {
    // Stage context contains object names, so classify the preserved source,
    // not timeout-like text supplied by an object name. Iterate nested stages.
    while let Some(source) = crate::data_movement::data_movement_stage_source(err) {
        err = source;
    }
    err
 }
 fn is_rebalance_transient_lock_error(err: &rustfs_lock::LockError) -> bool {
    match err {
        rustfs_lock::LockError::Timeout { .. } | rustfs_lock::LockError::Network { .. } => true,
@@ -319,7 +309,6 @@ pub(super) fn rebalance_listing_retry_delay(attempt: usize) -> Duration {
 }
 fn is_rebalance_lock_or_rpc_timeout(err: &Error) -> bool {
    let err = rebalance_error_source(err);
    match err {
        Error::Lock(rustfs_lock::LockError::Timeout { .. }) | Error::Lock(rustfs_lock::LockError::Network { .. }) => true,
        Error::Io(io_err) => is_rebalance_lock_or_rpc_timeout_message(&io_err.to_string()),
@@ -596,48 +585,3 @@ impl SetDisks {
        Ok(())
    }
 }
 #[cfg(test)]
 mod error_source_tests {
    use super::*;
    #[test]
    fn stage_wrapped_errors_select_the_source_backoff_policy() {
        let cases = [
            (
                Error::Lock(rustfs_lock::LockError::timeout(".rustfs.sys/pool.bin@latest", Duration::from_secs(5))),
                true,
            ),
            (
                Error::Lock(rustfs_lock::LockError::network(
                    "peer unavailable",
                    std::io::Error::from(std::io::ErrorKind::ConnectionReset),
                )),
                true,
            ),
            (Error::other("remote lock rpc timed out"), true),
            (Error::SlowDown, false),
            (Error::Io(std::io::Error::other(DiskError::Timeout)), false),
            (Error::FileAccessDenied, false),
        ];
        for (mut error, lock_backoff) in cases {
            for depth in 0..=3 {
                assert_eq!(
                    is_rebalance_lock_or_rpc_timeout(&error),
                    lock_backoff,
                    "wrong backoff at depth {depth}: {error:?}"
                );
                if !lock_backoff {
                    assert_eq!(rebalance_migration_retry_delay(1, &error), REBALANCE_MIGRATION_RETRY_BASE_DELAY * 2);
                }
                error = crate::data_movement::data_movement_stage_error_for_test(
                    "rebalance_object",
                    "put_object",
                    "bucket",
                    "remote lock rpc timed out",
                    error,
                );
            }
        }
    }
 }
@@ -326,15 +326,14 @@ impl SetDisks {
        let parity_blocks = Self::common_parity(&parities, default_parity_count as i32);
        if parity_blocks < 0 {
-            // A consistent layout can require more replies than the initial
+            // No parity value reached read quorum. Distinguish two cases:
-            // half-set probe. Reaching that probe alone is not corruption;
+            // enough disks answered with valid-looking metadata that simply
-            // only invalid or conflicting healthy replies establish that.
+            // cannot be reconciled (corrupt/foreign entries — retrying cannot
            // help, and heal should see Corrupt, rustfs#5801) versus too few
            // healthy answers (a genuine quorum condition where retry may
            // succeed once disks recover).
            let healthy_replies = errs.iter().filter(|err| err.is_none()).count();
-            let consistent_parity = parities
+            if healthy_replies >= expected_rquorum {
                .iter()
                .find(|&&parity| parity >= 0)
                .filter(|&&parity| parities.iter().filter(|&&candidate| candidate == parity).count() == healthy_replies);
            if healthy_replies >= expected_rquorum && consistent_parity.is_none() {
                error!(
                    "object_quorum_from_meta: irreconcilable parity across {healthy_replies} healthy replies (corrupt metadata), errs={errs:?}"
                );
@@ -1653,40 +1652,6 @@ mod tests {
        assert_eq!(err, DiskError::FileCorrupt);
    }
    #[test]
    fn consistent_parity_below_its_data_shard_quorum_is_not_corruption() {
        for (drive_count, parity) in [(6, 2), (8, 2), (12, 4)] {
            let data = drive_count - parity;
            let mut metas = (1..=drive_count)
                .map(|index| {
                    let mut info = FileInfo::new("bucket/object", data, parity);
                    info.size = 1024;
                    info.erasure.index = index;
                    info
                })
                .collect::<Vec<_>>();
            let mut errs = vec![Some(DiskError::DiskNotFound); drive_count];
            errs[..data].fill(None);
            assert_eq!(
                SetDisks::object_quorum_from_meta(&metas, &errs, parity).expect("exact data quorum should resolve"),
                (data as i32, data as i32)
            );
            errs[data - 1] = Some(DiskError::DiskNotFound);
            assert_eq!(
                SetDisks::object_quorum_from_meta(&metas, &errs, parity).expect_err("one fewer shard cannot resolve"),
                DiskError::ErasureReadQuorum,
                "layout {drive_count}/{parity} has consistent metadata but insufficient shards"
            );
            metas[0].erasure.parity_blocks = usize::MAX;
            assert_eq!(
                SetDisks::object_quorum_from_meta(&metas, &errs, parity).expect_err("corrupt healthy replies must be rejected"),
                DiskError::FileCorrupt
            );
        }
    }
    /// Too few healthy replies remains a genuine quorum condition where a
    /// retry may succeed once disks recover.
    #[test]
@@ -865,7 +865,6 @@ pub(crate) use core::io_primitives::{ENV_RUSTFS_PUT_RENAME_EARLY_ACK_ENABLE, ren
 mod ctx;
 mod metadata;
 mod ops;
 pub(crate) use ops::bucket::BucketInfoQuorum;
 #[cfg(test)]
 pub(crate) use ops::hermetic_set_disks_isolated;
@@ -21,72 +21,12 @@
 use super::super::{
    BUCKET_OP_IGNORED_ERRS, BucketInfo, BucketOperations, BucketOptions, DeleteBucketOptions, DiskError, Error, HashMap,
-    MakeBucketOptions, Result, SetDisks, is_reserved_or_invalid_bucket, join_all, reduce_read_quorum_errs,
+    MakeBucketOptions, Result, SetDisks, is_reserved_or_invalid_bucket, join_all, reduce_write_quorum_errs,
    reduce_write_quorum_errs,
 };
 use crate::api::bucket::metadata_sys;
 use crate::disk::DiskAPI;
 #[derive(Clone, Copy)]
 pub(crate) enum BucketInfoQuorum {
    Read,
    Write,
 }
 impl SetDisks {
    pub(crate) async fn stat_bucket_with_quorum(&self, bucket: &str, quorum: BucketInfoQuorum) -> Result<BucketInfo> {
        let disks = self.disk_inventory().await;
        let disk_count = disks.len();
        let mut futures = Vec::with_capacity(disk_count);
        for disk in disks {
            let bucket = bucket.to_string();
            futures.push(async move {
                match disk {
                    Some(disk) => disk.stat_volume(&bucket).await,
                    None => Err(DiskError::DiskNotFound),
                }
            });
        }
        let results = join_all(futures).await;
        let mut infos = Vec::with_capacity(results.len());
        let mut errs = Vec::with_capacity(results.len());
        for result in results {
            match result {
                Ok(info) => {
                    infos.push(Some(info));
                    errs.push(None);
                }
                Err(err) => {
                    infos.push(None);
                    errs.push(Some(err));
                }
            }
        }
        let error = match quorum {
            // Bucket mutations use a majority regardless of object storage
            // class. A namespace read must intersect that majority; object
            // readers still enforce the persisted layout's data-shard quorum.
            BucketInfoQuorum::Read => reduce_read_quorum_errs(&errs, BUCKET_OP_IGNORED_ERRS, disk_count.div_ceil(2).max(1)),
            BucketInfoQuorum::Write => reduce_write_quorum_errs(&errs, BUCKET_OP_IGNORED_ERRS, disk_count / 2 + 1),
        };
        if let Some(err) = error {
            return Err(err.into());
        }
        infos
            .into_iter()
            .flatten()
            .next()
            .map(|info| BucketInfo {
                name: info.name,
                created: info.created,
                ..Default::default()
            })
            .ok_or(Error::VolumeNotFound)
    }
    pub(crate) async fn list_bucket_for_scanner(&self, _opts: &BucketOptions) -> Result<(Vec<BucketInfo>, bool)> {
        let disks = self.disk_inventory().await;
        let write_quorum = (disks.len() / 2) + 1;
@@ -191,12 +131,59 @@ impl BucketOperations for SetDisks {
    #[tracing::instrument(skip(self))]
    async fn get_bucket_info(&self, bucket: &str, _opts: &BucketOptions) -> Result<BucketInfo> {
-        let mut info = self.stat_bucket_with_quorum(bucket, BucketInfoQuorum::Write).await?;
+        let disks = self.disk_inventory().await;
-        if let Ok(sys) = metadata_sys::get(bucket).await {
+        let write_quorum = (disks.len() / 2) + 1;
-            info.versioning = sys.versioning();
+
-            info.object_locking = sys.object_locking();
+        let mut futures = Vec::with_capacity(disks.len());
        for disk in disks {
            let bucket = bucket.to_string();
            futures.push(async move {
                match disk {
                    Some(disk) => disk.stat_volume(&bucket).await,
                    None => Err(DiskError::DiskNotFound),
                }
            });
        }
-        Ok(info)
+
        let results = join_all(futures).await;
        let mut infos = Vec::with_capacity(results.len());
        let mut errs = Vec::with_capacity(results.len());
        for result in results {
            match result {
                Ok(info) => {
                    infos.push(Some(info));
                    errs.push(None);
                }
                Err(err) => {
                    infos.push(None);
                    errs.push(Some(err));
                }
            }
        }
        if let Some(err) = reduce_write_quorum_errs(&errs, BUCKET_OP_IGNORED_ERRS, write_quorum) {
            return Err(err.into());
        }
        let mut versioning = false;
        let mut object_locking = false;
        if let Ok(sys) = metadata_sys::get(bucket).await {
            versioning = sys.versioning();
            object_locking = sys.object_locking();
        }
        infos
            .into_iter()
            .flatten()
            .next()
            .map(|info| BucketInfo {
                name: info.name,
                created: info.created,
                versioning,
                object_locking,
                ..Default::default()
            })
            .ok_or(Error::VolumeNotFound)
    }
    #[tracing::instrument(skip(self))]
@@ -19,7 +19,7 @@ use crate::bucket::{
 };
 use crate::error::is_err_bucket_not_found;
 use crate::runtime::sources as runtime_sources;
-use crate::set_disk::{BucketInfoQuorum, get_lock_acquire_timeout};
+use crate::set_disk::get_lock_acquire_timeout;
 use crate::storage_api_contracts::bucket::{BUCKET_LIFECYCLE_LOCK_OBJECT, SRBucketDeleteOp};
 use crate::storage_api_contracts::namespace::NamespaceLocking as _;
 use futures::stream::{self, StreamExt};
@@ -772,30 +772,17 @@ impl ECStore {
    #[instrument(skip(self))]
    pub(crate) async fn get_bucket_info_from_sets(&self, bucket: &str, opts: &BucketOptions) -> Result<BucketInfo> {
        self.get_bucket_info_from_sets_with_quorum(bucket, opts, BucketInfoQuorum::Write)
            .await
    }
    async fn get_bucket_info_from_sets_with_quorum(
        &self,
        bucket: &str,
        opts: &BucketOptions,
        quorum: BucketInfoQuorum,
    ) -> Result<BucketInfo> {
        // One host may participate in several pools after expansion. Resolve the
        // namespace against each erasure set so disks from different pools can
        // never be combined into one bucket quorum.
        // Bucket validation is request-path IO. Keep the previous peer fanout's
        // latency shape by probing every set concurrently; scanner listings use
        // a separate bounded path below because they run continuously.
-        let mut scoped_results = futures::future::join_all(self.bucket_sets().map(|(pool_index, set_index, set)| async move {
+        let mut scoped_results =
-            let result = match quorum {
+            futures::future::join_all(self.bucket_sets().map(|(pool_index, set_index, set)| async move {
-                BucketInfoQuorum::Read => set.stat_bucket_with_quorum(bucket, quorum).await,
+                (pool_index, set_index, set.get_bucket_info(bucket, opts).await)
-                BucketInfoQuorum::Write => set.get_bucket_info(bucket, opts).await,
+            }))
-            };
+            .await;
            (pool_index, set_index, result)
        }))
        .await;
        scoped_results.sort_unstable_by_key(|(pool_index, set_index, _)| (*pool_index, *set_index));
        let mut first_info = None;
@@ -819,11 +806,7 @@ impl ECStore {
    #[instrument(skip(self))]
    pub(super) async fn handle_get_bucket_info(&self, bucket: &str, opts: &BucketOptions) -> Result<BucketInfo> {
-        let mut info = match self.get_bucket_info_from_sets(bucket, opts).await {
+        let mut info = self.get_bucket_info_from_sets(bucket, opts).await?;
            Ok(info) => info,
            Err(Error::ErasureWriteQuorum) => return self.get_bucket_info_at_read_quorum(bucket, opts).await,
            Err(err) => return Err(err),
        };
        if let Ok(sys) = metadata_sys::get_in(&self.ctx, bucket).await {
            if should_override_created_from_metadata(sys.created) {
@@ -836,35 +819,6 @@ impl ECStore {
        Ok(info)
    }
    async fn get_bucket_info_at_read_quorum(&self, bucket: &str, opts: &BucketOptions) -> Result<BucketInfo> {
        // Lock order: bucket lifecycle -> internal metadata object read locks.
        // Keep create/delete from changing the namespace while a read quorum
        // confirms both physical presence and persisted bucket metadata.
        let guard = self.acquire_bucket_lifecycle_read_lock(bucket).await?;
        await_bucket_namespace_operation(Some(&guard), bucket, "bucket read quorum validation", async {
            let mut info = self
                .get_bucket_info_from_sets_with_quorum(bucket, opts, BucketInfoQuorum::Read)
                .await?;
            let (metadata, persisted) = metadata_sys::get_config_from_disk_with_presence_in(&self.ctx, bucket).await?;
            if !persisted {
                // A minority of directories left by failed creation is not an
                // authoritative bucket. Never turn fabricated defaults into
                // permission to serve degraded reads.
                return Err(Error::ErasureReadQuorum);
            }
            if metadata.name != bucket {
                return Err(Error::FileCorrupt);
            }
            if should_override_created_from_metadata(metadata.created) {
                info.created = Some(metadata.created);
            }
            info.versioning = metadata.versioning();
            info.object_locking = metadata.object_locking();
            Ok(info)
        })
        .await
    }
    #[instrument(skip(self))]
    pub(super) async fn handle_list_bucket(&self, opts: &BucketOptions) -> Result<Vec<BucketInfo>> {
        // TODO(backlog): support cached bucket listing via opts.cached
@@ -1095,7 +1049,7 @@ mod tests {
        run_physical_bucket_deletion, scan_metadata_less_residue, scan_metadata_less_residue_with_budget,
        should_override_created_from_metadata, validate_table_bucket_delete_allowed,
    };
-    use crate::bucket::metadata::{BucketMetadata, table_bucket_catalog_metadata_prefix};
+    use crate::bucket::metadata::table_bucket_catalog_metadata_prefix;
    use crate::bucket::metadata_sys;
    use crate::cluster::rpc::peer_s3_client::install_delete_bucket_empty_scan_barrier;
    use crate::disk::{BUCKET_META_PREFIX, DiskAPI, RUSTFS_META_BUCKET, STORAGE_FORMAT_FILE};
@@ -1122,7 +1076,6 @@ mod tests {
    use std::sync::atomic::{AtomicBool, Ordering};
    use std::time::{Duration, SystemTime};
    use time::OffsetDateTime;
    use tokio::io::AsyncReadExt;
    use tokio::sync::{Notify, OnceCell};
    use tokio_util::sync::CancellationToken;
    use uuid::Uuid;
@@ -1406,18 +1359,11 @@ mod tests {
    }
    async fn setup_multi_pool_bucket_test_env() -> (tempfile::TempDir, Arc<ECStore>) {
        setup_bucket_quorum_test_env(&[4, 4], None).await
    }
    async fn setup_bucket_quorum_test_env(
        drives_per_pool: &[usize],
        standard_parity: Option<usize>,
    ) -> (tempfile::TempDir, Arc<ECStore>) {
        let temp_dir = tempfile::tempdir().expect("multi-pool bucket test directory should be created");
        let mut pools = Vec::new();
-        for (pool_index, &drive_count) in drives_per_pool.iter().enumerate() {
+        for pool_index in 0..2 {
            let mut endpoints = Vec::new();
-            for disk_index in 0..drive_count {
+            for disk_index in 0..4 {
                let disk_path = temp_dir.path().join(format!("pool{pool_index}-disk{disk_index}"));
                tokio::fs::create_dir_all(&disk_path)
                    .await
@@ -1432,7 +1378,7 @@ mod tests {
            pools.push(PoolEndpoints {
                legacy: false,
                set_count: 1,
-                drives_per_set: drive_count,
+                drives_per_set: 4,
                endpoints: Endpoints::from(endpoints),
                cmd_line: format!("bucket-test-pool-{pool_index}"),
                platform: format!("OS: {} | Arch: {}", std::env::consts::OS, std::env::consts::ARCH),
@@ -1453,12 +1399,9 @@ mod tests {
        )
        .await
        .expect("multi-pool ECStore should initialize");
-        let mut storage_class_kvs = rustfs_config::server_config::KVS::new();
+        let storage_class =
-        if let Some(parity) = standard_parity {
+            crate::config::storageclass::lookup_config_for_pools_without_env(&rustfs_config::server_config::KVS::new(), &[4, 4])
-            storage_class_kvs.insert(crate::config::storageclass::CLASS_STANDARD.to_string(), format!("EC:{parity}"));
+                .expect("multi-pool storage class should match both four-disk pools");
        }
        let storage_class = crate::config::storageclass::lookup_config_for_pools_without_env(&storage_class_kvs, drives_per_pool)
            .expect("storage class should match every test erasure set");
        for pool in &ecstore.pools {
            for set in &pool.disk_set {
                set.set_test_storage_class_config(storage_class.clone());
@@ -2124,218 +2067,6 @@ mod tests {
        }
    }
    #[tokio::test]
    #[serial]
    async fn bucket_info_read_quorum_tracks_erasure_layout() {
        for (drive_count, parity) in [(2, 1), (3, 1), (4, 2), (5, 2), (6, 3), (8, 4), (6, 2), (12, 6)] {
            let (_temp_dir, store) = setup_bucket_quorum_test_env(&[drive_count], Some(parity)).await;
            metadata_sys::init_bucket_metadata_sys(store.clone(), Vec::new()).await;
            let bucket = format!("read-quorum-{drive_count}-{parity}");
            let object = "uncached-object";
            let body = b"erasure read quorum must follow the persisted layout".repeat(32_768);
            store
                .make_bucket(&bucket, &MakeBucketOptions::default())
                .await
                .expect("healthy namespace should accept bucket creation");
            store
                .put_object(&bucket, object, &mut PutObjReader::from_vec(body.clone()), &ObjectOptions::default())
                .await
                .expect("healthy erasure set should accept the seed object");
            let set = &store.pools[0].disk_set[0];
            let lock = set
                .new_ns_lock(&bucket, object)
                .await
                .expect("seed namespace lock should resolve");
            drop(
                lock.get_write_lock(Duration::from_secs(30))
                    .await
                    .expect("seed physical fanout must finish before taking disks offline"),
            );
            if (drive_count, parity) == (6, 3) {
                let mut kvs = rustfs_config::server_config::KVS::new();
                kvs.insert(crate::config::storageclass::CLASS_STANDARD.to_string(), "EC:2".to_string());
                set.set_test_storage_class_config(
                    crate::config::storageclass::lookup_config_for_pools_without_env(&kvs, &[drive_count])
                        .expect("a later storage-class change must not raise old objects' read quorum"),
                );
            }
            let offline_indexes = (0..parity).collect::<Vec<_>>();
            let offline = take_set_disks_offline(&store, set, &offline_indexes).await;
            let info = store
                .get_bucket_info(&bucket, &BucketOptions::default())
                .await
                .expect("bucket validation must admit the object's exact read quorum");
            assert_eq!(info.name, bucket);
            let mut reader = store
                .get_object_reader(&bucket, object, None, Default::default(), &ObjectOptions::default())
                .await
                .expect("the persisted layout should remain readable at its exact data-shard quorum");
            let mut restored = Vec::new();
            reader
                .stream
                .read_to_end(&mut restored)
                .await
                .expect("quorum read should reconstruct the body");
            assert_eq!(restored, body, "layout {drive_count}/{parity} must retain exact object contents");
            drop(reader);
            if drive_count - parity == drive_count / 2 {
                let error = store
                    .get_bucket_info_from_sets(&bucket, &BucketOptions::default())
                    .await
                    .expect_err("bucket mutations must retain their majority namespace check");
                assert_eq!(error, StorageError::ErasureWriteQuorum);
            }
            let below_quorum = take_set_disks_offline(&store, set, &[parity]).await;
            let read = store
                .get_object_reader(&bucket, object, None, Default::default(), &ObjectOptions::default())
                .await;
            match read {
                Ok(mut reader) => assert!(
                    reader.stream.read_to_end(&mut Vec::new()).await.is_err(),
                    "layout {drive_count}/{parity} must reject fewer than its data-shard quorum"
                ),
                Err(error) => assert!(
                    matches!(error, StorageError::ErasureReadQuorum | StorageError::InsufficientReadQuorum(_, _)),
                    "a missing shard must report read quorum loss, got {error}"
                ),
            }
            restore_set_disks(&store, set, below_quorum).await;
            restore_set_disks(&store, set, offline).await;
        }
    }
    #[tokio::test]
    #[serial]
    async fn bucket_info_read_quorum_is_scoped_to_each_erasure_set() {
        let (_temp_dir, store) = setup_bucket_quorum_test_env(&[4, 6], None).await;
        metadata_sys::init_bucket_metadata_sys(store.clone(), Vec::new()).await;
        let bucket = "read-quorum-mixed-pools";
        store
            .make_bucket(bucket, &MakeBucketOptions::default())
            .await
            .expect("healthy pools should accept bucket creation");
        let first_set = &store.pools[0].disk_set[0];
        let second_set = &store.pools[1].disk_set[0];
        let first_offline = take_set_disks_offline(&store, first_set, &[0, 1]).await;
        let second_offline = take_set_disks_offline(&store, second_set, &[0, 1, 2]).await;
        store
            .get_bucket_info(bucket, &BucketOptions::default())
            .await
            .expect("each set independently satisfies its namespace read quorum");
        for (set, extra_disk) in [(first_set, 2), (second_set, 3)] {
            let extra_offline = take_set_disks_offline(&store, set, &[extra_disk]).await;
            assert_eq!(
                store
                    .get_bucket_info(bucket, &BucketOptions::default())
                    .await
                    .expect_err("another pool must not subsidize a set below its read quorum"),
                StorageError::ErasureReadQuorum
            );
            restore_set_disks(&store, set, extra_offline).await;
        }
        restore_set_disks(&store, first_set, first_offline).await;
        restore_set_disks(&store, second_set, second_offline).await;
    }
    #[tokio::test]
    #[serial]
    async fn bucket_info_read_quorum_requires_authoritative_metadata() {
        for state in ["missing", "corrupt", "foreign", "incarnation"] {
            let (_temp_dir, store) = setup_bucket_quorum_test_env(&[4], None).await;
            metadata_sys::init_bucket_metadata_sys(store.clone(), Vec::new()).await;
            let bucket = format!("read-quorum-{state}-metadata");
            let mut metadata = if state == "missing" {
                store
                    .make_bucket_on_sets(&bucket, &MakeBucketOptions::default())
                    .await
                    .expect("simulate directories left before bucket metadata is published");
                BucketMetadata::new(&bucket)
            } else {
                store
                    .make_bucket(&bucket, &MakeBucketOptions::default())
                    .await
                    .expect("healthy bucket should publish metadata");
                metadata_sys::get_in(&store.ctx, &bucket)
                    .await
                    .expect("seed metadata should be cached")
                    .as_ref()
                    .clone()
            };
            let path = metadata.save_file_path();
            match state {
                "corrupt" => crate::config::com::save_config(store.clone(), &path, b"corrupt".to_vec())
                    .await
                    .expect("persist corrupt metadata while the cached copy remains valid"),
                "foreign" => {
                    metadata.name = "different-bucket".to_string();
                    let mut encoded = vec![1, 0, 1, 0];
                    encoded.extend(metadata.marshal_msg().expect("foreign metadata should encode"));
                    crate::config::com::save_config(store.clone(), &path, encoded)
                        .await
                        .expect("persist metadata for a different bucket at the requested path");
                }
                "incarnation" => crate::bucket::metadata::save_bucket_incarnation(store.clone(), &bucket, Uuid::new_v4())
                    .await
                    .expect("persist a different bucket generation"),
                _ => {}
            }
            let set = &store.pools[0].disk_set[0];
            let offline = take_set_disks_offline(&store, set, &[0, 1]).await;
            let error = store
                .get_bucket_info(&bucket, &BucketOptions::default())
                .await
                .expect_err("read admission must not trust residual directories or cached metadata");
            match state {
                "missing" => assert_eq!(error, StorageError::ErasureReadQuorum),
                "foreign" => assert_eq!(error, StorageError::FileCorrupt),
                "incarnation" => assert!(error.to_string().contains("sidecar does not match bucket metadata")),
                "corrupt" => assert!(error.to_string().contains("format invalid"), "unexpected corruption error: {error}"),
                _ => unreachable!(),
            }
            restore_set_disks(&store, set, offline).await;
        }
    }
    #[tokio::test]
    #[serial]
    async fn bucket_info_read_quorum_accepts_persisted_legacy_metadata() {
        let (_temp_dir, store) = setup_bucket_quorum_test_env(&[4], None).await;
        metadata_sys::init_bucket_metadata_sys(store.clone(), Vec::new()).await;
        let bucket = "interop";
        store
            .make_bucket_on_sets(bucket, &MakeBucketOptions::default())
            .await
            .expect("legacy bucket directories should exist");
        let hex = include_str!("../../tests/fixtures/minio/bucket_metadata.blob.hex")
            .split_whitespace()
            .collect::<String>();
        let body = (0..hex.len())
            .step_by(2)
            .map(|index| u8::from_str_radix(&hex[index..index + 2], 16).expect("pinned MinIO metadata fixture"))
            .collect();
        crate::config::com::save_config(store.clone(), &BucketMetadata::new(bucket).save_file_path(), body)
            .await
            .expect("legacy metadata should be persisted without an incarnation sidecar");
        let set = &store.pools[0].disk_set[0];
        let offline = take_set_disks_offline(&store, set, &[0, 1]).await;
        let info = store
            .get_bucket_info(bucket, &BucketOptions::default())
            .await
            .expect("persisted MinIO metadata should authorize reads at the namespace read quorum");
        assert_eq!(info.name, bucket);
        assert!(info.versioning);
        assert!(info.object_locking);
        restore_set_disks(&store, set, offline).await;
    }
    #[tokio::test]
    #[serial]
    async fn bucket_namespace_reads_report_missing_when_every_set_is_absent() {
@@ -2359,7 +2090,6 @@ mod tests {
    #[serial]
    async fn bucket_namespace_reads_fail_closed_when_any_set_loses_quorum() {
        let (_temp_dir, ecstore) = setup_multi_pool_bucket_test_env().await;
        metadata_sys::init_bucket_metadata_sys(ecstore.clone(), Vec::new()).await;
        let bucket = format!("degraded-expansion-{}", Uuid::new_v4().simple());
        ecstore.pools[0].disk_set[0]
            .make_bucket(&bucket, &MakeBucketOptions::default())
@@ -2367,7 +2097,6 @@ mod tests {
            .expect("bucket should be created in the original pool only");
        ecstore.pools[1].disk_set[0].disks.write().await[0] = None;
        ecstore.pools[1].disk_set[0].disks.write().await[1] = None;
        ecstore.pools[1].disk_set[0].disks.write().await[2] = None;
        let list_err = ecstore
            .list_bucket(&BucketOptions::default())
@@ -2379,7 +2108,7 @@ mod tests {
            .get_bucket_info(&bucket, &BucketOptions::default())
            .await
            .expect_err("bucket validation must fail when an expansion pool is unavailable");
-        assert_eq!(info_err, StorageError::ErasureReadQuorum);
+        assert_eq!(info_err, StorageError::ErasureWriteQuorum);
    }
    #[tokio::test]
@@ -18,6 +18,7 @@ use crate::services::rebalance::{REBAL_META_NAME, RebalStatus};
 use crate::set_disk::get_lock_acquire_timeout;
 use crate::storage_api_contracts::heal::HealOperations as _;
 use crate::storage_api_contracts::namespace::NamespaceLocking as _;
 use rustfs_heal_contracts::heal_channel::DriveState;
 use rustfs_lock::NamespaceLockGuard;
 use std::collections::BTreeSet;
 use tracing::trace;
@@ -378,6 +379,79 @@ impl ECStore {
        Ok(result)
    }
    /// Heal every pool metadata owner in the selected scope without allowing
    /// one healthy pool to hide another pool's failed repair.
    pub async fn heal_pool_metadata(&self, opts: &HealOpts) -> Result<Vec<HealResultItem>> {
        let scopes = self.heal_erasure_set_scopes(opts).await?;
        let mut results = Vec::new();
        for (pool_index, set_index) in scopes {
            if !self.replacement_pool_metadata_applies(pool_index, set_index)? {
                continue;
            }
            let set = &self.pools[pool_index].disk_set[set_index];
            let targets = set.set_endpoints.iter().map(ToString::to_string).collect::<Vec<_>>();
            if targets.is_empty()
                || targets.len() != set.set_drive_count
                || targets.iter().collect::<BTreeSet<_>>().len() != targets.len()
            {
                return Err(Error::SlowDown);
            }
            // Administrative remove/no-lock options apply to user objects,
            // never to the cluster's authoritative metadata transaction.
            let metadata_opts = HealOpts {
                dry_run: opts.dry_run,
                recreate: opts.recreate,
                scan_mode: opts.scan_mode,
                pool: Some(pool_index),
                set: Some(set_index),
                ..Default::default()
            };
            let (result, error) = self
                .handle_heal_object(RUSTFS_META_BUCKET, POOL_META_NAME, "", &metadata_opts)
                .await?;
            if let Some(error) = error {
                return Err(error);
            }
            if !opts.dry_run {
                let ok_state = DriveState::Ok.to_string();
                let complete = result.after.drives.len() == targets.len()
                    && targets.iter().all(|target| {
                        let mut outcomes = result.after.drives.iter().filter(|drive| drive.endpoint == *target);
                        outcomes.next().is_some_and(|drive| drive.state == ok_state) && outcomes.next().is_none()
                    });
                if !complete
                    || !set
                        .replacement_targets_have_version(RUSTFS_META_BUCKET, POOL_META_NAME, "", &targets)
                        .await?
                {
                    return Err(Error::SlowDown);
                }
            }
            results.push(result);
        }
        Ok(results)
    }
    /// Whether this replacement set owns the pool's metadata replica.
    ///
    /// Pool metadata follows normal object placement within each pool. A valid
    /// non-owner set has no replica to repair; missing metadata on the owner
    /// set still requires healing and target-specific readback.
    pub fn replacement_pool_metadata_applies(&self, pool_index: usize, set_index: usize) -> Result<bool> {
        let pool = self
            .pools
            .get(pool_index)
            .ok_or_else(|| invalid_heal_pool_index(pool_index, self.pools.len()))?;
        let selected = pool.get_disks_for_heal_object(
            POOL_META_NAME,
            &HealOpts {
                set: Some(set_index),
                ..Default::default()
            },
        )?;
        Ok(Arc::ptr_eq(&selected, &pool.get_disks_by_key(POOL_META_NAME)))
    }
    #[instrument(skip(self, targets), fields(pool_index, set_index, target_count = targets.len()))]
    pub async fn replacement_targets_have_version(
        &self,
@@ -829,6 +903,249 @@ mod tests {
        }
    }
    #[tokio::test]
    async fn replacement_pool_metadata_applies_to_the_written_replica_in_each_pool() {
        let mut store = minimal_heal_store().await;
        for pool_index in 0..store.pools.len() {
            assert!(
                store
                    .replacement_pool_metadata_applies(pool_index, 0)
                    .expect("a valid single-set pool should have a metadata owner")
            );
        }
        store.ctx = Arc::new(InstanceContext::new());
        for algorithm in [
            crate::disk::format::DistributionAlgoVersion::V1,
            crate::disk::format::DistributionAlgoVersion::V2,
            crate::disk::format::DistributionAlgoVersion::V3,
        ] {
            let mut temp_dirs = Vec::new();
            for pool_index in 0..store.pools.len() {
                let (dirs, mut pool) =
                    crate::core::sets::make_local_two_set_sets_for_pool_with_ctx(Arc::clone(&store.ctx), pool_index).await;
                temp_dirs.extend(dirs);
                Arc::get_mut(&mut pool)
                    .expect("fixture pool should have one owner")
                    .distribution_algo = algorithm.clone();
                store.pools[pool_index] = pool;
            }
            for (pool_index, pool) in store.pools.iter().enumerate() {
                let mut required_sets = 0;
                for set_index in 0..pool.disk_set.len() {
                    required_sets += usize::from(
                        store
                            .replacement_pool_metadata_applies(pool_index, set_index)
                            .expect("valid replacement topology should be classified before metadata exists"),
                    );
                }
                assert_eq!(required_sets, 1, "missing metadata cannot exempt the owner set");
                save_config(pool.clone(), POOL_META_NAME, b"pool metadata placement".to_vec())
                    .await
                    .expect("normal config writes should persist one metadata replica per pool");
                for (set_index, set) in pool.disk_set.iter().enumerate() {
                    let applies = store
                        .replacement_pool_metadata_applies(pool_index, set_index)
                        .expect("valid replacement topology should be classified");
                    let disks = set.disks.read().await.clone();
                    for disk in disks.iter().flatten() {
                        let replica = disk.read_xl(RUSTFS_META_BUCKET, POOL_META_NAME, false).await;
                        if applies {
                            replica.expect("the metadata owner must match actual persisted shards");
                        } else {
                            assert!(
                                matches!(replica, Err(crate::disk::error::DiskError::FileNotFound)),
                                "non-owner sets must have no persisted metadata shard; observed error: {:?}",
                                replica.as_ref().err()
                            );
                        }
                    }
                }
                assert!(
                    store
                        .replacement_pool_metadata_applies(pool_index, pool.disk_set.len())
                        .is_err()
                );
            }
        }
        assert!(store.replacement_pool_metadata_applies(store.pools.len(), 0).is_err());
    }
    #[tokio::test]
    #[serial_test::serial]
    async fn ordinary_pool_metadata_heal_repairs_each_owner_and_preserves_dry_run() {
        let (_temp_dirs, store, _other_store) = test_two_pool_stores(None).await;
        let first_missing = remove_pool_meta_shard(&store, 0).await;
        let second_missing = remove_pool_meta_shard(&store, 1).await;
        let destructive_options = HealOpts {
            remove: true,
            no_lock: true,
            ..Default::default()
        };
        let results = store
            .heal_pool_metadata(&HealOpts {
                dry_run: true,
                ..destructive_options
            })
            .await
            .expect("dry-run should inspect both metadata owners without requiring a commit");
        assert_eq!(results.len(), 2);
        assert!(
            first_missing
                .read_xl(RUSTFS_META_BUCKET, POOL_META_NAME, false)
                .await
                .is_err()
        );
        assert!(
            second_missing
                .read_xl(RUSTFS_META_BUCKET, POOL_META_NAME, false)
                .await
                .is_err()
        );
        let lock = store.pools[0]
            .new_ns_lock(RUSTFS_META_BUCKET, POOL_META_NAME)
            .await
            .expect("metadata namespace lock should be available");
        let guard = lock
            .get_read_lock(get_lock_acquire_timeout())
            .await
            .expect("a metadata reader should hold the shared fence");
        let error = temp_env::async_with_vars(
            [(rustfs_config::ENV_OBJECT_LOCK_ACQUIRE_TIMEOUT, Some("1"))],
            store.heal_pool_metadata(&destructive_options),
        )
        .await
        .expect_err("administrative no-lock cannot bypass the metadata write fence");
        assert!(matches!(error, Error::Lock(rustfs_lock::LockError::Timeout { .. })));
        drop(guard);
        let results = store
            .heal_pool_metadata(&destructive_options)
            .await
            .expect("every metadata owner should be repaired");
        assert_eq!(results.len(), 2);
        assert!(first_missing.read_xl(RUSTFS_META_BUCKET, POOL_META_NAME, false).await.is_ok());
        assert!(
            second_missing
                .read_xl(RUSTFS_META_BUCKET, POOL_META_NAME, false)
                .await
                .is_ok()
        );
    }
    #[tokio::test]
    #[serial_test::serial]
    async fn ordinary_pool_metadata_heal_does_not_hide_missing_later_pool() {
        let (_temp_dirs, store, _other_store) = test_two_pool_stores(None).await;
        delete_config(store.pools[1].clone(), POOL_META_NAME)
            .await
            .expect("the second pool metadata replica should be removed");
        let error = store
            .heal_pool_metadata(&HealOpts::default())
            .await
            .expect_err("the healthy first pool must not hide the second owner's missing replica");
        assert!(!matches!(error, Error::NoHealRequired));
        let second_set = store.pools[1].get_disks_by_key(POOL_META_NAME);
        for disk in second_set.disks.read().await.iter().flatten() {
            assert!(disk.read_xl(RUSTFS_META_BUCKET, POOL_META_NAME, false).await.is_err());
        }
    }
    #[tokio::test]
    async fn ordinary_pool_metadata_heal_skips_only_valid_non_owner_sets() {
        let mut store = minimal_heal_store().await;
        store.ctx = Arc::new(InstanceContext::new());
        for algorithm in [
            crate::disk::format::DistributionAlgoVersion::V1,
            crate::disk::format::DistributionAlgoVersion::V2,
            crate::disk::format::DistributionAlgoVersion::V3,
        ] {
            let mut temp_dirs = Vec::new();
            for pool_index in 0..store.pools.len() {
                let (dirs, mut pool) =
                    crate::core::sets::make_local_two_set_sets_for_pool_with_ctx(Arc::clone(&store.ctx), pool_index).await;
                temp_dirs.extend(dirs);
                Arc::get_mut(&mut pool)
                    .expect("fixture pool should have one owner")
                    .distribution_algo = algorithm.clone();
                store.pools[pool_index] = pool;
            }
            for pool_index in 0..store.pools.len() {
                let owner = (0..store.pools[pool_index].disk_set.len())
                    .find(|set_index| {
                        store
                            .replacement_pool_metadata_applies(pool_index, *set_index)
                            .expect("valid metadata placement")
                    })
                    .expect("every pool must have one metadata owner");
                let non_owner = 1 - owner;
                assert!(
                    store
                        .heal_pool_metadata(&HealOpts {
                            pool: Some(pool_index),
                            set: Some(non_owner),
                            ..Default::default()
                        })
                        .await
                        .expect("valid non-owner should need no metadata write")
                        .is_empty()
                );
                assert!(
                    store
                        .heal_pool_metadata(&HealOpts {
                            pool: Some(pool_index),
                            set: Some(owner),
                            ..Default::default()
                        })
                        .await
                        .is_err(),
                    "an owner with no authoritative metadata must fail"
                );
                assert!(
                    store
                        .heal_pool_metadata(&HealOpts {
                            pool: Some(pool_index),
                            set: Some(2),
                            ..Default::default()
                        })
                        .await
                        .is_err(),
                    "invalid sets cannot claim the non-owner exemption"
                );
            }
        }
        assert!(
            store
                .heal_pool_metadata(&HealOpts {
                    pool: Some(2),
                    ..Default::default()
                })
                .await
                .is_err()
        );
    }
    #[tokio::test]
    #[serial_test::serial]
    async fn ordinary_pool_metadata_heal_requires_every_owner_endpoint() {
        let (_temp_dirs, store, _other_store) = test_two_pool_stores(None).await;
        let owner = store.pools[0].get_disks_by_key(POOL_META_NAME);
        let offline_disk = owner.disks.write().await[0]
            .take()
            .expect("fixture owner disk should start online");
        let result = store
            .heal_pool_metadata(&HealOpts {
                pool: Some(0),
                ..Default::default()
            })
            .await;
        assert!(result.is_err(), "a surviving metadata shard must not hide an offline owner endpoint");
        owner.disks.write().await[0] = Some(offline_disk);
    }
    async fn remove_pool_meta_shard(store: &ECStore, pool_idx: usize) -> DiskStore {
        let target_set = store.pools[pool_idx].get_disks_by_key(POOL_META_NAME);
        let missing_disk = target_set.disks.read().await[0]
@@ -18932,6 +18932,124 @@ mod tests {
        .await;
    }
    #[cfg(feature = "test-util")]
    #[test]
    #[serial_test::serial(storage_class_env)]
    fn tier_config_init_recovers_after_initial_reload_failure_without_another_mutation() {
        run_large_stack_async_test("tier-config-init-recovery", || async {
            use crate::services::tier::tier::{TIER_DRIVER_TEST_FACTORY, TierDriverTestFactory};
            let temp_dir = tempfile::tempdir().expect("create tier startup recovery store dir");
            let (ctx, store, _shutdown) = build_isolated_test_store(temp_dir.path(), "tier-startup-recovery", &[4]).await;
            let manager = ctx.tier_config_mgr();
            let candidate = TierConfigMgr::new();
            let tier_name = "STARTUP-RECOVERY";
            let backend = register_mock_tier(&candidate, tier_name).await;
            let backend_identity = TierConfigMgr::acquire_operation_lease(&candidate, tier_name)
                .await
                .expect("candidate backend identity should resolve")
                .backend_identity();
            let candidate_digest = {
                let candidate = candidate.read().await;
                candidate
                    .save_tiering_config(store.clone())
                    .await
                    .expect("committed tier config should persist");
                tier_config_candidate_digest(&candidate).expect("committed candidate digest should build")
            };
            let config_path = format!("{}/{}", com::CONFIG_PREFIX, TIER_CONFIG_FILE);
            let config_bytes = com::read_config(store.clone(), &config_path)
                .await
                .expect("committed tier config bytes should load");
            let config_etag = store
                .get_object_info(RUSTFS_META_BUCKET, &config_path, &ObjectOptions::default())
                .await
                .expect("committed tier config metadata should load")
                .etag
                .expect("committed tier config should have an ETag");
            let mutation_id = uuid::Uuid::new_v4();
            let intent = TierMutationIntent {
                mutation_id,
                revision: 2,
                kind: TierMutationIntentKind::Add,
                state: TierMutationIntentState::Committed,
                old_config_etag: None,
                committed_config_etag: Some(config_etag),
                candidate_digest,
                affected_targets: vec![TierMutationIntentTarget {
                    tier_name: tier_name.to_string(),
                    old_backend_identity: None,
                    new_backend_identity: Some(backend_identity),
                }],
                expires_at_unix_nanos: 1,
            };
            // Persist the restart state directly: no peer Commit or runtime block
            // installation may leave a notification that masks a missing startup wakeup.
            save_tier_mutation_intent_record(store.clone(), &intent)
                .await
                .expect("committed restart intent should persist without notifying the manager");
            com::save_config(store.clone(), &config_path, vec![0])
                .await
                .expect("controlled invalid config should persist");
            runtime_sources::TEST_TIER_CONFIG_MGR
                .scope(manager.clone(), async {
                    let err = runtime_sources::init_tier_config_mgr(store.clone())
                        .await
                        .expect_err("initial reload must report the controlled config parse failure");
                    assert!(err.to_string().contains("tierConfigInit: no data"), "unexpected startup failure: {err}");
                })
                .await;
            assert!(manager.read().await.tiers.is_empty(), "failed startup must not publish the candidate");
            let blocked = match TierConfigMgr::acquire_operation_lease(&manager, tier_name).await {
                Ok(_) => panic!("failed startup must retain the recovered committed fence"),
                Err(err) => err,
            };
            assert!(TierConfigMgr::operation_lease_blocked_by_mutation(&blocked));
            // Restore only the durable bytes. Recovery must come from the worker
            // started by init, after the test-only handle scope has already ended.
            com::save_config(store.clone(), &config_path, config_bytes)
                .await
                .expect("restoring committed config bytes should remove the startup failure");
            // Reload only publishes this Add into an empty manager, so it has
            // no replaced backend to construct or probe. Only the lease check
            // below needs a driver; keep its factory on the observing task.
            let driver_factory: TierDriverTestFactory = Arc::new(move |_| Ok(Box::new(backend.clone())));
            TIER_DRIVER_TEST_FACTORY
                .scope(driver_factory, async {
                    tokio::time::timeout(Duration::from_secs(10), async {
                        loop {
                            match TierConfigMgr::acquire_operation_lease(&manager, tier_name).await {
                                Ok(lease) => {
                                    drop(lease);
                                    match load_tier_mutation_intent_record(store.clone(), mutation_id).await {
                                        Err(Error::ConfigNotFound) => break,
                                        Ok(retained) => assert_eq!(retained, intent),
                                        Err(err) => panic!("committed recovery intent should remain readable: {err}"),
                                    }
                                }
                                Err(err) => assert!(
                                    TierConfigMgr::operation_lease_blocked_by_mutation(&err),
                                    "recovery must retain the fence until the tier becomes available: {err}"
                                ),
                            }
                            tokio::time::sleep(Duration::from_millis(10)).await;
                        }
                    })
                    .await
                    .expect("startup worker must publish the tier and clean its fence without another mutation");
                })
                .await;
            let recovered = manager.read().await;
            assert_eq!(
                tier_config_candidate_digest(&recovered).expect("recovered config digest should build"),
                candidate_digest,
                "startup recovery must publish the committed configuration"
            );
        });
    }
    #[cfg(feature = "test-util")]
    #[tokio::test]
    #[serial_test::serial(storage_class_env)]
@@ -34,7 +34,7 @@ use std::time::{Duration, UNIX_EPOCH};
 use tokio::sync::{RwLock, Semaphore};
 use tracing::{debug, error, warn};
-use super::{DiskStore, EcstoreError};
+use super::{DiskStore, EcstoreError, POOL_META_NAME, RUSTFS_META_BUCKET};
 /// Outcome of classifying an error returned by [`HealStorageAPI::heal_object`].
 enum HealObjectOutcome {
@@ -68,6 +68,13 @@ struct PageConcurrencyGuard {
    set_label: String,
 }
 struct ErasureSetPassCounters<'a> {
    processed_objects: &'a mut u64,
    successful_objects: &'a mut u64,
    failed_objects: &'a mut u64,
    skipped_objects: &'a mut u64,
 }
 impl PageConcurrencyGuard {
    fn new(in_flight: Arc<AtomicUsize>, set_label: String) -> Self {
        let current = in_flight.fetch_add(1, Ordering::SeqCst) + 1;
@@ -106,6 +113,7 @@ pub struct ErasureSetHealer {
    heal_opts: HealOpts,
    source: HealRequestSource,
    target_endpoints: Arc<[String]>,
    pool_metadata_target_endpoints: Arc<[String]>,
    replacement_task_id: Option<String>,
    replacement_target_identities: Option<Arc<[ReplacementTargetIdentity]>>,
    mainline_pacer: Option<Arc<super::pacing::MainlinePacer>>,
@@ -355,6 +363,7 @@ impl ErasureSetHealer {
            heal_opts,
            source,
            target_endpoints: Vec::new().into(),
            pool_metadata_target_endpoints: Vec::new().into(),
            replacement_task_id: None,
            replacement_target_identities: None,
            mainline_pacer: None,
@@ -378,6 +387,13 @@ impl ErasureSetHealer {
        self
    }
    pub(crate) fn with_pool_metadata_targets(mut self, mut target_endpoints: Vec<String>) -> Self {
        target_endpoints.sort_unstable();
        target_endpoints.dedup();
        self.pool_metadata_target_endpoints = target_endpoints.into();
        self
    }
    pub(crate) fn with_replacement_identity_fence(
        mut self,
        replacement_target_identities: Option<Vec<ReplacementTargetIdentity>>,
@@ -834,6 +850,21 @@ impl ErasureSetHealer {
            current_object_index = 0;
        }
        if failed_objects == 0 && skipped_objects == 0 && failed_buckets == 0 {
            self.heal_pool_metadata(
                set_disk_id,
                &mut ErasureSetPassCounters {
                    processed_objects: &mut processed_objects,
                    successful_objects: &mut successful_objects,
                    failed_objects: &mut failed_objects,
                    skipped_objects: &mut skipped_objects,
                },
                resume_manager,
                checkpoint_manager,
            )
            .await?;
        }
        // 5. finalize. Only declare the set healed when nothing failed AND
        // nothing was transiently skipped — otherwise the resume/checkpoint
        // state must survive so the failed/skipped versions are retried instead
@@ -919,6 +950,251 @@ impl ErasureSetHealer {
        Ok(())
    }
    async fn heal_pool_metadata(
        &self,
        set_disk_id: &str,
        counters: &mut ErasureSetPassCounters<'_>,
        resume_manager: &ResumeManager,
        checkpoint_manager: &CheckpointManager,
    ) -> Result<()> {
        let mut metadata_opts = self.heal_opts;
        metadata_opts.remove = false;
        metadata_opts.no_lock = false;
        if self.replacement_task_id.is_none() {
            let (pool_index, set_index) = crate::heal::utils::parse_set_disk_id(set_disk_id)?;
            if metadata_opts.pool.is_some_and(|pool| pool != pool_index) || metadata_opts.set.is_some_and(|set| set != set_index)
            {
                return Err(Error::TaskExecutionFailed {
                    message: format!("Pool metadata scope does not match resumed set {set_disk_id}"),
                });
            }
            metadata_opts.pool = Some(pool_index);
            metadata_opts.set = Some(set_index);
        }
        let target_endpoints = if self.replacement_task_id.is_some() || self.pool_metadata_target_endpoints.is_empty() {
            self.target_endpoints.as_ref()
        } else {
            self.pool_metadata_target_endpoints.as_ref()
        };
        let target_scoped_recreate = !metadata_opts.dry_run && metadata_opts.recreate && !target_endpoints.is_empty();
        let ordinary_heal = self.replacement_task_id.is_none() && !target_scoped_recreate;
        if !ordinary_heal {
            if target_endpoints.is_empty() {
                return Err(Error::TaskExecutionFailed {
                    message: "Replacement pool metadata heal requires target endpoints".to_string(),
                });
            }
            if !self.storage.replacement_pool_metadata_applies(&metadata_opts).await? {
                return Ok(());
            }
        }
        let object_key = format!("{RUSTFS_META_BUCKET}/{POOL_META_NAME}");
        let checkpoint_key = compose_key(&object_key, None);
        let checkpoint = checkpoint_manager.get_checkpoint().await;
        if checkpoint.processed_objects.contains(&checkpoint_key)
            || checkpoint.failed_objects.contains(&checkpoint_key)
            || checkpoint.skipped_objects.contains(&checkpoint_key)
        {
            return Ok(());
        }
        drop(checkpoint);
        self.verify_replacement_identity_fence("pool metadata").await?;
        resume_manager
            .set_current_item(Some(RUSTFS_META_BUCKET.to_string()), Some(POOL_META_NAME.to_string()))
            .await?;
        let result = if ordinary_heal {
            match self.storage.heal_pool_metadata(&metadata_opts).await {
                Ok(results) if results.is_empty() => return Ok(()),
                Ok(results) => {
                    let [result] = results.as_slice() else {
                        return Err(Error::TaskExecutionFailed {
                            message: format!("Pool metadata returned multiple replicas for set {set_disk_id}"),
                        });
                    };
                    (result_object_size_u64(result), Ok(()))
                }
                Err(err @ Error::TaskCancelled) | Err(err @ Error::TaskTimeout) => return Err(err),
                Err(err) => match Self::classify_heal_object_error(&err) {
                    HealObjectOutcome::Absent | HealObjectOutcome::Transient => {
                        (0, Err(Error::transient_skip(format!("Pool metadata heal must be retried: {err}"))))
                    }
                    HealObjectOutcome::Failed => (0, Err(err)),
                },
            }
        } else {
            match self
                .storage
                .heal_object(RUSTFS_META_BUCKET, POOL_META_NAME, None, &metadata_opts)
                .await
            {
                Ok((result, None)) if target_outcomes_complete(&result, target_endpoints) => {
                    let object_size = result_object_size_u64(&result);
                    match self
                        .storage
                        .replacement_targets_have_version(
                            RUSTFS_META_BUCKET,
                            POOL_META_NAME,
                            None,
                            &metadata_opts,
                            target_endpoints,
                        )
                        .await
                    {
                        Ok(true) => (object_size, Ok(())),
                        Ok(false) => (
                            object_size,
                            Err(Error::transient_skip(
                                "Skipped replacement pool metadata heal because target readback did not confirm the committed version",
                            )),
                        ),
                        Err(err) => (
                            object_size,
                            Err(Error::transient_skip(format!(
                                "Skipped replacement pool metadata heal because target readback failed: {err}"
                            ))),
                        ),
                    }
                }
                Ok((result, None)) => (
                    result_object_size_u64(&result),
                    Err(Error::transient_skip(
                        "Skipped replacement pool metadata heal because a replacement target was not committed",
                    )),
                ),
                Ok((result, Some(err))) => {
                    let object_size = result_object_size_u64(&result);
                    match Self::classify_heal_object_error(&err) {
                        HealObjectOutcome::Absent | HealObjectOutcome::Transient => (
                            object_size,
                            Err(Error::transient_skip(format!(
                                "Skipped replacement pool metadata heal due to transient error: {err}"
                            ))),
                        ),
                        HealObjectOutcome::Failed => (object_size, Err(err)),
                    }
                }
                Err(err @ Error::TaskCancelled) | Err(err @ Error::TaskTimeout) => return Err(err),
                Err(err) => match Self::classify_heal_object_error(&err) {
                    HealObjectOutcome::Absent | HealObjectOutcome::Transient => (
                        0,
                        Err(Error::transient_skip(format!(
                            "Skipped replacement pool metadata heal due to transient error: {err}"
                        ))),
                    ),
                    HealObjectOutcome::Failed => (0, Err(err)),
                },
            }
        };
        let (object_size, result) = result;
        let mut bytes_processed = self.progress.read().await.bytes_processed;
        let mut telemetry_unknown = false;
        let checkpoint_outcome = match result {
            Ok(()) => {
                telemetry_unknown |= !increment_counter(counters.successful_objects);
                telemetry_unknown |= !add_bytes(&mut bytes_processed, object_size);
                debug!(
                    target: "rustfs::heal::erasure_healer",
                    event = EVENT_HEAL_ERASURE_OBJECT_STATE,
                    component = LOG_COMPONENT_HEAL,
                    subsystem = LOG_SUBSYSTEM_ERASURE_HEALER,
                    set_disk_id,
                    bucket = RUSTFS_META_BUCKET,
                    object = POOL_META_NAME,
                    state = "healed",
                    "Pool metadata healed"
                );
                CheckpointObjectOutcome::Processed
            }
            Err(Error::TransientSkip { message }) => {
                telemetry_unknown |= !increment_counter(counters.skipped_objects);
                telemetry_unknown |= !add_bytes(&mut bytes_processed, object_size);
                warn!(
                    target: "rustfs::heal::erasure_healer",
                    event = EVENT_HEAL_ERASURE_OBJECT_STATE,
                    component = LOG_COMPONENT_HEAL,
                    subsystem = LOG_SUBSYSTEM_ERASURE_HEALER,
                    set_disk_id,
                    bucket = RUSTFS_META_BUCKET,
                    object = POOL_META_NAME,
                    state = "transient_skip",
                    error = %message,
                    "Pool metadata heal skipped due to transient error"
                );
                CheckpointObjectOutcome::Skipped
            }
            Err(err) => {
                telemetry_unknown |= !increment_counter(counters.failed_objects);
                telemetry_unknown |= !add_bytes(&mut bytes_processed, object_size);
                warn!(
                    target: "rustfs::heal::erasure_healer",
                    event = EVENT_HEAL_ERASURE_OBJECT_STATE,
                    component = LOG_COMPONENT_HEAL,
                    subsystem = LOG_SUBSYSTEM_ERASURE_HEALER,
                    set_disk_id,
                    bucket = RUSTFS_META_BUCKET,
                    object = POOL_META_NAME,
                    state = "failed",
                    error = %err,
                    "Pool metadata heal failed"
                );
                CheckpointObjectOutcome::Failed
            }
        };
        telemetry_unknown |= !increment_counter(counters.processed_objects);
        let (outcome_record, counter_unknown, skipped_new_versions, skipped_ilm_expired) = {
            let mut progress = self.progress.write().await;
            progress.set_current_object(Some(object_key.clone()));
            progress.update_object_progress(
                *counters.processed_objects,
                *counters.successful_objects,
                *counters.failed_objects,
                *counters.skipped_objects,
                bytes_processed,
            );
            if telemetry_unknown {
                progress.mark_unknown();
            }
            (
                CheckpointObjectOutcomeRecord {
                    object: checkpoint_key,
                    outcome: checkpoint_outcome,
                    successful: progress.objects_healed,
                    failed: progress.objects_failed,
                    skipped: progress.skipped_objects,
                    bytes: progress.bytes_processed,
                    skipped_new_versions: progress.skipped_new_versions,
                    skipped_ilm_expired: progress.skipped_ilm_expired,
                    counter_unknown: progress.counter_unknown,
                },
                progress.counter_unknown,
                progress.skipped_new_versions,
                progress.skipped_ilm_expired,
            )
        };
        checkpoint_manager.record_object_outcome(outcome_record).await?;
        resume_manager
            .update_progress_with_bytes(
                *counters.processed_objects,
                *counters.successful_objects,
                *counters.failed_objects,
                *counters.skipped_objects,
                bytes_processed,
            )
            .await?;
        resume_manager
            .set_skipped_version_counts(skipped_new_versions, skipped_ilm_expired)
            .await?;
        if counter_unknown {
            resume_manager.mark_counter_unknown().await?;
        }
        Ok(())
    }
    /// heal single bucket with resume
    #[allow(clippy::too_many_arguments)]
    #[tracing::instrument(skip(self, current_object_index, processed_objects, successful_objects, failed_objects, skipped_objects, resume_manager, checkpoint_manager), fields(bucket = %bucket, bucket_index = bucket_index))]
@@ -1304,7 +1580,9 @@ impl ErasureSetHealer {
                        );
                        CheckpointObjectOutcome::Processed
                    }
-                    Err(err @ Error::TaskCancelled) | Err(err @ Error::TaskTimeout) => return Err(err),
+                    Err(err @ Error::TaskCancelled) | Err(err @ Error::TaskTimeout) => {
                        return Err(err);
                    }
                    Err(Error::TransientSkip { message }) => {
                        telemetry_unknown |= !increment_counter(skipped_objects);
                        telemetry_unknown |= !add_bytes(&mut bytes_processed, object_size);
@@ -1679,7 +1957,8 @@ mod resume_loop_tests {
    use crate::heal::storage::{HealLifecycleExpiryContext, HealListItem, HealObjectInfo, HealStorageAPI};
    use crate::heal::storage_api::status::BucketInfo;
    use crate::heal::{
-        BUCKET_META_PREFIX, DiskOption, DiskStore, EcstoreError, Endpoint, HealDiskExt as _, RUSTFS_META_BUCKET, new_disk,
+        BUCKET_META_PREFIX, DiskOption, DiskStore, EcstoreError, Endpoint, HealDiskExt as _, POOL_META_NAME, RUSTFS_META_BUCKET,
        new_disk,
    };
    use crate::{Error, Result};
    use rustfs_heal_contracts::heal_channel::{HealOpts, HealRequestSource};
@@ -1773,6 +2052,28 @@ mod resume_loop_tests {
        assert!(!target_outcomes_complete(&duplicate, &["replacement-a".to_string()]));
    }
    fn replacement_target_ok_result(endpoint: &str, object: &str) -> HealResultItem {
        HealResultItem {
            object: object.to_string(),
            object_size: 1024,
            before: Infos {
                drives: vec![HealDriveInfo {
                    endpoint: endpoint.to_string(),
                    state: "missing".to_string(),
                    ..Default::default()
                }],
            },
            after: Infos {
                drives: vec![HealDriveInfo {
                    endpoint: endpoint.to_string(),
                    state: "ok".to_string(),
                    ..Default::default()
                }],
            },
            ..Default::default()
        }
    }
    #[derive(Clone)]
    struct Page {
        items: Vec<HealListItem>,
@@ -1783,6 +2084,8 @@ mod resume_loop_tests {
    #[derive(Clone)]
    enum HealOutcome {
        Ok,
        /// The object has no metadata on any disk in the selected set.
        FileNotFound,
        /// The version vanished before heal ran (deleted mid-heal).
        VersionNotFound,
        /// A transient infrastructure condition (offline disk / unmet quorum):
@@ -1808,6 +2111,10 @@ mod resume_loop_tests {
        /// Target-specific physical readback evidence per `compose_key`; the
        /// fake models a healthy backend unless a test explicitly revokes it.
        replacement_commit_evidence: Mutex<HashMap<String, ReplacementCommitEvidence>>,
        ordinary_pool_metadata_required: AtomicBool,
        pool_metadata_opts: Mutex<Vec<HealOpts>>,
        pool_metadata_not_applicable: AtomicBool,
        fail_pool_metadata_scope: AtomicBool,
        lifecycle_expired: Mutex<HashSet<String>>,
        /// every heal_object call recorded as (name, version_id)
        heal_calls: Mutex<Vec<(String, Option<String>)>>,
@@ -1896,11 +2203,14 @@ mod resume_loop_tests {
        }
        async fn heal_object(
            &self,
-            _bucket: &str,
+            bucket: &str,
            object: &str,
            version_id: Option<&str>,
-            _opts: &HealOpts,
+            opts: &HealOpts,
        ) -> Result<(HealResultItem, Option<Error>)> {
            if bucket == RUSTFS_META_BUCKET && object == POOL_META_NAME {
                self.pool_metadata_opts.lock().expect("metadata options").push(*opts);
            }
            self.heal_calls
                .lock()
                .unwrap()
@@ -1909,6 +2219,7 @@ mod resume_loop_tests {
            let outcome = self.outcomes.lock().unwrap().get(&key).cloned().unwrap_or(HealOutcome::Ok);
            match outcome {
                HealOutcome::Ok => Ok((self.results.lock().unwrap().get(&key).cloned().unwrap_or_default(), None)),
                HealOutcome::FileNotFound => Ok((HealResultItem::default(), Some(Error::Storage(EcstoreError::FileNotFound)))),
                HealOutcome::VersionNotFound => {
                    Ok((HealResultItem::default(), Some(Error::Storage(EcstoreError::FileVersionNotFound))))
                }
@@ -1922,6 +2233,30 @@ mod resume_loop_tests {
        async fn heal_format(&self, _dry: bool) -> Result<(HealResultItem, Option<Error>)> {
            Ok((HealResultItem::default(), None))
        }
        async fn heal_pool_metadata(&self, opts: &HealOpts) -> Result<Vec<HealResultItem>> {
            if !self.ordinary_pool_metadata_required.load(Ordering::SeqCst) {
                return Ok(Vec::new());
            }
            if !self.replacement_pool_metadata_applies(opts).await? {
                return Ok(Vec::new());
            }
            let (result, error) = self.heal_object(RUSTFS_META_BUCKET, POOL_META_NAME, None, opts).await?;
            if let Some(error) = error {
                return Err(error);
            }
            Ok(vec![result])
        }
        async fn replacement_pool_metadata_applies(&self, opts: &HealOpts) -> Result<bool> {
            if self.fail_pool_metadata_scope.load(Ordering::SeqCst) {
                return Err(Error::other("injected pool metadata scope failure"));
            }
            if self.pool_metadata_not_applicable.load(Ordering::SeqCst) {
                assert_eq!(opts.pool, Some(0));
                assert_eq!(opts.set, Some(1));
                return Ok(false);
            }
            Ok(true)
        }
        async fn replacement_targets_have_version(
            &self,
            _bucket: &str,
@@ -2413,6 +2748,115 @@ mod resume_loop_tests {
        assert!(state.completed, "successful data heal must be persisted before cleanup is attempted");
    }
    #[tokio::test]
    async fn ordinary_set_heals_pool_metadata_without_replacement_generation_or_targets() {
        let env = make_env().await;
        env.storage.ordinary_pool_metadata_required.store(true, Ordering::SeqCst);
        env.storage
            .set_result(POOL_META_NAME, None, replacement_target_ok_result("metadata-disk", POOL_META_NAME));
        assert!(env.healer.replacement_task_id.is_none());
        assert!(env.healer.target_endpoints.is_empty());
        env.healer
            .execute_heal_with_resume(&[], "pool_0_set_0", &env.resume, &env.checkpoint)
            .await
            .expect("ordinary set recovery should repair metadata even without user buckets");
        assert_eq!(env.storage.calls(), vec![(POOL_META_NAME.to_string(), None)]);
        {
            let opts = env.storage.pool_metadata_opts.lock().expect("metadata options");
            assert_eq!(opts.len(), 1);
            assert_eq!((opts[0].pool, opts[0].set), (Some(0), Some(0)));
        }
        let state = env.resume.get_state().await;
        assert!(state.completed);
        assert_eq!(state.successful_objects, 1, "metadata must enter durable completion counters");
    }
    #[tokio::test]
    async fn ordinary_set_pool_metadata_respects_non_owner_and_dry_run() {
        let mut env = make_env().await;
        env.storage.ordinary_pool_metadata_required.store(true, Ordering::SeqCst);
        env.storage.pool_metadata_not_applicable.store(true, Ordering::SeqCst);
        env.healer.heal_opts.pool = Some(0);
        env.healer.heal_opts.set = Some(1);
        env.healer
            .execute_heal_with_resume(&[], "pool_0_set_1", &env.resume, &env.checkpoint)
            .await
            .expect("a valid non-owner set must not invent a metadata replica");
        assert!(env.storage.calls().is_empty());
        assert_eq!(env.resume.get_state().await.successful_objects, 0);
        let mut env = make_env().await;
        env.storage.ordinary_pool_metadata_required.store(true, Ordering::SeqCst);
        env.healer.heal_opts.dry_run = true;
        env.healer.heal_opts.remove = true;
        env.healer.heal_opts.no_lock = true;
        env.storage.set_replacement_commit_evidence(POOL_META_NAME, None, false);
        env.healer
            .execute_heal_with_resume(&[], "pool_0_set_0", &env.resume, &env.checkpoint)
            .await
            .expect("ordinary dry-run metadata work must not require a replacement commit");
        assert_eq!(env.storage.calls(), vec![(POOL_META_NAME.to_string(), None)]);
        let opts = env.storage.pool_metadata_opts.lock().expect("metadata options");
        assert_eq!(opts.len(), 1);
        assert!(opts[0].dry_run);
        assert!(!opts[0].remove);
        assert!(!opts[0].no_lock);
    }
    #[tokio::test]
    async fn ordinary_set_missing_pool_metadata_preserves_retry_state() {
        let env = make_env().await;
        env.storage.ordinary_pool_metadata_required.store(true, Ordering::SeqCst);
        env.storage.set_outcome(POOL_META_NAME, None, HealOutcome::FileNotFound);
        let error = env
            .healer
            .execute_heal_with_resume(&[], "pool_0_set_0", &env.resume, &env.checkpoint)
            .await
            .expect_err("missing required pool metadata must prevent ordinary set completion");
        assert!(matches!(error, Error::TransientSkip { .. }));
        let state = env.resume.get_state().await;
        assert!(!state.completed);
        assert_eq!(state.retry_count, 1);
        assert!(CheckpointManager::has_checkpoint(&env.healer.disk, &env.task_id).await);
    }
    #[tokio::test]
    async fn ordinary_set_pool_metadata_timeout_keeps_control_error() {
        let env = make_env().await;
        env.storage.ordinary_pool_metadata_required.store(true, Ordering::SeqCst);
        env.storage.set_outcome(POOL_META_NAME, None, HealOutcome::Timeout);
        let error = env
            .healer
            .execute_heal_with_resume(&[], "pool_0_set_0", &env.resume, &env.checkpoint)
            .await
            .expect_err("metadata timeout must abort the ordinary set pass");
        assert!(matches!(error, Error::TaskTimeout));
        assert!(!env.resume.get_state().await.completed);
    }
    #[tokio::test]
    async fn ordinary_set_pool_metadata_rejects_mismatched_explicit_scope() {
        let mut env = make_env().await;
        env.storage.ordinary_pool_metadata_required.store(true, Ordering::SeqCst);
        env.healer.heal_opts.pool = Some(1);
        let error = env
            .healer
            .execute_heal_with_resume(&[], "pool_0_set_0", &env.resume, &env.checkpoint)
            .await
            .expect_err("explicit metadata scope must agree with the resumed set");
        assert!(matches!(error, Error::TaskExecutionFailed { .. }));
        assert!(env.storage.calls().is_empty(), "scope mismatch must fail before metadata mutation");
        assert!(!env.resume.get_state().await.completed);
    }
    #[tokio::test]
    async fn replacement_completion_keeps_resume_artifacts_until_marker_cleanup() {
        let env = make_env_with_targets(vec!["replacement-a".to_string()]).await;
@@ -2444,6 +2888,8 @@ mod resume_loop_tests {
            HealRequestSource::AutoHeal,
        )
        .with_replacement_targets(vec!["replacement-a".to_string()], Some(replacement_task_id.clone()));
        env.storage
            .set_result(POOL_META_NAME, None, replacement_target_ok_result("replacement-a", POOL_META_NAME));
        healer
            .heal_erasure_set(&["b".to_string()], "pool_0_set_0")
@@ -2461,9 +2907,263 @@ mod resume_loop_tests {
            CheckpointManager::has_checkpoint(&env.healer.disk, &replacement_task_id).await,
            "the checkpoint must survive until the caller clears the healing marker"
        );
        assert_eq!(env.storage.calls(), vec![(POOL_META_NAME.to_string(), None)]);
        drop(checkpoint);
    }
    #[tokio::test]
    async fn replacement_pool_metadata_non_owner_completes_but_missing_owner_retries() {
        for owns_pool_metadata in [false, true] {
            let env = make_env_with_targets(vec!["replacement-a".to_string()]).await;
            let replacement_task_id = ResumeUtils::generate_task_id();
            let set_index = usize::from(!owns_pool_metadata);
            let set_disk_id = format!("pool_0_set_{set_index}");
            ResumeManager::new_replacement_intent(
                env.healer.disk.clone(),
                replacement_task_id.clone(),
                set_disk_id.clone(),
                vec!["b".to_string()],
                vec!["replacement-a".to_string()],
                vec![crate::heal::resume::ReplacementTargetIdentity {
                    endpoint: "replacement-a".to_string(),
                    canonical_path: "/mnt/replacement-a".to_string(),
                    physical_device_ids: vec!["device-a".to_string()],
                    filesystem_identity: "1:2:3".to_string(),
                }],
            )
            .await
            .expect("replacement intent should persist");
            env.storage
                .pool_metadata_not_applicable
                .store(!owns_pool_metadata, Ordering::SeqCst);
            env.storage.set_outcome(POOL_META_NAME, None, HealOutcome::FileNotFound);
            let healer = ErasureSetHealer::new(
                env.storage.clone(),
                Arc::new(RwLock::new(HealProgress::new())),
                CancellationToken::new(),
                env.healer.disk.clone(),
                HealOpts {
                    pool: Some(0),
                    set: Some(set_index),
                    ..Default::default()
                },
                HealRequestSource::AutoHeal,
            )
            .with_replacement_targets(vec!["replacement-a".to_string()], Some(replacement_task_id.clone()));
            let result = healer.heal_erasure_set(&["b".to_string()], &set_disk_id).await;
            let state = ResumeManager::load_replacement_intent(env.healer.disk.clone(), &replacement_task_id)
                .await
                .expect("replacement state must remain until marker cleanup")
                .get_state()
                .await;
            if owns_pool_metadata {
                let error = result.expect_err("missing metadata in the owner set must keep replacement incomplete");
                assert!(error.to_string().contains("Replacement erasure set heal incomplete"));
                assert!(!state.completed);
                assert_eq!(state.replacement_phase, crate::heal::resume::ReplacementPhase::Intent);
                assert_eq!(state.retry_count, 1);
                assert_eq!(env.storage.calls(), vec![(POOL_META_NAME.to_string(), None)]);
            } else {
                result.expect("a non-owner set must complete without a pool metadata replica");
                assert!(state.completed);
                assert_eq!(state.replacement_phase, crate::heal::resume::ReplacementPhase::Verified);
                assert_eq!(state.retry_count, 0);
                assert!(env.storage.calls().is_empty(), "non-owner sets must not attempt pool metadata repair");
            }
        }
    }
    #[tokio::test]
    async fn replacement_pool_metadata_unknown_scope_cannot_complete() {
        let env = make_env_with_targets(vec!["replacement-a".to_string()]).await;
        let healer = ErasureSetHealer::new(
            env.storage.clone(),
            Arc::new(RwLock::new(HealProgress::new())),
            CancellationToken::new(),
            env.healer.disk.clone(),
            HealOpts {
                pool: Some(0),
                set: Some(0),
                ..Default::default()
            },
            HealRequestSource::AutoHeal,
        )
        .with_replacement_targets(vec!["replacement-a".to_string()], Some("generation-a".to_string()));
        env.storage.fail_pool_metadata_scope.store(true, Ordering::SeqCst);
        env.storage
            .set_result(POOL_META_NAME, None, replacement_target_ok_result("replacement-a", POOL_META_NAME));
        let mut processed_objects = 0;
        let mut successful_objects = 0;
        let mut failed_objects = 0;
        let mut skipped_objects = 0;
        let error = healer
            .heal_pool_metadata(
                "pool_0_set_0",
                &mut super::ErasureSetPassCounters {
                    processed_objects: &mut processed_objects,
                    successful_objects: &mut successful_objects,
                    failed_objects: &mut failed_objects,
                    skipped_objects: &mut skipped_objects,
                },
                &env.resume,
                &env.checkpoint,
            )
            .await
            .expect_err("unknown metadata placement must keep replacement incomplete");
        assert!(error.to_string().contains("injected pool metadata scope failure"));
        assert!(env.storage.calls().is_empty());
        assert_eq!((processed_objects, successful_objects, failed_objects, skipped_objects), (0, 0, 0, 0));
    }
    #[tokio::test]
    async fn replacement_pool_metadata_readback_failure_schedules_retry() {
        let env = make_env_with_targets(vec!["replacement-a".to_string()]).await;
        let replacement_task_id = ResumeUtils::generate_task_id();
        ResumeManager::new_replacement_intent(
            env.healer.disk.clone(),
            replacement_task_id.clone(),
            "pool_0_set_0".to_string(),
            vec!["b".to_string()],
            vec!["replacement-a".to_string()],
            vec![crate::heal::resume::ReplacementTargetIdentity {
                endpoint: "replacement-a".to_string(),
                canonical_path: "/mnt/replacement-a".to_string(),
                physical_device_ids: vec!["device-a".to_string()],
                filesystem_identity: "1:2:3".to_string(),
            }],
        )
        .await
        .expect("replacement intent should persist");
        env.storage
            .set_result(POOL_META_NAME, None, replacement_target_ok_result("replacement-a", POOL_META_NAME));
        env.storage.set_replacement_commit_evidence(POOL_META_NAME, None, false);
        let healer = ErasureSetHealer::new(
            env.storage.clone(),
            Arc::new(RwLock::new(HealProgress::new())),
            CancellationToken::new(),
            env.healer.disk.clone(),
            HealOpts::default(),
            HealRequestSource::AutoHeal,
        )
        .with_replacement_targets(vec!["replacement-a".to_string()], Some(replacement_task_id.clone()));
        let error = healer
            .heal_erasure_set(&["b".to_string()], "pool_0_set_0")
            .await
            .expect_err("unconfirmed pool metadata readback must keep the replacement incomplete");
        assert!(error.to_string().contains("Replacement erasure set heal incomplete"));
        let state = ResumeManager::load_replacement_intent(env.healer.disk.clone(), &replacement_task_id)
            .await
            .expect("replacement retry state must remain")
            .get_state()
            .await;
        assert!(!state.completed);
        assert_eq!(state.replacement_phase, crate::heal::resume::ReplacementPhase::Intent);
        assert_eq!(state.retry_count, 1);
        assert_eq!(env.storage.calls(), vec![(POOL_META_NAME.to_string(), None)]);
    }
    #[tokio::test]
    async fn admin_recreate_target_heals_pool_metadata_before_completion() {
        let env = make_env_with_targets(vec!["replacement-a".to_string()]).await;
        let healer = ErasureSetHealer::new(
            env.storage.clone(),
            Arc::new(RwLock::new(HealProgress::new())),
            CancellationToken::new(),
            env.healer.disk.clone(),
            HealOpts {
                recreate: true,
                remove: true,
                no_lock: true,
                ..Default::default()
            },
            HealRequestSource::Admin,
        )
        .with_pool_metadata_targets(vec!["replacement-a".to_string()]);
        env.storage
            .set_result(POOL_META_NAME, None, replacement_target_ok_result("replacement-a", POOL_META_NAME));
        healer
            .execute_heal_with_resume(&["b".to_string()], "pool_0_set_0", &env.resume, &env.checkpoint)
            .await
            .expect("admin recreate should heal and verify pool metadata");
        assert!(env.resume.get_state().await.completed);
        assert_eq!(env.storage.calls(), vec![(POOL_META_NAME.to_string(), None)]);
        let opts = env.storage.pool_metadata_opts.lock().expect("metadata options");
        assert_eq!(opts.len(), 1);
        assert_eq!((opts[0].pool, opts[0].set), (Some(0), Some(0)));
        assert!(opts[0].recreate);
        assert!(!opts[0].remove);
        assert!(!opts[0].no_lock);
    }
    #[tokio::test]
    async fn admin_recreate_pool_metadata_validates_owner_scope_before_io() {
        for (non_owner, unknown_scope, pool) in [(true, false, None), (false, true, None), (false, false, Some(1))] {
            let env = make_env_with_targets(vec!["replacement-a".to_string()]).await;
            env.storage.pool_metadata_not_applicable.store(non_owner, Ordering::SeqCst);
            env.storage.fail_pool_metadata_scope.store(unknown_scope, Ordering::SeqCst);
            let healer = ErasureSetHealer::new(
                env.storage.clone(),
                Arc::new(RwLock::new(HealProgress::new())),
                CancellationToken::new(),
                env.healer.disk.clone(),
                HealOpts {
                    recreate: true,
                    pool,
                    ..Default::default()
                },
                HealRequestSource::Admin,
            )
            .with_pool_metadata_targets(vec!["replacement-a".to_string()]);
            let set_disk_id = if non_owner { "pool_0_set_1" } else { "pool_0_set_0" };
            let result = healer
                .execute_heal_with_resume(&[], set_disk_id, &env.resume, &env.checkpoint)
                .await;
            assert_eq!(result.is_ok(), non_owner, "only a known non-owner may skip metadata: {result:?}");
            assert!(env.storage.calls().is_empty(), "scope validation must precede metadata I/O");
            assert_eq!(env.resume.get_state().await.completed, non_owner);
        }
    }
    #[tokio::test]
    async fn admin_recreate_pool_metadata_readback_failure_keeps_resume_state() {
        let env = make_env_with_targets(vec!["replacement-a".to_string()]).await;
        let healer = ErasureSetHealer::new(
            env.storage.clone(),
            Arc::new(RwLock::new(HealProgress::new())),
            CancellationToken::new(),
            env.healer.disk.clone(),
            HealOpts {
                recreate: true,
                pool: Some(0),
                set: Some(0),
                ..Default::default()
            },
            HealRequestSource::Admin,
        )
        .with_pool_metadata_targets(vec!["replacement-a".to_string()]);
        env.storage
            .set_result(POOL_META_NAME, None, replacement_target_ok_result("replacement-a", POOL_META_NAME));
        env.storage.set_replacement_commit_evidence(POOL_META_NAME, None, false);
        let error = healer
            .execute_heal_with_resume(&["b".to_string()], "pool_0_set_0", &env.resume, &env.checkpoint)
            .await
            .expect_err("unconfirmed admin recreate pool metadata must keep the set incomplete");
        assert!(error.to_string().contains("Erasure set heal incomplete"));
        let state = env.resume.get_state().await;
        assert!(!state.completed);
        assert_eq!(state.retry_count, 1);
        assert_eq!(env.storage.calls(), vec![(POOL_META_NAME.to_string(), None)]);
    }
    #[tokio::test]
    async fn retry_exhaustion_keeps_resume_artifacts_for_recovery() {
        let env = make_env().await;
@@ -506,6 +506,25 @@ fn active_heal_for_dedup_key(active_heals: &HashMap<String, Arc<HealTask>>, key:
        .map(|(task_id, task)| (task_id.clone(), task.heal_type.clone()))
 }
 fn request_matches_task(request: &HealRequest, task: &HealTask) -> bool {
    request.heal_type == task.heal_type
        && request.options == task.options
        && request.priority == task.priority
        && request.source == task.source
        && request.retry_attempts == task.retry_attempts
        && request.heal_endpoints == task.heal_endpoints
 }
 fn request_matches_request(request: &HealRequest, existing: &HealRequest) -> bool {
    request.heal_type == existing.heal_type
        && request.options == existing.options
        && request.priority == existing.priority
        && request.source == existing.source
        && request.force_start == existing.force_start
        && request.retry_attempts == existing.retry_attempts
        && request.heal_endpoints == existing.heal_endpoints
 }
 fn retrying_heal_for_dedup_key(retrying_heals: &HashMap<String, RetryingHeal>, key: &str) -> Option<(String, HealType)> {
    retrying_heals
        .iter()
@@ -1508,7 +1527,7 @@ impl HealManager {
        request: HealRequest,
        preserve_alias: bool,
    ) -> Result<HealAdmissionReceipt> {
-        self.submit_heal_request_with_receipt_alias_and_mrf_notice(request, preserve_alias, None)
+        self.submit_heal_request_with_receipt_alias_and_mrf_notice(request, preserve_alias, true, None)
            .await
    }
@@ -1544,14 +1563,27 @@ impl HealManager {
        request: HealRequest,
        mrf_notice_target: MrfRepairNoticeTarget,
    ) -> Result<HealAdmissionReceipt> {
-        self.submit_heal_request_with_receipt_alias_and_mrf_notice(request, true, Some(mrf_notice_target))
+        self.submit_heal_request_with_receipt_alias_and_mrf_notice(request, true, true, Some(mrf_notice_target))
            .await
    }
    pub(crate) async fn durable_mrf_repair_anchor(
        &self,
        intent: &rustfs_common::mrf_channel::MrfIntent,
    ) -> Option<rustfs_common::mrf_channel::MrfDurableRepairAnchor> {
        match self.storage.mrf_bucket_incarnation_id(intent.bucket.as_ref()).await {
            Ok(Some(bucket_incarnation_id)) => {
                rustfs_common::mrf_channel::MrfDurableRepairAnchor::from_intent(intent, bucket_incarnation_id)
            }
            Ok(None) | Err(_) => None,
        }
    }
    async fn submit_heal_request_with_receipt_alias_and_mrf_notice(
        &self,
        request: HealRequest,
        preserve_alias: bool,
        accept_same_request_id_replay: bool,
        mrf_notice_target: Option<MrfRepairNoticeTarget>,
    ) -> Result<HealAdmissionReceipt> {
        let admission_start = Instant::now();
@@ -1613,6 +1645,68 @@ impl HealManager {
        pause_duplicate_admission_after_active_lock(&request.id).await;
        let mut queue = self.heal_queue.lock().await;
        let retrying_heals = self.retrying_heals.lock().await;
        let request_id_admission = active_heals
            .get(&request.id)
            .map(|task| (request_matches_task(&request, task), "active"))
            .or_else(|| {
                queue
                    .requests()
                    .find(|queued| queued.id == request.id)
                    .map(|queued| (request_matches_request(&request, queued), "queued"))
            })
            .or_else(|| {
                retrying_heals
                    .get(&request.id)
                    .map(|retrying| (request_matches_request(&request, &retrying.request), "retrying"))
            });
        if let Some((matches_existing, duplicate_state)) = request_id_admission {
            let admission = if matches_existing {
                if accept_same_request_id_replay {
                    HealAdmissionResult::Accepted
                } else {
                    Self::duplicate_admission_for_request(&request, &config)
                }
            } else {
                HealAdmissionResult::Dropped(HealAdmissionDropReason::AlreadyRunning)
            };
            if matches!(admission, HealAdmissionResult::Accepted | HealAdmissionResult::Merged)
                && let Some(target) = mrf_notice_target
            {
                let mut targets = lock_mrf_repair_notice_targets(&self.mrf_repair_notice_targets);
                Self::insert_mrf_repair_notice_target(&mut targets, &request.id, target);
            }
            drop(retrying_heals);
            drop(queue);
            drop(active_heals);
            let lock_phase = lock_phase_start.elapsed();
            Self::record_admission_metric(request.source, admission, "duplicate");
            self.record_admission_observation(HealAdmissionObservation {
                source,
                result: admission,
                context: "duplicate",
                force_start,
                displaced: false,
                start_duration: admission_start.elapsed(),
                lock_phase,
            });
            debug!(
                target: "rustfs::heal::manager",
                event = EVENT_HEAL_QUEUE_ADMISSION,
                component = LOG_COMPONENT_HEAL,
                subsystem = LOG_SUBSYSTEM_MANAGER,
                request_id = %request.id,
                duplicate_state,
                result = admission.result_label(),
                reason = admission.reason_label(),
                "Heal queue admission reused an existing request id"
            );
            return Ok(HealAdmissionReceipt {
                result: admission,
                task_id: request.id,
            });
        }
        let duplicate = (!request.force_start).then(|| {
            active_heal_for_dedup_key(&active_heals, &dedup_key)
                .map(|(task_id, _)| (task_id, "active"))
@@ -1819,7 +1913,10 @@ impl HealManager {
    /// Submit heal request.
    pub async fn submit_heal_request(&self, request: HealRequest) -> Result<HealAdmissionResult> {
-        Ok(self.submit_heal_request_with_receipt_and_alias(request, true).await?.result)
+        Ok(self
            .submit_heal_request_with_receipt_alias_and_mrf_notice(request, true, false, None)
            .await?
            .result)
    }
    /// Get task status
@@ -733,23 +733,30 @@ pub(super) fn mrf_verified_repair_event_for_target(
        HealObjectDisposition::AuthoritativelyAbsent => MrfVerifiedRepairDisposition::AuthoritativelyAbsent,
        _ => return None,
    };
-    if target.kind != MrfKind::PartialWrite {
+    let expected_kind = match target.kind {
-        return None;
+        MrfKind::DecodeFailure => HealObjectKind::Decode,
-    }
+        MrfKind::MetadataCorruption => HealObjectKind::Metadata,
-    let expected_kind = HealObjectKind::Object;
+        MrfKind::PartialWrite => HealObjectKind::Object,
    };
    if outcome.identity.kind != expected_kind
        || outcome.identity.bucket.as_str() != target.bucket.as_ref()
        || outcome.identity.object.as_str() != target.object.as_ref()
    {
        return None;
    }
-    let version_id = target.version_id.filter(|bytes| *bytes != [0; 16]);
+    let version_id = (!matches!(target.kind, MrfKind::MetadataCorruption))
        .then_some(target.version_id)
        .flatten()
        .filter(|bytes| *bytes != [0; 16]);
    let expected_version = version_id.map(|bytes| uuid::Uuid::from_bytes(bytes).to_string());
    if outcome.identity.version_id != expected_version {
        return None;
    }
-    let expected_pool = target.scope.and_then(|scope| usize::try_from(scope.pool_index).ok());
+    let scope = (!matches!(target.kind, MrfKind::MetadataCorruption))
-    let expected_set = target.scope.and_then(|scope| usize::try_from(scope.set_index).ok());
+        .then_some(target.scope)
        .flatten();
    let expected_pool = scope.and_then(|scope| usize::try_from(scope.pool_index).ok());
    let expected_set = scope.and_then(|scope| usize::try_from(scope.set_index).ok());
    if outcome.identity.pool_index != expected_pool || outcome.identity.set_index != expected_set {
        return None;
    }
@@ -759,7 +766,7 @@ pub(super) fn mrf_verified_repair_event_for_target(
        bucket: target.bucket.clone(),
        object: target.object.clone(),
        version_id,
-        scope: target.scope,
+        scope,
        lease: target.lease,
        bucket_incarnation_id,
        disposition,
@@ -555,6 +555,10 @@ async fn completed_retention_scheduler_preserves_progress_aliases_and_atomic_han
 #[async_trait::async_trait]
 impl HealStorageAPI for MockStorage {
    async fn heal_pool_metadata(&self, _opts: &HealOpts) -> Result<Vec<HealResultItem>> {
        Ok(Vec::new())
    }
    async fn get_object_meta(&self, _bucket: &str, _object: &str) -> Result<Option<HealObjectInfo>> {
        Ok(None)
    }
@@ -1142,18 +1146,48 @@ fn mrf_verified_repair_event_requires_positive_exact_identity() {
    assert_eq!(event.bucket_incarnation_id, incarnation);
    assert_eq!(event.disposition, MrfVerifiedRepairDisposition::Repaired);
-    assert!(
+    let decode_target = MrfRepairNoticeTarget {
-        mrf_verified_repair_event_for_target(
+        kind: MrfKind::DecodeFailure,
-            &MrfRepairNoticeTarget {
+        ..target.clone()
-                kind: MrfKind::DecodeFailure,
+    };
-                ..target.clone()
+    let decode_outcome = HealObjectOutcome {
-            },
+        identity: HealObjectIdentity {
-            &matching
+            kind: HealObjectKind::Decode,
-        )
+            ..matching.identity.clone()
-        .is_none(),
+        },
-        "only receipt-producing partial-write object heals can publish verified events today"
+        ..matching.clone()
    };
    let decode_event =
        mrf_verified_repair_event_for_target(&decode_target, &decode_outcome).expect("decode repairs publish exact proofs");
    assert_eq!(decode_event.kind, MrfKind::DecodeFailure);
    assert_eq!(
        decode_event.scope,
        Some(MrfScope {
            pool_index: 1,
            set_index: 2
        })
    );
    let metadata_target = MrfRepairNoticeTarget {
        kind: MrfKind::MetadataCorruption,
        ..target.clone()
    };
    let metadata_outcome = HealObjectOutcome {
        identity: HealObjectIdentity {
            kind: HealObjectKind::Metadata,
            version_id: None,
            pool_index: None,
            set_index: None,
            ..matching.identity.clone()
        },
        ..matching.clone()
    };
    let metadata_event =
        mrf_verified_repair_event_for_target(&metadata_target, &metadata_outcome).expect("metadata repairs publish exact proofs");
    assert_eq!(metadata_event.kind, MrfKind::MetadataCorruption);
    assert_eq!(metadata_event.version_id, None);
    assert_eq!(metadata_event.scope, None);
    for rejected in [
        HealObjectOutcome {
            disposition: HealObjectDisposition::Unknown,
@@ -1839,7 +1873,7 @@ async fn test_submit_heal_request_returns_merged_for_duplicate() {
    let storage: Arc<dyn HealStorageAPI> = Arc::new(MockStorage);
    let manager = HealManager::new(storage, None);
-    let request = HealRequest::new(
+    let mut request = HealRequest::new(
        HealType::Object {
            bucket: "bucket".to_string(),
            object: "object".to_string(),
@@ -1856,6 +1890,7 @@ async fn test_submit_heal_request_returns_merged_for_duplicate() {
            .expect("first request should be accepted"),
        HealAdmissionResult::Accepted
    );
    request.id = uuid::Uuid::new_v4().to_string();
    assert_eq!(
        manager
            .submit_heal_request(request)
@@ -3695,7 +3730,7 @@ async fn test_submit_heal_request_returns_merged_before_full_for_duplicate() {
        }),
    );
-    let request = HealRequest::new(
+    let mut request = HealRequest::new(
        HealType::Object {
            bucket: "bucket".to_string(),
            object: "object".to_string(),
@@ -3712,6 +3747,7 @@ async fn test_submit_heal_request_returns_merged_before_full_for_duplicate() {
            .expect("first request should be accepted"),
        HealAdmissionResult::Accepted
    );
    request.id = uuid::Uuid::new_v4().to_string();
    assert_eq!(
        manager
            .submit_heal_request(request)
@@ -4537,6 +4573,48 @@ async fn test_force_start_marks_dedup_key_for_future_duplicates() {
    );
 }
 #[tokio::test]
 async fn same_request_id_replay_reuses_existing_task_without_force_start_duplication() {
    let storage: Arc<dyn HealStorageAPI> = Arc::new(MockStorage);
    let manager = HealManager::new(storage, None);
    let mut original = admin_prefix_request("bucket", "logs/");
    original.force_start = true;
    let original_id = original.id.clone();
    let accepted = manager
        .submit_heal_request_with_receipt(original.clone())
        .await
        .expect("original forceStart request should queue");
    assert_eq!(accepted.result, HealAdmissionResult::Accepted);
    assert_eq!(accepted.task_id, original_id);
    let replayed = manager
        .submit_heal_request_with_receipt(original.clone())
        .await
        .expect("same request id and payload should reuse the existing task");
    assert_eq!(replayed.result, HealAdmissionResult::Accepted);
    assert_eq!(replayed.task_id, original_id);
    assert_eq!(
        manager.get_queue_length().await,
        1,
        "exact forceStart replay must not create a second queued task"
    );
    let mut changed = original;
    changed.options.remove_corrupted = true;
    let changed = manager
        .submit_heal_request_with_receipt(changed)
        .await
        .expect("same request id with a changed payload should fail closed");
    assert_eq!(changed.result, HealAdmissionResult::Dropped(HealAdmissionDropReason::AlreadyRunning));
    assert_eq!(changed.task_id, original_id);
    assert_eq!(
        manager.get_queue_length().await,
        1,
        "same-id conflict must not displace or duplicate the original task"
    );
 }
 #[test]
 fn test_running_heal_set_counts_groups_set_scoped_tasks() {
    let storage: Arc<dyn HealStorageAPI> = Arc::new(MockStorage);
@@ -27,10 +27,10 @@ pub mod task;
 pub mod utils;
 use storage_api::owner::{
-    ECSTORE_BUCKET_META_PREFIX, ECSTORE_DATA_USAGE_CACHE_NAME, ECSTORE_HEALING_MARKER_PATH, ECSTORE_RUSTFS_META_BUCKET,
+    ECSTORE_BUCKET_META_PREFIX, ECSTORE_DATA_USAGE_CACHE_NAME, ECSTORE_HEALING_MARKER_PATH, ECSTORE_POOL_META_NAME,
-    EcstoreConditionalFileUpdate, EcstoreDeleteOptions, EcstoreDiskAPI, EcstoreDiskBytes, EcstoreDiskError, EcstoreDiskOption,
+    ECSTORE_RUSTFS_META_BUCKET, EcstoreConditionalFileUpdate, EcstoreDeleteOptions, EcstoreDiskAPI, EcstoreDiskBytes,
-    EcstoreDiskResult, EcstoreDiskStore, EcstoreEndpoint, EcstoreErrorType, EcstoreStorageError, EcstoreStore, ObjectIO,
+    EcstoreDiskError, EcstoreDiskOption, EcstoreDiskResult, EcstoreDiskStore, EcstoreEndpoint, EcstoreErrorType,
-    ObjectOperations, ecstore_local_disk_map_read, ecstore_new_disk,
+    EcstoreStorageError, EcstoreStore, ObjectIO, ObjectOperations, ecstore_local_disk_map_read, ecstore_new_disk,
 };
 pub use erasure_healer::ErasureSetHealer;
@@ -41,6 +41,7 @@ pub use task::{HealOptions, HealPriority, HealRequest, HealTask, HealType};
 pub(crate) const DATA_USAGE_CACHE_NAME: &str = ECSTORE_DATA_USAGE_CACHE_NAME;
 pub(crate) const BUCKET_META_PREFIX: &str = ECSTORE_BUCKET_META_PREFIX;
 pub(crate) const RUSTFS_META_BUCKET: &str = ECSTORE_RUSTFS_META_BUCKET;
 pub(crate) const POOL_META_NAME: &str = ECSTORE_POOL_META_NAME;
 /// Marker written to every local disk while the process runs; removed by
 /// [`clear_unclean_shutdown_markers`] on graceful shutdown. Finding it at
@@ -110,7 +110,8 @@ impl HealObjectReceipt {
            && self.identity.version_id == expected.version_id
            && self.identity.pool_index == expected.pool_index
            && self.identity.set_index == expected.set_index
-            && self.identity.bucket_incarnation_id.is_some()
+            && self.identity.bucket_incarnation_id == expected.bucket_incarnation_id
            && expected.bucket_incarnation_id.is_some()
    }
 }
@@ -496,21 +497,31 @@ mod canonical_outcome_tests {
    #[test]
    fn positive_receipt_requires_exact_identity_and_bucket_incarnation() {
-        let expected = item(HealObjectDisposition::Unknown).identity;
+        let incarnation = Uuid::new_v4();
        let expected = HealObjectIdentity {
            bucket_incarnation_id: Some(incarnation),
            ..item(HealObjectDisposition::Unknown).identity
        };
        let mut receipt = HealObjectReceipt {
            identity: expected.clone(),
            disposition: HealObjectDisposition::Repaired,
        };
        receipt.identity.bucket_incarnation_id = None;
        assert!(
            !receipt.verified_for(&expected),
            "a positive storage receipt without bucket incarnation must remain untrusted"
        );
        let incarnation = Uuid::new_v4();
        receipt.identity.bucket_incarnation_id = Some(incarnation);
        assert!(receipt.verified_for(&expected));
        receipt.identity.bucket_incarnation_id = Some(Uuid::new_v4());
        assert!(
            !receipt.verified_for(&expected),
            "a storage receipt for a different bucket incarnation must not clear the requested responsibility"
        );
        receipt.identity.version_id = Some("older-version".to_string());
        assert!(
            !receipt.verified_for(&expected),
@@ -20,6 +20,7 @@ use rustfs_madmin::heal_commands::HealResultItem;
 use serde::{Deserialize, Serialize};
 use std::sync::Arc;
 use tracing::{debug, error, warn};
 use uuid::Uuid;
 use super::outcome::{HealObjectDisposition, HealObjectIdentity, HealObjectKind, HealObjectReceipt};
 use super::progress::stable_generation;
@@ -355,6 +356,14 @@ pub trait HealStorageAPI: Send + Sync {
    /// Get bucket info
    async fn get_bucket_info(&self, bucket: &str) -> Result<Option<BucketInfo>>;
    /// Return the current bucket incarnation for exact MRF durable proof
    /// matching. Alternate backends that cannot expose this must return
    /// `None`, leaving replay anchors retained instead of acknowledged with an
    /// incomplete identity.
    async fn mrf_bucket_incarnation_id(&self, _bucket: &str) -> Result<Option<Uuid>> {
        Ok(None)
    }
    /// Aggregate usage-cache baselines for the requested buckets.
    async fn erasure_set_usage_baseline(&self, _buckets: &[String]) -> Result<Option<HealBucketUsageBaseline>> {
        Ok(None)
@@ -383,6 +392,11 @@ pub trait HealStorageAPI: Send + Sync {
    /// Check object exists
    async fn object_exists(&self, bucket: &str, object: &str) -> Result<bool>;
    /// Stable bucket incarnation observed before an object heal starts.
    async fn bucket_incarnation_id(&self, _bucket: &str) -> Result<Option<Uuid>> {
        Ok(None)
    }
    /// Heal object using ecstore
    async fn heal_object(
        &self,
@@ -422,6 +436,23 @@ pub trait HealStorageAPI: Send + Sync {
        Err(Error::other("target-scoped replacement format is unsupported"))
    }
    /// Heal each pool metadata replica owned by the selected live scope.
    ///
    /// A successful result requires every applicable owner to finish; an empty
    /// result is valid only for a known scope with no metadata replica. Backends
    /// without pool metadata must explicitly implement that empty result.
    async fn heal_pool_metadata(&self, _opts: &HealOpts) -> Result<Vec<HealResultItem>> {
        Err(Error::other("pool metadata healing is unsupported"))
    }
    /// Whether the selected replacement set owns the pool metadata replica.
    ///
    /// Only a topology-aware backend may exempt a valid non-owner set. The
    /// conservative default requires the existing repair and readback checks.
    async fn replacement_pool_metadata_applies(&self, _opts: &HealOpts) -> Result<bool> {
        Ok(true)
    }
    /// Read target-specific physical evidence for one replacement version.
    ///
    /// This is only used by automatic replacement healing after the normal
@@ -809,6 +840,14 @@ impl HealStorageAPI for ECStoreHealStorage {
        }
    }
    async fn mrf_bucket_incarnation_id(&self, bucket: &str) -> Result<Option<Uuid>> {
        self.ecstore
            .bucket_incarnation_id(bucket)
            .await
            .map(Some)
            .map_err(Error::Storage)
    }
    async fn erasure_set_usage_baseline(&self, buckets: &[String]) -> Result<Option<HealBucketUsageBaseline>> {
        if buckets.is_empty() {
            return Ok(None);
@@ -1028,6 +1067,14 @@ impl HealStorageAPI for ECStoreHealStorage {
        }
    }
    async fn bucket_incarnation_id(&self, bucket: &str) -> Result<Option<Uuid>> {
        self.ecstore
            .bucket_incarnation_id(bucket)
            .await
            .map(Some)
            .map_err(Error::Storage)
    }
    async fn heal_object(
        &self,
        bucket: &str,
@@ -1238,6 +1285,22 @@ impl HealStorageAPI for ECStoreHealStorage {
            .map_err(Error::Storage)
    }
    async fn heal_pool_metadata(&self, opts: &HealOpts) -> Result<Vec<HealResultItem>> {
        self.ecstore.heal_pool_metadata(opts).await.map_err(Error::Storage)
    }
    async fn replacement_pool_metadata_applies(&self, opts: &HealOpts) -> Result<bool> {
        let pool_index = opts
            .pool
            .ok_or_else(|| Error::other("replacement pool metadata is missing pool scope"))?;
        let set_index = opts
            .set
            .ok_or_else(|| Error::other("replacement pool metadata is missing set scope"))?;
        self.ecstore
            .replacement_pool_metadata_applies(pool_index, set_index)
            .map_err(Error::Storage)
    }
    async fn replacement_targets_have_version(
        &self,
        bucket: &str,
@@ -29,6 +29,7 @@ pub(crate) use rustfs_ecstore::api::error::{Error as EcstoreErrorType, StorageEr
 pub(crate) use rustfs_ecstore::api::runtime::local_disk_map_read as ecstore_local_disk_map_read;
 pub(crate) use rustfs_ecstore::api::storage::{
    ECStore as EcstoreStore, HealLifecycleExpiryContext as EcstoreHealLifecycleExpiryContext,
    POOL_META_NAME as ECSTORE_POOL_META_NAME,
 };
 use rustfs_storage_api as storage_contracts;
@@ -36,10 +37,11 @@ pub(crate) mod owner {
    pub(crate) use super::storage_contracts::{ObjectIO, ObjectOperations};
    pub(crate) use super::{
-        ECSTORE_BUCKET_META_PREFIX, ECSTORE_DATA_USAGE_CACHE_NAME, ECSTORE_HEALING_MARKER_PATH, ECSTORE_RUSTFS_META_BUCKET,
+        ECSTORE_BUCKET_META_PREFIX, ECSTORE_DATA_USAGE_CACHE_NAME, ECSTORE_HEALING_MARKER_PATH, ECSTORE_POOL_META_NAME,
-        EcstoreConditionalFileUpdate, EcstoreDeleteOptions, EcstoreDiskAPI, EcstoreDiskBytes, EcstoreDiskError,
+        ECSTORE_RUSTFS_META_BUCKET, EcstoreConditionalFileUpdate, EcstoreDeleteOptions, EcstoreDiskAPI, EcstoreDiskBytes,
-        EcstoreDiskResult, EcstoreDiskStore, EcstoreEndpoint, EcstoreErrorType, EcstoreHealLifecycleExpiryContext,
+        EcstoreDiskError, EcstoreDiskResult, EcstoreDiskStore, EcstoreEndpoint, EcstoreErrorType,
-        EcstoreStorageError, EcstoreStore, ecstore_load_admin_data_usage_from_backend_cached, ecstore_local_disk_map_read,
+        EcstoreHealLifecycleExpiryContext, EcstoreStorageError, EcstoreStore, ecstore_load_admin_data_usage_from_backend_cached,
        ecstore_local_disk_map_read,
    };
    pub(crate) use super::{EcstoreDiskOption, ecstore_new_disk};
@@ -106,7 +106,7 @@ const EVENT_HEAL_ERASURE_SET_STAGE: &str = "heal_erasure_set_stage";
 const EVENT_HEAL_ERASURE_SET_RESULT: &str = "heal_erasure_set_result";
 /// Heal type
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, PartialEq, Eq)]
 pub enum HealType {
    /// Cluster heal
    Cluster,
@@ -209,7 +209,7 @@ impl HealPriority {
 }
 /// Heal options
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
 pub struct HealOptions {
    /// Scan mode
    pub scan_mode: HealScanMode,
@@ -574,6 +574,18 @@ impl HealTask {
        }
    }
    pub(super) async fn outcome_bucket_incarnation_id(&self, bucket: &str, dry_run: bool) -> Result<Option<Uuid>> {
        if dry_run {
            return Ok(None);
        }
        match self.await_with_control(self.storage.bucket_incarnation_id(bucket)).await {
            Ok(incarnation_id) => Ok(incarnation_id),
            Err(Error::TaskCancelled) => Err(Error::TaskCancelled),
            Err(Error::TaskTimeout) => Err(Error::TaskTimeout),
            Err(_) => Ok(None),
        }
    }
    fn single_object_identity(&self) -> Option<HealObjectIdentity> {
        let (bucket, object, version) = match &self.heal_type {
            HealType::Object {
@@ -597,6 +609,9 @@ impl HealTask {
        expected: HealObjectIdentity,
        receipt: Option<HealObjectReceipt>,
    ) -> bool {
        if self.options.dry_run || self.cancel_token.is_cancelled() {
            return false;
        }
        let Some(receipt) = receipt else {
            return false;
        };
@@ -328,6 +328,23 @@ impl HealTask {
            }
        }
        if !self.options.dry_run {
            let metadata_opts = HealOpts {
                dry_run: self.options.dry_run,
                recreate: self.options.recreate_missing,
                scan_mode: self.options.scan_mode,
                pool: self.options.pool_index,
                set: self.options.set_index,
                ..Default::default()
            };
            for result in self
                .await_with_control(self.storage.heal_pool_metadata(&metadata_opts))
                .await?
            {
                self.record_result_item(result).await;
            }
        }
        if failed > 0 {
            let failure = BatchHealFailure {
                scope: "cluster".to_string(),
@@ -403,6 +420,7 @@ impl HealTask {
        };
        for (set_disk_id, heal_opts) in listing_scopes {
            let bucket_incarnation_id = self.outcome_bucket_incarnation_id(bucket, heal_opts.dry_run).await?;
            let mut continuation_token: Option<String> = None;
            let mut deferred = DeferredWindow::default();
            let mut inline_retry: Option<DeferredObject> = None;
@@ -501,13 +519,15 @@ impl HealTask {
                    let retry_attempt = item.attempt;
                    let mut telemetry_unknown = false;
                    let object = item.name.as_str();
-                    let identity =
+                    let mut identity =
                        self.outcome_identity(bucket, object, item.version_id.as_deref(), heal_opts.pool, heal_opts.set);
                    identity.bucket_incarnation_id = bucket_incarnation_id;
                    let mut disposition = if heal_opts.dry_run {
                        HealObjectDisposition::DryRunObserved
                    } else {
                        HealObjectDisposition::Unknown
                    };
                    let mut recorded_authoritative_outcome = false;
                    let mut detail = None;
                    {
                        let mut progress = self.progress.write().await;
@@ -520,23 +540,37 @@ impl HealTask {
                        Some(Error::other("heal object retry age exhausted"))
                    } else {
                        match self
-                            .await_with_control(
+                            .await_with_control(self.storage.heal_object_with_receipt(
-                                self.storage
+                                bucket,
-                                    .heal_object(bucket, object, item.version_id.as_deref(), &heal_opts),
+                                object,
-                            )
+                                item.version_id.as_deref(),
                                &heal_opts,
                            ))
                            .await
                        {
-                            Ok((result, None)) => match unavailable_recreate_error(&result, &heal_opts) {
+                            Ok(storage_result) if storage_result.error.is_none() => {
-                                Some(error) => Some(error),
+                                match unavailable_recreate_error(&storage_result.item, &heal_opts) {
-                                None => {
+                                    Some(error) => Some(error),
-                                    telemetry_unknown |= !increment_counter(&mut healed);
+                                    None => {
-                                    telemetry_unknown |=
+                                        telemetry_unknown |= !increment_counter(&mut healed);
-                                        !add_bytes(&mut bytes, u64::try_from(result.object_size).unwrap_or(u64::MAX));
+                                        telemetry_unknown |= !add_bytes(
-                                    self.record_result_item(result).await;
+                                            &mut bytes,
-                                    None
+                                            u64::try_from(storage_result.item.object_size).unwrap_or(u64::MAX),
                                        );
                                        recorded_authoritative_outcome = self
                                            .record_verified_storage_receipt(identity.clone(), storage_result.receipt)
                                            .await;
                                        self.record_result_item(storage_result.item).await;
                                        None
                                    }
                                }
-                            },
+                            }
-                            Ok((_, Some(err))) if is_missing_object_dir_heal_result(object, &err) => {
+                            Ok(storage_result)
                                if storage_result
                                    .error
                                    .as_ref()
                                    .is_some_and(|err| is_missing_object_dir_heal_result(object, err)) =>
                            {
                                telemetry_unknown |= !increment_counter(&mut healed);
                                debug!(
                                    target: "rustfs::heal::task",
@@ -551,7 +585,8 @@ impl HealTask {
                                );
                                None
                            }
-                            Ok((_, Some(err))) | Err(err) => Some(err),
+                            Ok(storage_result) => storage_result.error,
                            Err(err) => Some(err),
                        }
                    };
@@ -674,11 +709,13 @@ impl HealTask {
                        continue;
                    }
-                    self.outcome.write().await.record(HealObjectOutcome {
+                    if !recorded_authoritative_outcome {
-                        identity,
+                        self.outcome.write().await.record(HealObjectOutcome {
-                        disposition,
+                            identity,
-                        detail,
+                            disposition,
-                    });
+                            detail,
                        });
                    }
                    let mut progress = self.progress.write().await;
                    progress.update_object_progress(
@@ -451,6 +451,11 @@ impl HealTask {
            self.source,
        )
        .with_replacement_targets(replacement_targets, is_auto_replacement.then(|| self.id.clone()))
        .with_pool_metadata_targets(if self.options.recreate_missing && !self.options.dry_run {
            self.heal_endpoints.clone()
        } else {
            Vec::new()
        })
        .with_replacement_identity_fence(replacement_target_identities.clone())
        .with_mainline_pacer(self.mainline_pacer.clone());
@@ -162,6 +162,9 @@ impl HealTask {
            pool: self.options.pool_index,
            set: self.options.set_index,
        };
        let mut expected_identity =
            self.outcome_identity(bucket, object, version_id, self.options.pool_index, self.options.set_index);
        expected_identity.bucket_incarnation_id = self.outcome_bucket_incarnation_id(bucket, self.options.dry_run).await?;
        let heal_fut = self.storage.heal_object_with_receipt(bucket, object, version_id, &heal_opts);
        let heal_result = if self.source == HealRequestSource::ReadRepair {
@@ -266,8 +269,6 @@ impl HealTask {
                    let mut progress = self.progress.write().await;
                    progress.update_object_progress(1, 1, 0, 0, object_size);
                }
                let expected_identity =
                    self.outcome_identity(bucket, object, version_id, self.options.pool_index, self.options.set_index);
                self.record_verified_storage_receipt(expected_identity, storage_result.receipt)
                    .await;
                self.record_result_item(result).await;
@@ -14,6 +14,7 @@
 use super::super::{DiskOption, DiskStore, Endpoint, new_disk};
 use super::*;
 use crate::heal::POOL_META_NAME;
 use crate::heal::storage::HealStorageObjectResult;
 mod deferred_retry;
@@ -42,6 +43,7 @@ mod canonical_outcome {
    #[tokio::test(start_paused = true)]
    async fn cluster_retries_only_the_failed_listing_page() {
        let storage = Arc::new(MockStorage {
            pool_metadata_required: true,
            recoverable_second_page_failures: Mutex::new(Some(1)),
            ..Default::default()
        });
@@ -66,7 +68,7 @@ mod canonical_outcome {
        assert_eq!(task.get_progress().await.objects_scanned, 2);
        assert_eq!(
            storage.heal_object_calls.lock().expect("object calls").as_slice(),
-            ["object-a", "object-b"]
+            ["object-a", "object-b", POOL_META_NAME]
        );
        assert_eq!(
            storage.listing_tokens.lock().expect("listing tokens").as_slice(),
@@ -154,6 +156,126 @@ mod canonical_outcome {
        );
    }
    #[tokio::test]
    async fn bucket_heal_records_matching_positive_storage_receipts() {
        let incarnation = Uuid::new_v4();
        let storage = Arc::new(MockStorage {
            heal_object_receipts: Mutex::new(HashMap::from([
                (
                    "object-a".to_string(),
                    VecDeque::from([object_receipt("object-a", None, HealObjectDisposition::Repaired, incarnation)]),
                ),
                (
                    "object-b".to_string(),
                    VecDeque::from([object_receipt("object-b", None, HealObjectDisposition::Repaired, incarnation)]),
                ),
            ])),
            bucket_incarnation_id: Mutex::new(Some(incarnation)),
            ..Default::default()
        });
        let task = bucket_task(storage);
        task.execute()
            .await
            .expect("bucket heal should record verified object receipts");
        let outcome = task.get_outcome().await;
        assert_eq!(outcome.execution, HealExecutionOutcome::Completed);
        assert_eq!(outcome.counters.healed, 2);
        assert_eq!(outcome.counters.unknown, 0);
        assert_eq!(outcome.objects.len(), 2);
        assert!(outcome.objects.iter().all(|item| {
            item.identity.bucket_incarnation_id == Some(incarnation) && item.disposition == HealObjectDisposition::Repaired
        }));
    }
    #[tokio::test]
    async fn bucket_heal_keeps_repairing_when_bucket_incarnation_is_unavailable() {
        let storage = Arc::new(MockStorage {
            heal_object_receipts: Mutex::new(HashMap::from([(
                "object-a".to_string(),
                VecDeque::from([object_receipt(
                    "object-a",
                    None,
                    HealObjectDisposition::Repaired,
                    Uuid::new_v4(),
                )]),
            )])),
            bucket_incarnation_unavailable: Mutex::new(true),
            ..Default::default()
        });
        let task = bucket_task(storage.clone());
        task.execute()
            .await
            .expect("bucket heal should continue when only proof ownership is unavailable");
        let outcome = task.get_outcome().await;
        assert_eq!(outcome.execution, HealExecutionOutcome::Completed);
        assert_eq!(outcome.counters.healed, 0);
        assert_eq!(outcome.counters.unknown, 2);
        assert!(
            outcome
                .objects
                .iter()
                .all(|item| item.disposition == HealObjectDisposition::Unknown)
        );
        assert_eq!(storage.healed_objects.lock().expect("healed objects").len(), 2);
    }
    #[tokio::test]
    async fn bucket_heal_rejects_stale_receipts_without_double_recording() {
        let expected_incarnation = Uuid::new_v4();
        let storage = Arc::new(MockStorage {
            heal_object_receipts: Mutex::new(HashMap::from([
                (
                    "object-a".to_string(),
                    VecDeque::from([object_receipt(
                        "object-a",
                        None,
                        HealObjectDisposition::Repaired,
                        Uuid::new_v4(),
                    )]),
                ),
                (
                    "object-b".to_string(),
                    VecDeque::from([object_receipt(
                        "object-b",
                        None,
                        HealObjectDisposition::Repaired,
                        expected_incarnation,
                    )]),
                ),
            ])),
            bucket_incarnation_id: Mutex::new(Some(expected_incarnation)),
            ..Default::default()
        });
        let task = bucket_task(storage);
        task.execute()
            .await
            .expect("stale bucket receipt should not fail the legacy heal");
        let outcome = task.get_outcome().await;
        assert_eq!(outcome.execution, HealExecutionOutcome::Completed);
        assert_eq!(outcome.counters.healed, 1);
        assert_eq!(outcome.counters.unknown, 1);
        assert_eq!(outcome.objects.len(), 2);
        let object_a = outcome
            .objects
            .iter()
            .find(|item| item.identity.object == "object-a")
            .expect("stale receipt object outcome");
        assert_eq!(object_a.disposition, HealObjectDisposition::Unknown);
        let object_b = outcome
            .objects
            .iter()
            .find(|item| item.identity.object == "object-b")
            .expect("matching receipt object outcome");
        assert_eq!(object_b.disposition, HealObjectDisposition::Repaired);
        assert_eq!(object_b.identity.bucket_incarnation_id, Some(expected_incarnation));
    }
    #[tokio::test(start_paused = true)]
    async fn exhausted_object_does_not_abort_other_objects_or_erase_counts() {
        let storage = Arc::new(MockStorage::default());
@@ -313,6 +435,96 @@ mod canonical_outcome {
        );
    }
    #[tokio::test]
    async fn mixed_grace_and_repaired_receipt_transfer_only_repaired_responsibility() {
        let incarnation = Uuid::new_v4();
        let storage = Arc::new(MockStorage {
            heal_object_outcomes: Mutex::new(HashMap::from([(
                "object-a".to_string(),
                VecDeque::from([MockHealObjectOutcome::DanglingGraceDeferred]),
            )])),
            heal_object_receipts: Mutex::new(HashMap::from([(
                "object-b".to_string(),
                VecDeque::from([object_receipt("object-b", None, HealObjectDisposition::Repaired, incarnation)]),
            )])),
            bucket_incarnation_id: Mutex::new(Some(incarnation)),
            ..Default::default()
        });
        let task = bucket_task(storage);
        task.execute()
            .await
            .expect("mixed grace and repaired receipt should complete");
        let outcome = task.get_outcome().await;
        assert_eq!(outcome.coverage, HealTraversalCoverage::Complete);
        assert_eq!(outcome.counters.processed, 2);
        assert_eq!(outcome.counters.healed, 1);
        assert_eq!(outcome.counters.skipped, 1);
        let deferred = outcome
            .objects
            .iter()
            .find(|item| item.identity.object == "object-a")
            .expect("grace object should remain recorded");
        assert!(matches!(
            deferred.disposition,
            HealObjectDisposition::Deferred {
                reason: HealDeferredReason::DanglingDeleteGrace,
                ..
            }
        ));
        assert_ne!(deferred.disposition, HealObjectDisposition::Repaired);
        let repaired = outcome
            .objects
            .iter()
            .find(|item| item.identity.object == "object-b")
            .expect("receipt-backed object should be recorded");
        assert_eq!(repaired.identity.bucket_incarnation_id, Some(incarnation));
        assert_eq!(repaired.disposition, HealObjectDisposition::Repaired);
    }
    #[tokio::test]
    async fn bucket_heal_records_matching_positive_storage_receipt() {
        let incarnation = Uuid::new_v4();
        let storage = Arc::new(MockStorage {
            heal_object_receipts: Mutex::new(HashMap::from([(
                "object-a".to_string(),
                VecDeque::from([object_receipt("object-a", None, HealObjectDisposition::Repaired, incarnation)]),
            )])),
            bucket_incarnation_id: Mutex::new(Some(incarnation)),
            ..Default::default()
        });
        let task = bucket_task(storage);
        task.execute().await.expect("bucket heal should complete");
        let outcome = task.get_outcome().await;
        assert_eq!(outcome.counters.processed, 2);
        assert_eq!(outcome.counters.healed, 1);
        assert_eq!(outcome.counters.unknown, 1);
        assert_eq!(
            outcome
                .objects
                .iter()
                .filter(|item| item.identity.object == "object-a")
                .count(),
            1
        );
        let repaired = outcome
            .objects
            .iter()
            .find(|item| item.identity.object == "object-a")
            .expect("receipt-backed bucket object should be recorded");
        assert_eq!(repaired.identity.bucket_incarnation_id, Some(incarnation));
        assert_eq!(repaired.disposition, HealObjectDisposition::Repaired);
        let legacy = outcome
            .objects
            .iter()
            .find(|item| item.identity.object == "object-b")
            .expect("legacy bucket object should still be recorded");
        assert_eq!(legacy.disposition, HealObjectDisposition::Unknown);
    }
    #[tokio::test]
    async fn grace_single_object_is_completed_but_deferred() {
        let storage = Arc::new(MockStorage {
@@ -344,6 +556,18 @@ mod canonical_outcome {
                    .lock()
                    .expect("existence fixture")
                    .insert("object".to_string(), MockObjectExists::TransientSkip("retry later"));
            } else {
                let incarnation = Uuid::new_v4();
                *storage.bucket_incarnation_id.lock().expect("bucket incarnation") = Some(incarnation);
                storage.heal_object_receipts.lock().expect("receipt fixture").insert(
                    "object".to_string(),
                    VecDeque::from([object_receipt(
                        "object",
                        None,
                        HealObjectDisposition::VerifiedHealthy,
                        incarnation,
                    )]),
                );
            }
            let mut request = HealRequest::object("bucket-a".to_string(), "object".to_string(), None);
            request.options.dry_run = !transient;
@@ -1103,6 +1327,7 @@ struct MockStorage {
    retry_test_events: Mutex<Vec<String>>,
    listed: Mutex<bool>,
    list_each_bucket: bool,
    pool_metadata_required: bool,
    fail_second_listing_page: bool,
    recoverable_second_page_failures: Mutex<Option<usize>>,
    listing_tokens: Mutex<Vec<Option<String>>>,
@@ -1116,6 +1341,13 @@ struct MockStorage {
    heal_object_outcome: Mutex<Option<MockHealObjectOutcome>>,
    heal_object_outcomes: Mutex<HashMap<String, VecDeque<MockHealObjectOutcome>>>,
    heal_object_receipts: Mutex<HashMap<String, VecDeque<HealObjectReceipt>>>,
    bucket_incarnation_id: Mutex<Option<Uuid>>,
    bucket_incarnation_calls: AtomicU64,
    bucket_incarnation_error: Mutex<Option<Error>>,
    block_bucket_incarnation: bool,
    bucket_incarnation_started: tokio::sync::Notify,
    bucket_incarnation_after_object_heal: Mutex<Option<Uuid>>,
    bucket_incarnation_unavailable: Mutex<bool>,
    format_no_heal_required: Mutex<bool>,
    format_error: Mutex<Option<Error>>,
    global_format_calls: Mutex<u32>,
@@ -1219,14 +1451,19 @@ async fn execute_emits_heal_trace_task_state() {
    assert_eq!(trace_attr_string(&completed, "state").as_deref(), Some("completed"));
 }
-fn object_receipt(object: &str, version_id: Option<&str>, disposition: HealObjectDisposition) -> HealObjectReceipt {
+fn object_receipt(
    object: &str,
    version_id: Option<&str>,
    disposition: HealObjectDisposition,
    bucket_incarnation_id: Uuid,
 ) -> HealObjectReceipt {
    HealObjectReceipt {
        identity: HealObjectIdentity {
            kind: HealObjectKind::Object,
            bucket: "bucket-a".to_string(),
            object: object.to_string(),
            version_id: version_id.map(ToOwned::to_owned),
-            bucket_incarnation_id: Some(Uuid::new_v4()),
+            bucket_incarnation_id: Some(bucket_incarnation_id),
            pool_index: None,
            set_index: None,
        },
@@ -1236,16 +1473,23 @@ fn object_receipt(object: &str, version_id: Option<&str>, disposition: HealObjec
 #[tokio::test]
 async fn object_heal_records_matching_positive_storage_receipt() {
    let incarnation = Uuid::new_v4();
    let storage = Arc::new(MockStorage {
        heal_object_receipts: Mutex::new(HashMap::from([(
            "object-a".to_string(),
-            VecDeque::from([object_receipt("object-a", Some("version-a"), HealObjectDisposition::Repaired)]),
+            VecDeque::from([object_receipt(
                "object-a",
                Some("version-a"),
                HealObjectDisposition::Repaired,
                incarnation,
            )]),
        )])),
        bucket_incarnation_id: Mutex::new(Some(incarnation)),
        ..Default::default()
    });
    let task = HealTask::from_request(
        HealRequest::object("bucket-a".to_string(), "object-a".to_string(), Some("version-a".to_string())),
-        storage,
+        storage.clone(),
    );
    task.execute().await.expect("mock object heal should complete");
@@ -1258,19 +1502,276 @@ async fn object_heal_records_matching_positive_storage_receipt() {
    assert_eq!(object.identity.version_id.as_deref(), Some("version-a"));
    assert!(object.identity.bucket_incarnation_id.is_some());
    assert_eq!(object.disposition, HealObjectDisposition::Repaired);
    assert_eq!(
        storage.bucket_incarnation_calls.load(Ordering::Relaxed),
        1,
        "latch the owner exactly once before repair"
    );
 }
 #[tokio::test]
-async fn object_heal_rejects_mismatched_or_legacy_storage_receipts() {
+async fn object_heal_owner_lookup_failure_preserves_unverified_repair() {
    let storage = Arc::new(MockStorage {
        bucket_incarnation_error: Mutex::new(Some(Error::other("owner metadata unavailable"))),
        heal_object_receipts: Mutex::new(HashMap::from([(
            "object-a".to_string(),
            VecDeque::from([object_receipt(
                "object-a",
                None,
                HealObjectDisposition::Repaired,
                Uuid::new_v4(),
            )]),
        )])),
        ..Default::default()
    });
    let task = HealTask::from_request(HealRequest::object("bucket-a".to_string(), "object-a".to_string(), None), storage.clone());
    task.execute().await.expect("missing receipt owner must not prevent repair");
    assert_eq!(storage.heal_object_calls.lock().expect("heal calls").as_slice(), ["object-a"]);
    let outcome = task.get_outcome().await;
    assert_eq!(outcome.counters.healed, 0);
    assert_eq!(outcome.counters.unknown, 1);
    assert_eq!(
        outcome.objects.front().expect("unverified outcome").disposition,
        HealObjectDisposition::Unknown
    );
 }
 #[tokio::test]
 async fn object_heal_dry_run_skips_owner_lookup_and_positive_receipts() {
    let storage = Arc::new(MockStorage {
        bucket_incarnation_error: Mutex::new(Some(Error::other("dry-run must not query the receipt owner"))),
        heal_object_receipts: Mutex::new(HashMap::from([(
            "object-a".to_string(),
            VecDeque::from([object_receipt(
                "object-a",
                None,
                HealObjectDisposition::Repaired,
                Uuid::new_v4(),
            )]),
        )])),
        ..Default::default()
    });
    let mut request = HealRequest::object("bucket-a".to_string(), "object-a".to_string(), None);
    request.options.dry_run = true;
    let task = HealTask::from_request(request, storage.clone());
    task.execute().await.expect("dry-run should complete without owner metadata");
    assert!(storage.object_heal_opts.lock().expect("heal options")[0].dry_run);
    assert_eq!(storage.bucket_incarnation_calls.load(Ordering::Relaxed), 0);
    let outcome = task.get_outcome().await;
    assert_eq!(outcome.counters.healed, 0);
    assert_eq!(outcome.counters.unknown, 0);
    assert_eq!(outcome.counters.skipped, 1);
    assert_eq!(
        outcome.objects.front().expect("dry-run outcome").disposition,
        HealObjectDisposition::DryRunObserved
    );
 }
 #[tokio::test(start_paused = true)]
 async fn object_heal_owner_lookup_obeys_task_timeout() {
    let storage = Arc::new(MockStorage {
        block_bucket_incarnation: true,
        ..Default::default()
    });
    let mut request = HealRequest::object("bucket-a".to_string(), "object-a".to_string(), None);
    request.options.timeout = Some(Duration::from_secs(5));
    let task = HealTask::from_request(request, storage.clone());
    let result = tokio::time::timeout(Duration::from_secs(60), task.execute())
        .await
        .expect("owner lookup must honor the task deadline");
    assert!(matches!(result, Err(Error::TaskTimeout)));
    assert!(storage.heal_object_calls.lock().expect("heal calls").is_empty());
 }
 #[tokio::test]
 async fn object_heal_owner_lookup_obeys_cancellation() {
    let storage = Arc::new(MockStorage {
        block_bucket_incarnation: true,
        ..Default::default()
    });
    let mut request = HealRequest::object("bucket-a".to_string(), "object-a".to_string(), None);
    request.options.timeout = None;
    let task = HealTask::from_request(request, storage.clone());
    let (result, ()) = tokio::time::timeout(Duration::from_secs(5), async {
        tokio::join!(task.execute(), async {
            storage.bucket_incarnation_started.notified().await;
            task.cancel().await.expect("cancel pending owner lookup");
        })
    })
    .await
    .expect("cancellation must interrupt owner lookup");
    assert!(matches!(result, Err(Error::TaskCancelled)));
    assert!(storage.heal_object_calls.lock().expect("heal calls").is_empty());
 }
 #[tokio::test]
 async fn cancelled_object_heal_rejects_matching_positive_storage_receipt() {
    let incarnation = Uuid::new_v4();
    let storage = Arc::new(MockStorage::default());
    let task = HealTask::from_request(
        HealRequest::object("bucket-a".to_string(), "object-a".to_string(), Some("version-a".to_string())),
        storage,
    );
    task.cancel().await.expect("task cancellation should succeed");
    let expected = HealObjectIdentity {
        kind: HealObjectKind::Object,
        bucket: "bucket-a".to_string(),
        object: "object-a".to_string(),
        version_id: Some("version-a".to_string()),
        bucket_incarnation_id: Some(incarnation),
        pool_index: None,
        set_index: None,
    };
    let accepted = task
        .record_verified_storage_receipt(
            expected,
            Some(object_receipt(
                "object-a",
                Some("version-a"),
                HealObjectDisposition::Repaired,
                incarnation,
            )),
        )
        .await;
    let outcome = task.get_outcome().await;
    assert!(!accepted);
    assert_eq!(outcome.counters.healed, 0);
    assert!(outcome.objects.is_empty());
 }
 #[tokio::test]
 async fn failed_object_heal_rejects_matching_positive_storage_receipt() {
    let incarnation = Uuid::new_v4();
    let storage = Arc::new(MockStorage {
        heal_object_outcome: Mutex::new(Some(MockHealObjectOutcome::OkWithOtherError("commit failed"))),
        heal_object_receipts: Mutex::new(HashMap::from([(
            "object-a".to_string(),
            VecDeque::from([object_receipt(
                "object-a",
                Some("version-a"),
                HealObjectDisposition::Repaired,
                incarnation,
            )]),
        )])),
        bucket_incarnation_id: Mutex::new(Some(incarnation)),
        ..Default::default()
    });
    let task = HealTask::from_request(
        HealRequest::object("bucket-a".to_string(), "object-a".to_string(), Some("version-a".to_string())),
        storage,
    );
    let result = task.execute().await;
    let outcome = task.get_outcome().await;
    assert!(result.is_err());
    assert_eq!(outcome.counters.healed, 0);
    assert_eq!(outcome.counters.unchanged, 0);
    assert!(outcome.objects.iter().all(|object| {
        !matches!(
            object.disposition,
            HealObjectDisposition::Repaired
                | HealObjectDisposition::VerifiedHealthy
                | HealObjectDisposition::AuthoritativelyAbsent
        )
    }));
 }
 #[tokio::test]
 async fn transient_quorum_object_heal_rejects_matching_positive_storage_receipt() {
    let incarnation = Uuid::new_v4();
    let storage = Arc::new(MockStorage {
        heal_object_outcome: Mutex::new(Some(MockHealObjectOutcome::OkWithReadQuorum)),
        heal_object_receipts: Mutex::new(HashMap::from([(
            "object-a".to_string(),
            VecDeque::from([object_receipt(
                "object-a",
                Some("version-a"),
                HealObjectDisposition::Repaired,
                incarnation,
            )]),
        )])),
        bucket_incarnation_id: Mutex::new(Some(incarnation)),
        ..Default::default()
    });
    let task = HealTask::from_request(
        HealRequest::object("bucket-a".to_string(), "object-a".to_string(), Some("version-a".to_string())),
        storage,
    );
    let result = task.execute().await;
    let outcome = task.get_outcome().await;
    assert!(result.is_err());
    assert_eq!(outcome.counters.healed, 0);
    assert_eq!(outcome.counters.unchanged, 0);
    assert!(outcome.objects.iter().all(|object| {
        !matches!(
            object.disposition,
            HealObjectDisposition::Repaired
                | HealObjectDisposition::VerifiedHealthy
                | HealObjectDisposition::AuthoritativelyAbsent
        )
    }));
 }
 #[tokio::test]
 async fn object_heal_latches_expected_incarnation_before_repair() {
    let original_incarnation = Uuid::new_v4();
    let successor_incarnation = Uuid::new_v4();
    let storage = Arc::new(MockStorage {
        heal_object_receipts: Mutex::new(HashMap::from([(
            "object-a".to_string(),
            VecDeque::from([object_receipt(
                "object-a",
-                Some("old-version"),
+                Some("version-a"),
-                HealObjectDisposition::Repaired,
+                HealObjectDisposition::VerifiedHealthy,
                original_incarnation,
            )]),
        )])),
        bucket_incarnation_id: Mutex::new(Some(original_incarnation)),
        bucket_incarnation_after_object_heal: Mutex::new(Some(successor_incarnation)),
        ..Default::default()
    });
    let task = HealTask::from_request(
        HealRequest::object("bucket-a".to_string(), "object-a".to_string(), Some("version-a".to_string())),
        storage,
    );
    task.execute().await.expect("mock object heal should complete");
    let outcome = task.get_outcome().await;
    assert_eq!(outcome.counters.unchanged, 1);
    assert_eq!(outcome.counters.unknown, 0);
    let object = outcome.objects.front().expect("latched receipt should be recorded");
    assert_eq!(object.identity.bucket_incarnation_id, Some(original_incarnation));
    assert_eq!(object.disposition, HealObjectDisposition::VerifiedHealthy);
 }
 #[tokio::test]
 async fn object_heal_rejects_mismatched_or_legacy_storage_receipts() {
    let expected_incarnation = Uuid::new_v4();
    let storage = Arc::new(MockStorage {
        heal_object_receipts: Mutex::new(HashMap::from([(
            "object-a".to_string(),
            VecDeque::from([object_receipt(
                "object-a",
                Some("version-a"),
                HealObjectDisposition::Repaired,
                Uuid::new_v4(),
            )]),
        )])),
        bucket_incarnation_id: Mutex::new(Some(expected_incarnation)),
        ..Default::default()
    });
    let task = HealTask::from_request(
@@ -1361,6 +1862,7 @@ enum MockHealObjectOutcome {
    RetryableLock,
    RetryableLockTimeout,
    OkWithOtherError(&'static str),
    OkWithReadQuorum,
    ErrOther(&'static str),
    DanglingGraceDeferred,
    UnavailableDrive(DriveState),
@@ -1445,6 +1947,34 @@ impl HealStorageAPI for MockStorage {
            .collect())
    }
    async fn heal_pool_metadata(&self, opts: &HealOpts) -> Result<Vec<HealResultItem>> {
        if !self.pool_metadata_required {
            return Ok(Vec::new());
        }
        let scopes = self.erasure_set_scopes.lock().expect("metadata scopes").clone();
        let scopes = if scopes.is_empty() {
            vec![(opts.pool.unwrap_or(0), opts.set.unwrap_or(0))]
        } else {
            scopes
        };
        let mut results = Vec::new();
        for (pool, set) in scopes {
            let scoped_opts = HealOpts {
                pool: Some(pool),
                set: Some(set),
                ..*opts
            };
            let (result, error) = self
                .heal_object(RUSTFS_META_BUCKET, crate::heal::POOL_META_NAME, None, &scoped_opts)
                .await?;
            if let Some(error) = error {
                return Err(error);
            }
            results.push(result);
        }
        Ok(results)
    }
    async fn object_exists(&self, _bucket: &str, object: &str) -> Result<bool> {
        if let Some(result) = self.object_exists_by_name.lock().unwrap().get(object).copied() {
            return match result {
@@ -1456,6 +1986,21 @@ impl HealStorageAPI for MockStorage {
        Ok(self.object_exists.lock().unwrap().unwrap_or(true))
    }
    async fn bucket_incarnation_id(&self, _bucket: &str) -> Result<Option<Uuid>> {
        self.bucket_incarnation_calls.fetch_add(1, Ordering::Relaxed);
        self.bucket_incarnation_started.notify_one();
        if self.block_bucket_incarnation {
            std::future::pending::<()>().await;
        }
        if let Some(error) = self.bucket_incarnation_error.lock().expect("owner lookup error").take() {
            return Err(error);
        }
        if *self.bucket_incarnation_unavailable.lock().unwrap() {
            return Err(Error::Other("bucket incarnation unavailable".to_string()));
        }
        Ok(*self.bucket_incarnation_id.lock().unwrap())
    }
    async fn heal_object(
        &self,
        bucket: &str,
@@ -1513,6 +2058,13 @@ impl HealStorageAPI for MockStorage {
                MockHealObjectOutcome::RetryableSlowDown => {
                    Ok((HealResultItem::default(), Some(Error::Storage(EcstoreError::SlowDown))))
                }
                MockHealObjectOutcome::OkWithReadQuorum => Ok((
                    HealResultItem::default(),
                    Some(Error::Storage(EcstoreError::InsufficientReadQuorum(
                        bucket.to_string(),
                        object.to_string(),
                    ))),
                )),
                MockHealObjectOutcome::PermanentOther(message) => Err(Error::other(message)),
                MockHealObjectOutcome::OkWithOtherError(message) => Ok((HealResultItem::default(), Some(Error::other(message)))),
                MockHealObjectOutcome::ErrOther(message) => Err(Error::other(message)),
@@ -1552,6 +2104,13 @@ impl HealStorageAPI for MockStorage {
                MockHealObjectOutcome::RetryableSlowDown => {
                    Ok((HealResultItem::default(), Some(Error::Storage(EcstoreError::SlowDown))))
                }
                MockHealObjectOutcome::OkWithReadQuorum => Ok((
                    HealResultItem::default(),
                    Some(Error::Storage(EcstoreError::InsufficientReadQuorum(
                        bucket.to_string(),
                        object.to_string(),
                    ))),
                )),
            };
        }
        if bucket == RUSTFS_META_BUCKET && object == format!("{BUCKET_META_PREFIX}/{DATA_USAGE_CACHE_NAME}") {
@@ -1566,6 +2125,9 @@ impl HealStorageAPI for MockStorage {
            return Ok((HealResultItem::default(), Some(Error::Disk(DiskError::FileNotFound))));
        }
        self.healed_objects.lock().unwrap().push(object.to_string());
        if let Some(bucket_incarnation_id) = self.bucket_incarnation_after_object_heal.lock().unwrap().take() {
            *self.bucket_incarnation_id.lock().unwrap() = Some(bucket_incarnation_id);
        }
        Ok((
            HealResultItem {
                object_size: 1,
@@ -2153,9 +2715,187 @@ async fn test_recursive_bucket_heal_treats_missing_continuation_token_as_end() {
    );
 }
 #[tokio::test]
 async fn root_heal_restores_pool_metadata_without_user_buckets() {
    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        listed_buckets: Mutex::new(Some(Vec::new())),
        ..Default::default()
    });
    assert!(storage.pool_metadata_required);
    let task = HealTask::from_request(
        HealRequest::new(HealType::Cluster, HealOptions::default(), HealPriority::Normal),
        storage.clone(),
    );
    task.execute().await.expect("root heal should restore required pool metadata");
    assert_eq!(
        storage.heal_object_calls.lock().expect("heal calls").as_slice(),
        [crate::heal::POOL_META_NAME]
    );
    assert!(matches!(task.get_status().await, HealTaskStatus::Completed));
 }
 #[tokio::test]
 async fn root_heal_pool_metadata_cannot_hide_a_later_owner_failure() {
    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        listed_buckets: Mutex::new(Some(Vec::new())),
        erasure_set_scopes: Mutex::new(vec![(0, 0), (1, 1)]),
        ..Default::default()
    });
    storage.heal_object_outcomes.lock().expect("metadata outcomes").insert(
        crate::heal::POOL_META_NAME.to_string(),
        VecDeque::from([
            MockHealObjectOutcome::UnavailableDrive(DriveState::Ok),
            MockHealObjectOutcome::OkWithReadQuorum,
        ]),
    );
    let task = HealTask::from_request(
        HealRequest::new(HealType::Cluster, HealOptions::default(), HealPriority::Normal),
        storage.clone(),
    );
    let error = task
        .execute()
        .await
        .expect_err("one healthy owner cannot satisfy another owner's recovery");
    assert!(matches!(error, Error::Storage(EcstoreError::InsufficientReadQuorum(_, _))));
    {
        let opts = storage.object_heal_opts.lock().expect("owner options");
        assert_eq!(
            opts.iter().map(|opts| (opts.pool, opts.set)).collect::<Vec<_>>(),
            vec![(Some(0), Some(0)), (Some(1), Some(1))]
        );
    }
    assert!(!matches!(task.get_status().await, HealTaskStatus::Completed));
 }
 #[tokio::test]
 async fn root_heal_pool_metadata_does_not_inherit_remove_or_no_lock() {
    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        listed_buckets: Mutex::new(Some(Vec::new())),
        ..Default::default()
    });
    let task = HealTask::from_request(
        HealRequest::new(
            HealType::Cluster,
            HealOptions {
                remove_corrupted: true,
                no_lock: true,
                ..Default::default()
            },
            HealPriority::Normal,
        ),
        storage.clone(),
    );
    task.execute()
        .await
        .expect("metadata repair should retain its write fence and reject destructive options");
    let opts = storage.object_heal_opts.lock().expect("metadata options");
    assert_eq!(opts.len(), 1, "an empty user namespace must still inspect metadata");
    assert!(!opts[0].dry_run);
    assert!(!opts[0].remove);
    assert!(!opts[0].no_lock);
 }
 #[tokio::test]
 async fn root_heal_pool_metadata_failure_does_not_prevent_user_repairs() {
    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        ..Default::default()
    });
    storage.heal_object_outcomes.lock().expect("metadata outcome").insert(
        crate::heal::POOL_META_NAME.to_string(),
        VecDeque::from([MockHealObjectOutcome::OkWithReadQuorum]),
    );
    let task = HealTask::from_request(
        HealRequest::new(
            HealType::Cluster,
            HealOptions {
                recursive: true,
                ..Default::default()
            },
            HealPriority::Normal,
        ),
        storage.clone(),
    );
    let error = task
        .execute()
        .await
        .expect_err("unrecovered metadata must still fail root completion");
    assert!(matches!(error, Error::Storage(EcstoreError::InsufficientReadQuorum(_, _))));
    assert_eq!(
        storage.heal_object_calls.lock().expect("heal calls").as_slice(),
        ["object-a", "object-b", crate::heal::POOL_META_NAME]
    );
    assert!(!matches!(task.get_status().await, HealTaskStatus::Completed));
 }
 #[tokio::test]
 async fn root_heal_pool_metadata_preserves_typed_quorum_failure() {
    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        listed_buckets: Mutex::new(Some(Vec::new())),
        ..Default::default()
    });
    storage.heal_object_outcomes.lock().expect("metadata outcome").insert(
        crate::heal::POOL_META_NAME.to_string(),
        VecDeque::from([MockHealObjectOutcome::OkWithReadQuorum]),
    );
    let task = HealTask::from_request(HealRequest::new(HealType::Cluster, HealOptions::default(), HealPriority::Normal), storage);
    let error = task
        .execute()
        .await
        .expect_err("metadata quorum failure must prevent root completion");
    assert!(matches!(error, Error::Storage(EcstoreError::InsufficientReadQuorum(_, _))));
    assert!(!matches!(task.get_status().await, HealTaskStatus::Completed));
 }
 #[tokio::test(start_paused = true)]
 async fn root_heal_pool_metadata_obeys_task_timeout() {
    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        listed_buckets: Mutex::new(Some(Vec::new())),
        retry_test_delays: HashMap::from([(crate::heal::POOL_META_NAME.to_string(), Duration::from_secs(10))]),
        ..Default::default()
    });
    let task = HealTask::from_request(
        HealRequest::new(
            HealType::Cluster,
            HealOptions {
                timeout: Some(Duration::from_millis(10)),
                ..Default::default()
            },
            HealPriority::Normal,
        ),
        storage,
    );
    let error = task
        .execute()
        .await
        .expect_err("metadata work must stay inside the root task budget");
    assert!(matches!(error, Error::TaskTimeout));
    assert!(!matches!(task.get_status().await, HealTaskStatus::Completed));
 }
 #[tokio::test]
 async fn test_cluster_heal_visits_bucket_objects() {
-    let storage = Arc::new(MockStorage::default());
+    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        ..Default::default()
    });
    let request = HealRequest::new(
        HealType::Cluster,
        HealOptions {
@@ -2171,11 +2911,104 @@ async fn test_cluster_heal_visits_bucket_objects() {
    assert_eq!(
        storage.healed_objects.lock().unwrap().as_slice(),
-        ["object-a".to_string(), "object-b".to_string()]
+        ["object-a".to_string(), "object-b".to_string(), POOL_META_NAME.to_string()]
    );
    assert!(matches!(task.get_status().await, HealTaskStatus::Completed));
 }
 #[tokio::test]
 async fn cluster_recreate_heals_pool_metadata_after_user_buckets() {
    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        ..Default::default()
    });
    let request = HealRequest::new(
        HealType::Cluster,
        HealOptions {
            recursive: true,
            recreate_missing: true,
            timeout: None,
            ..Default::default()
        },
        HealPriority::Normal,
    );
    let task = HealTask::from_request(request, storage.clone());
    task.execute()
        .await
        .expect("cluster recreate heal should include pool metadata");
    assert_eq!(
        storage.heal_object_calls.lock().expect("object calls").as_slice(),
        ["object-a".to_string(), "object-b".to_string(), POOL_META_NAME.to_string()]
    );
    let opts = storage.object_heal_opts.lock().expect("object opts");
    assert!(opts.last().expect("pool metadata opts").recreate);
 }
 #[tokio::test]
 async fn cluster_recreate_fails_when_pool_metadata_heal_fails() {
    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        ..Default::default()
    });
    storage.heal_object_outcomes.lock().expect("object outcomes").insert(
        POOL_META_NAME.to_string(),
        VecDeque::from([MockHealObjectOutcome::ErrOther("pool metadata missing")]),
    );
    let request = HealRequest::new(
        HealType::Cluster,
        HealOptions {
            recursive: true,
            recreate_missing: true,
            timeout: None,
            ..Default::default()
        },
        HealPriority::Normal,
    );
    let task = HealTask::from_request(request, storage.clone());
    let err = task
        .execute()
        .await
        .expect_err("cluster recreate heal must not hide pool metadata failure");
    assert!(matches!(err, Error::Other(message) if message == "pool metadata missing"));
    assert_eq!(
        storage.heal_object_calls.lock().expect("object calls").as_slice(),
        ["object-a".to_string(), "object-b".to_string(), POOL_META_NAME.to_string()]
    );
 }
 #[tokio::test]
 async fn cluster_dry_run_does_not_heal_pool_metadata() {
    let storage = Arc::new(MockStorage {
        pool_metadata_required: true,
        ..Default::default()
    });
    let request = HealRequest::new(
        HealType::Cluster,
        HealOptions {
            recursive: true,
            dry_run: true,
            recreate_missing: true,
            timeout: None,
            ..Default::default()
        },
        HealPriority::Normal,
    );
    let task = HealTask::from_request(request, storage.clone());
    task.execute()
        .await
        .expect("dry-run cluster heal should preserve existing coverage");
    assert_eq!(
        storage.heal_object_calls.lock().expect("object calls").as_slice(),
        ["object-a".to_string(), "object-b".to_string()]
    );
 }
 #[tokio::test]
 async fn object_heal_skips_dangling_delete_grace_without_failing_task() {
    let storage = Arc::new(MockStorage {
@@ -29,6 +29,7 @@ use rustfs_heal::heal::{
    storage::{ECStoreHealStorage, HealStorageAPI},
 };
 use serial_test::serial;
 use sha2::{Digest, Sha256};
 #[cfg(unix)]
 use std::{
    fs::{File, OpenOptions},
@@ -48,6 +49,12 @@ use storage_api::endpoint_index::{Endpoint, EndpointServerPools, Endpoints, Pool
 const META_BUCKET: &str = ".rustfs.sys";
 const JOURNAL_REL: &str = "buckets/.heal/mrf/journal.bin";
 const SCOPED_JOURNAL_REL: &str = "buckets/.heal/mrf/journal-scoped.bin";
 const COMMITTED_PAYLOAD_REL: &str = ".heal-mrf-snapshot.0.bin";
 const COMMITTED_MANIFEST_REL: &str = ".heal-mrf-commit.0.bin";
 const COMMITTED_PAYLOAD_RELS: [&str; 2] = [".heal-mrf-snapshot.0.bin", ".heal-mrf-snapshot.1.bin"];
 const COMMITTED_MANIFEST_RELS: [&str; 2] = [".heal-mrf-commit.0.bin", ".heal-mrf-commit.1.bin"];
 const COMMITTED_MAGIC: &[u8; 8] = b"RFMRFC01";
 const COMMITTED_MANIFEST_LEN: usize = 8 + 1 + 16 + 8 + 8 + 32 + 32;
 async fn heal_env() -> (Vec<std::path::PathBuf>, Arc<dyn HealStorageAPI>) {
    heal_env_at(None).await
@@ -195,6 +202,29 @@ fn write_journal_to_disks(disk_paths: &[std::path::PathBuf], data: &[u8]) {
    write_journal_path_to_disks(disk_paths, JOURNAL_REL, data);
 }
 fn committed_manifest(owner: uuid::Uuid, sequence: u64, payload: &[u8]) -> Vec<u8> {
    let mut manifest = Vec::with_capacity(COMMITTED_MANIFEST_LEN);
    manifest.extend_from_slice(COMMITTED_MAGIC);
    manifest.push(1);
    manifest.extend_from_slice(owner.as_bytes());
    manifest.extend_from_slice(&sequence.to_le_bytes());
    manifest.extend_from_slice(
        &u64::try_from(payload.len())
            .expect("fixture payload length fits")
            .to_le_bytes(),
    );
    manifest.extend_from_slice(&Sha256::digest(payload));
    manifest.extend_from_slice(&Sha256::digest(&manifest));
    assert_eq!(manifest.len(), COMMITTED_MANIFEST_LEN, "committed fixture manifest length");
    manifest
 }
 fn write_committed_snapshot_to_disks(disk_paths: &[std::path::PathBuf], sequence: u64, payload: &[u8]) {
    let manifest = committed_manifest(uuid::Uuid::new_v4(), sequence, payload);
    write_journal_path_to_disks(disk_paths, COMMITTED_PAYLOAD_REL, payload);
    write_journal_path_to_disks(disk_paths, COMMITTED_MANIFEST_REL, &manifest);
 }
 fn journal_exists_on_all_disks(disk_paths: &[std::path::PathBuf], relative_path: &str) -> bool {
    disk_paths
        .iter()
@@ -207,6 +237,44 @@ fn journal_matches_on_all_disks(disk_paths: &[PathBuf], relative_path: &str, exp
        .all(|path| std::fs::read(path.join(META_BUCKET).join(relative_path)).is_ok_and(|actual| actual == expected))
 }
 fn committed_checkpoint_matches_on_all_disks(disk_paths: &[PathBuf], sequence: u64, expected_payload: &[u8]) -> bool {
    disk_paths.iter().all(|path| {
        let root = path.join(META_BUCKET);
        COMMITTED_PAYLOAD_RELS
            .into_iter()
            .zip(COMMITTED_MANIFEST_RELS)
            .any(|(payload_rel, manifest_rel)| {
                let Ok(payload) = std::fs::read(root.join(payload_rel)) else {
                    return false;
                };
                if payload != expected_payload {
                    return false;
                }
                let Ok(manifest) = std::fs::read(root.join(manifest_rel)) else {
                    return false;
                };
                if manifest.len() != COMMITTED_MANIFEST_LEN || &manifest[..8] != COMMITTED_MAGIC || manifest[8] != 1 {
                    return false;
                }
                let Ok(recorded_sequence) = <[u8; 8]>::try_from(&manifest[25..33]).map(u64::from_le_bytes) else {
                    return false;
                };
                let Ok(recorded_len) = <[u8; 8]>::try_from(&manifest[33..41]).map(u64::from_le_bytes) else {
                    return false;
                };
                let Ok(expected_len) = u64::try_from(expected_payload.len()) else {
                    return false;
                };
                if recorded_sequence != sequence || recorded_len != expected_len {
                    return false;
                }
                let payload_digest: [u8; 32] = Sha256::digest(expected_payload).into();
                let manifest_digest: [u8; 32] = Sha256::digest(&manifest[..COMMITTED_MANIFEST_LEN - 32]).into();
                payload_digest.as_slice() == &manifest[41..73] && manifest_digest.as_slice() == &manifest[73..]
            })
    })
 }
 async fn wait_until<F, Fut>(deadline: Duration, mut probe: F) -> bool
 where
    F: FnMut() -> Fut,
@@ -255,11 +323,12 @@ async fn decode_failure_intent_maps_to_urgent_mrf_heal_request() {
 }
 /// A journal left behind by a previous process must be replayed into the
-/// manager queue and then removed, and a torn tail must not block replay of
+/// manager queue, and a torn tail must not block replay of the intact records.
-/// the intact records.
+/// The partial-write record keeps the legacy journal as the durable anchor
 /// until an exact verified repair proof can discharge it.
 #[tokio::test(flavor = "multi_thread", worker_threads = 4)]
 #[serial]
-async fn journal_replay_arms_intents_and_deletes_the_file() {
+async fn journal_replay_arms_intents_and_retains_unproven_partial_write_anchor() {
    let (disk_paths, storage) = heal_env().await;
    // The journal reader resolves disks through the process-local disk map;
@@ -285,14 +354,14 @@ async fn journal_replay_arms_intents_and_deletes_the_file() {
    assert!(
        disk_paths
            .iter()
-            .all(|path| !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()),
+            .all(|path| Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()),
-        "the journal file must be removed after a successful replay"
+        "partial-write replay must retain the legacy journal until durable proof"
    );
    assert!(
        disk_paths
            .iter()
            .all(|path| !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()),
-        "the authoritative journal file must also be removed after replay"
+        "missing authoritative journal remains absent"
    );
    let snapshot = manager.operations_snapshot().await;
@@ -300,6 +369,76 @@ async fn journal_replay_arms_intents_and_deletes_the_file() {
    assert!(snapshot.queued_by_priority.normal >= 1, "the partial-write record must replay as Normal");
 }
 /// A committed checkpoint published by the new two-slot writer is the
 /// authoritative startup snapshot. Legacy mirrors are fallback-only and must
 /// not be merged with or preferred over the committed epoch.
 #[tokio::test(flavor = "multi_thread", worker_threads = 4)]
 #[serial]
 async fn committed_snapshot_replay_takes_precedence_over_stale_legacy_mirror() {
    let (disk_paths, storage) = heal_env().await;
    register_local_disks(&disk_paths, "mrf-committed-replay-test").await;
    let committed = scoped_journal_record(3, "committed-bucket", "committed-object", Some([9u8; 16]), 0, 0, 0);
    let stale_legacy = journal_record(1, "legacy-bucket", "legacy-object", None, 0);
    write_committed_snapshot_to_disks(&disk_paths, 7, &committed);
    write_journal_path_to_disks(&disk_paths, SCOPED_JOURNAL_REL, &stale_legacy);
    write_journal_path_to_disks(&disk_paths, JOURNAL_REL, &stale_legacy);
    let manager = make_manager(storage);
    let replayed = mrf_queue::replay_journal_once(&manager).await;
    assert_eq!(replayed, 1, "only the committed snapshot epoch may replay");
    let snapshot = manager.operations_snapshot().await;
    assert_eq!(snapshot.queued_by_source.mrf, 1);
    assert_eq!(
        snapshot.queued_by_priority.normal, 1,
        "the committed partial-write record must replay instead of the stale legacy decode-failure"
    );
    assert_eq!(
        snapshot.queued_by_priority.urgent, 0,
        "stale legacy decode-failure records must not be mixed into committed replay"
    );
    assert!(
        journal_exists_on_all_disks(&disk_paths, COMMITTED_MANIFEST_REL),
        "the committed checkpoint remains until the accepted partial-write has proof"
    );
 }
 /// A damaged committed checkpoint is ambiguous: replay must not fall back to
 /// older legacy bytes or delete any recovery anchor until another process can
 /// publish a valid successor.
 #[tokio::test(flavor = "multi_thread", worker_threads = 4)]
 #[serial]
 async fn damaged_committed_snapshot_blocks_legacy_fallback_and_retains_anchors() {
    let (disk_paths, storage) = heal_env().await;
    register_local_disks(&disk_paths, "mrf-damaged-committed-replay-test").await;
    let committed = scoped_journal_record(3, "damaged-committed-bucket", "committed-object", Some([8u8; 16]), 0, 0, 0);
    let stale_legacy = journal_record(1, "damaged-legacy-bucket", "legacy-object", None, 0);
    write_journal_path_to_disks(&disk_paths, COMMITTED_PAYLOAD_REL, &committed);
    let mut manifest = committed_manifest(uuid::Uuid::new_v4(), 9, &committed);
    manifest[25] ^= 1;
    write_journal_path_to_disks(&disk_paths, COMMITTED_MANIFEST_REL, &manifest);
    write_journal_path_to_disks(&disk_paths, SCOPED_JOURNAL_REL, &stale_legacy);
    write_journal_path_to_disks(&disk_paths, JOURNAL_REL, &stale_legacy);
    let manager = make_manager(storage);
    let replayed = mrf_queue::replay_journal_once(&manager).await;
    assert_eq!(replayed, 0, "damaged committed state must fail closed");
    assert_eq!(
        manager.operations_snapshot().await.queued_by_source.mrf,
        0,
        "stale legacy bytes must not be replayed when committed state is ambiguous"
    );
    assert!(
        journal_exists_on_all_disks(&disk_paths, COMMITTED_MANIFEST_REL)
            && journal_exists_on_all_disks(&disk_paths, COMMITTED_PAYLOAD_REL)
            && journal_matches_on_all_disks(&disk_paths, SCOPED_JOURNAL_REL, &stale_legacy)
            && journal_matches_on_all_disks(&disk_paths, JOURNAL_REL, &stale_legacy),
        "all recovery anchors must remain after a fail-closed committed read"
    );
 }
 /// A canonical snapshot and its compatibility mirror may differ after a
 /// partial flush. Replay must choose the complete canonical epoch instead of
 /// combining records that never coexisted in memory.
@@ -322,21 +461,24 @@ async fn authoritative_journal_is_not_merged_with_legacy_mirror() {
    assert_eq!(snapshot.queued_by_source.mrf, 1);
    assert!(
        disk_paths.iter().all(|path| {
-            !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
+            Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
-                && !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
+                && Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
        }),
-        "replay cleanup must remove both journal paths"
+        "accepted replay responsibilities remain anchored until a verified repair proof"
    );
    // A scoped-only snapshot is valid during a rollout where no legacy
    // compatibility mirror was written. Missing legacy files must not leave
    // the runtime in a permanent cleanup-retry state.
    let (disk_paths, storage) = heal_env().await;
    register_local_disks(&disk_paths, "mrf-scoped-authoritative-test").await;
    let manager = make_manager(storage);
    let scoped_only = journal_record(1, "scoped-only-bucket", "scoped-only-object", None, 0);
    write_journal_path_to_disks(&disk_paths, SCOPED_JOURNAL_REL, &scoped_only);
    assert_eq!(mrf_queue::replay_journal_once(&manager).await, 1);
    assert!(disk_paths.iter().all(|path| {
        !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
-            && !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
+            && Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
    }));
    let scoped_v2 = scoped_journal_record(1, "scoped-v2-bucket", "scoped-v2-object", None, 0, 3, 7);
@@ -350,12 +492,12 @@ async fn authoritative_journal_is_not_merged_with_legacy_mirror() {
    );
    assert_eq!(
        manager.operations_snapshot().await.queued_by_source.mrf,
-        3,
+        2,
-        "only the three authoritative/scoped-only epochs should have reached the manager"
+        "only the scoped-only and scoped-v2 authoritative epochs should have reached the manager"
    );
    assert!(disk_paths.iter().all(|path| {
-        !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
+        Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
-            && !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
+            && Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
    }));
 }
@@ -400,10 +542,13 @@ async fn authoritative_journal_replay_preserves_kind_and_scope_identity() {
        snapshot.queued_by_priority.urgent, 1,
        "decode-failure repair must not merge with object repair responsibility"
    );
-    assert!(disk_paths.iter().all(|path| {
+    assert!(
-        !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
+        disk_paths.iter().all(|path| {
-            && !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
+            Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
-    }));
+                && Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
        }),
        "partial-write responsibilities keep both replay anchors until proof"
    );
 }
 /// If replay reaches a full heal-manager queue, the old journal remains the
@@ -521,16 +666,18 @@ fn mrf_successor_flush_child_process_fixture() {
            }),
        ));
        mrf_queue::spawn_mrf_consumer(manager.clone());
-        let expected_successor = journal_record(1, "successor-bucket", "second-object", None, 2);
+        let mut expected_successor = journal_record(1, "successor-bucket", "second-object", None, 2);
        expected_successor.extend(journal_record(1, "successor-bucket", "first-object", None, 0));
        let flushed = wait_until(Duration::from_secs(10), || async {
            manager.operations_snapshot().await.queued_by_source.mrf == 1
                && committed_checkpoint_matches_on_all_disks(&disk_paths, 2, &expected_successor)
                && journal_matches_on_all_disks(&disk_paths, SCOPED_JOURNAL_REL, &expected_successor)
                && journal_matches_on_all_disks(&disk_paths, JOURNAL_REL, &expected_successor)
        })
        .await;
        assert!(
            flushed,
-            "child process must publish the pending successor snapshot before the delete phase"
+            "child process must publish the committed pending successor before the delete phase"
        );
    });
    std::process::exit(78);
@@ -568,16 +715,18 @@ fn mrf_successor_flush_waiting_child_process_fixture() {
            }),
        ));
        mrf_queue::spawn_mrf_consumer(manager.clone());
-        let expected_successor = journal_record(1, "service-kill-bucket", "second-object", None, 2);
+        let mut expected_successor = journal_record(1, "service-kill-bucket", "second-object", None, 2);
        expected_successor.extend(journal_record(1, "service-kill-bucket", "first-object", None, 0));
        let flushed = wait_until(Duration::from_secs(10), || async {
            manager.operations_snapshot().await.queued_by_source.mrf == 1
                && committed_checkpoint_matches_on_all_disks(&disk_paths, 2, &expected_successor)
                && journal_matches_on_all_disks(&disk_paths, SCOPED_JOURNAL_REL, &expected_successor)
                && journal_matches_on_all_disks(&disk_paths, JOURNAL_REL, &expected_successor)
        })
        .await;
        assert!(
            flushed,
-            "child process must publish the pending successor snapshot before it can be killed"
+            "child process must publish the committed pending successor before it can be killed"
        );
        std::fs::write(&ready_path, b"ready").expect("write ready marker");
        loop {
@@ -608,7 +757,8 @@ fn mrf_authoritative_fsync_waiting_child_process_fixture() {
        write_journal_path_to_disks(&disk_paths, SCOPED_JOURNAL_REL, &startup);
        write_journal_path_to_disks(&disk_paths, JOURNAL_REL, &startup);
-        let successor = journal_record(1, "fsync-kill-bucket", "second-object", None, 2);
+        let mut successor = journal_record(1, "fsync-kill-bucket", "second-object", None, 2);
        successor.extend(journal_record(1, "fsync-kill-bucket", "first-object", None, 0));
        write_journal_path_to_disks_synced(&disk_paths, SCOPED_JOURNAL_REL, &successor);
        assert!(
            journal_matches_on_all_disks(&disk_paths, SCOPED_JOURNAL_REL, &successor)
@@ -665,9 +815,8 @@ async fn journal_replay_retains_child_process_anchor_when_manager_is_full() {
    );
 }
-/// If a process crashes after flushing a smaller successor snapshot but before
+/// A successor flush must preserve both pending work and accepted work whose
-/// deleting the startup anchor, the restarted process must replay the
+/// repair has not been proven when the process restarts.
 /// successor tail rather than losing it or merging it with stale records.
 #[tokio::test(flavor = "multi_thread", worker_threads = 4)]
 #[serial]
 async fn journal_replay_survives_successor_flush_before_delete() {
@@ -682,7 +831,8 @@ async fn journal_replay_survives_successor_flush_before_delete() {
    assert_eq!(status.code(), Some(78), "child process did not reach the successor flush boundary");
    let (disk_paths, storage) = heal_env_at(Some(temp_dir.path())).await;
-    let expected_successor = journal_record(1, "successor-bucket", "second-object", None, 2);
+    let mut expected_successor = journal_record(1, "successor-bucket", "second-object", None, 2);
    expected_successor.extend(journal_record(1, "successor-bucket", "first-object", None, 0));
    assert!(
        journal_matches_on_all_disks(&disk_paths, SCOPED_JOURNAL_REL, &expected_successor),
        "restarted process must see the pending successor snapshot"
@@ -690,24 +840,24 @@ async fn journal_replay_survives_successor_flush_before_delete() {
    let restarted = make_manager(storage);
    let replayed = mrf_queue::replay_journal_once(&restarted).await;
-    assert_eq!(replayed, 1, "restart after successor flush must replay only the still-pending tail");
+    assert_eq!(replayed, 2, "restart must replay both the admitted and pending responsibilities");
    assert_eq!(
        restarted.operations_snapshot().await.queued_by_source.mrf,
-        1,
+        2,
-        "the successor tail must be accepted after restart"
+        "both unproven successor responsibilities must be accepted after restart"
    );
    assert!(
        disk_paths.iter().all(|path| {
-            !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
+            Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
-                && !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
+                && Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
        }),
-        "a fully consumed successor snapshot may be deleted after restart replay"
+        "the accepted successor remains anchored until a verified repair proof"
    );
 }
 /// A service-style hard kill after successor flush must be equivalent to a
-/// crash at the flush-before-delete boundary: restart may replay the smaller
+/// crash at the flush-before-delete boundary: restart must recover every
-/// successor snapshot, but must not lose or merge stale startup records.
+/// unproven responsibility from the successor snapshot.
 #[tokio::test(flavor = "multi_thread", worker_threads = 4)]
 #[serial]
 #[cfg(unix)]
@@ -735,7 +885,8 @@ async fn journal_replay_survives_service_kill_after_successor_flush() {
    assert!(!status.success(), "child fixture must be terminated instead of exiting cleanly");
    let (disk_paths, storage) = heal_env_at(Some(temp_dir.path())).await;
-    let expected_successor = journal_record(1, "service-kill-bucket", "second-object", None, 2);
+    let mut expected_successor = journal_record(1, "service-kill-bucket", "second-object", None, 2);
    expected_successor.extend(journal_record(1, "service-kill-bucket", "first-object", None, 0));
    assert!(
        journal_matches_on_all_disks(&disk_paths, SCOPED_JOURNAL_REL, &expected_successor),
        "restarted process must see the successor snapshot produced before the kill"
@@ -743,25 +894,25 @@ async fn journal_replay_survives_service_kill_after_successor_flush() {
    let restarted = make_manager(storage);
    let replayed = mrf_queue::replay_journal_once(&restarted).await;
-    assert_eq!(replayed, 1, "restart after service kill must replay only the still-pending tail");
+    assert_eq!(replayed, 2, "service-kill restart must preserve every unproven responsibility");
    assert_eq!(
        restarted.operations_snapshot().await.queued_by_source.mrf,
-        1,
+        2,
-        "the successor tail must be accepted after service kill restart"
+        "both unproven responsibilities must be accepted after service kill restart"
    );
    assert!(
        disk_paths.iter().all(|path| {
-            !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
+            Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
-                && !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
+                && Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
        }),
-        "a fully consumed successor snapshot may be deleted after service-kill restart replay"
+        "the accepted successor remains anchored until a verified repair proof after service-kill restart"
    );
 }
 /// A hard kill between the authoritative successor fsync and the legacy mirror
-/// rewrite must prefer the canonical successor tail over the stale legacy
+/// rewrite must prefer the canonical successor over the stale legacy startup
-/// startup epoch. This models the mixed-version boundary conservatively: new
+/// epoch while retaining every unproven responsibility. New readers must not
-/// readers must not merge epochs, while the old mirror remains crash-visible.
+/// merge epochs, while the old mirror remains crash-visible.
 #[tokio::test(flavor = "multi_thread", worker_threads = 4)]
 #[serial]
 #[cfg(unix)]
@@ -789,7 +940,8 @@ async fn journal_replay_survives_sigkill_after_authoritative_successor_fsync_bef
    assert!(!status.success(), "child fixture must be terminated instead of exiting cleanly");
    let (disk_paths, storage) = heal_env_at(Some(temp_dir.path())).await;
-    let expected_successor = journal_record(1, "fsync-kill-bucket", "second-object", None, 2);
+    let mut expected_successor = journal_record(1, "fsync-kill-bucket", "second-object", None, 2);
    expected_successor.extend(journal_record(1, "fsync-kill-bucket", "first-object", None, 0));
    let stale_startup = {
        let mut startup = journal_record(1, "fsync-kill-bucket", "first-object", None, 0);
        startup.extend(journal_record(1, "fsync-kill-bucket", "second-object", None, 0));
@@ -806,17 +958,17 @@ async fn journal_replay_survives_sigkill_after_authoritative_successor_fsync_bef
    let restarted = make_manager(storage);
    let replayed = mrf_queue::replay_journal_once(&restarted).await;
-    assert_eq!(replayed, 1, "new reader must replay only the authoritative successor tail");
+    assert_eq!(replayed, 2, "new reader must recover every responsibility in the authoritative successor");
    assert_eq!(
        restarted.operations_snapshot().await.queued_by_source.mrf,
-        1,
+        2,
-        "the successor tail must be accepted after the fsync-boundary restart"
+        "both responsibilities must be accepted after the fsync-boundary restart"
    );
    assert!(
        disk_paths.iter().all(|path| {
-            !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
+            Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
-                && !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
+                && Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
        }),
-        "a fully consumed authoritative successor may clean both epochs after restart replay"
+        "the accepted authoritative successor remains anchored until a verified repair proof"
    );
 }
@@ -2098,11 +2098,7 @@ impl KmsBackend for LocalKmsBackend {
        let _write_guard = self.client.lock_key_for_write(key_id).await;
        // First, load the key from disk to get the master key
-        let mut master_key = self
+        let mut master_key = self.client.load_master_key(key_id).await?;
            .client
            .load_master_key(key_id)
            .await
            .map_err(|_| KmsError::key_not_found(format!("Key {key_id} not found")))?;
        let (deletion_date_str, deletion_date_dt) = if request.force_immediate.unwrap_or(false) {
            // Tombstone first: mark the record Deleted before removing the
@@ -2205,11 +2201,7 @@ impl KmsBackend for LocalKmsBackend {
        let _write_guard = self.client.lock_key_for_write(key_id).await;
        // Load the key from disk to get the master key
-        let mut master_key = self
+        let mut master_key = self.client.load_master_key(key_id).await?;
            .client
            .load_master_key(key_id)
            .await
            .map_err(|_| KmsError::key_not_found(format!("Key {key_id} not found")))?;
        if master_key.status != KeyStatus::PendingDeletion {
            return Err(KmsError::invalid_key_state(format!("Key {key_id} is not pending deletion")));
@@ -3038,6 +3030,111 @@ mod tests {
        assert!(matches!(error, KmsError::InvalidKey { .. }));
    }
    #[tokio::test]
    async fn delete_key_preserves_directory_io_error() {
        let (client, temp_dir) = create_dev_mode_client().await;
        client.create_key("existing-key", "AES_256", None).await.expect("create key");
        let backend = LocalKmsBackend { client };
        let offline_dir = TempDir::new().expect("create offline directory");
        let offline_key_dir = offline_dir.path().join("keys");
        fs::rename(temp_dir.path(), &offline_key_dir)
            .await
            .expect("move key directory offline");
        fs::write(temp_dir.path(), b"not a directory")
            .await
            .expect("replace key directory with a file");
        let error = backend
            .delete_key(DeleteKeyRequest {
                key_id: "existing-key".to_string(),
                ..Default::default()
            })
            .await
            .expect_err("unreadable storage must prevent scheduling deletion");
        fs::remove_file(temp_dir.path()).await.expect("remove replacement file");
        fs::rename(&offline_key_dir, temp_dir.path())
            .await
            .expect("restore key directory");
        assert!(matches!(error, KmsError::IoError { .. }), "got {error:?}");
        let key = backend
            .client
            .load_master_key("existing-key")
            .await
            .expect("read retained key");
        assert_eq!(key.status, KeyStatus::Active, "failed deletion must not mutate key state");
    }
    #[tokio::test]
    async fn cancel_key_deletion_preserves_directory_io_error() {
        let (client, temp_dir) = create_dev_mode_client().await;
        client.create_key("existing-key", "AES_256", None).await.expect("create key");
        let backend = LocalKmsBackend { client };
        backend
            .delete_key(DeleteKeyRequest {
                key_id: "existing-key".to_string(),
                ..Default::default()
            })
            .await
            .expect("schedule key deletion");
        let offline_dir = TempDir::new().expect("create offline directory");
        let offline_key_dir = offline_dir.path().join("keys");
        fs::rename(temp_dir.path(), &offline_key_dir)
            .await
            .expect("move key directory offline");
        fs::write(temp_dir.path(), b"not a directory")
            .await
            .expect("replace key directory with a file");
        let error = backend
            .cancel_key_deletion(CancelKeyDeletionRequest {
                key_id: "existing-key".to_string(),
            })
            .await
            .expect_err("unreadable storage must prevent cancelling deletion");
        fs::remove_file(temp_dir.path()).await.expect("remove replacement file");
        fs::rename(&offline_key_dir, temp_dir.path())
            .await
            .expect("restore key directory");
        assert!(matches!(error, KmsError::IoError { .. }), "got {error:?}");
        let key = backend
            .client
            .load_master_key("existing-key")
            .await
            .expect("read retained key");
        assert_eq!(
            key.status,
            KeyStatus::PendingDeletion,
            "failed cancellation must retain the deletion state"
        );
    }
    #[tokio::test]
    async fn test_load_master_key_directory_replaced_by_file_is_io_error() {
        let (client, temp_dir) = create_dev_mode_client().await;
        client.create_key("existing-key", "AES_256", None).await.expect("create key");
        let offline_dir = TempDir::new().expect("create offline directory");
        let offline_key_dir = offline_dir.path().join("keys");
        fs::rename(temp_dir.path(), &offline_key_dir)
            .await
            .expect("move key directory offline");
        fs::write(temp_dir.path(), b"not a directory")
            .await
            .expect("replace key directory with a file");
        let error = client
            .load_master_key("existing-key")
            .await
            .expect_err("a file in place of the key directory must fail");
        fs::remove_file(temp_dir.path()).await.expect("remove replacement file");
        fs::rename(&offline_key_dir, temp_dir.path())
            .await
            .expect("restore key directory");
        assert!(matches!(error, KmsError::IoError { .. }), "got {error:?}");
    }
    #[tokio::test]
    async fn test_load_master_key_accepts_legacy_rfc3339_timestamp() {
        let (client, _temp_dir) = create_dev_mode_client().await;
@@ -577,16 +577,13 @@ impl KmsServiceManager {
        Some(service_version.probe_worker.as_ref()?.status())
    }
-    /// Check backend health without changing the service lifecycle state.
+    /// Health check for the KMS service
    ///
    /// A transient backend failure leaves the published service available for
    /// subsequent checks and operations. Readiness uses the background probe
    /// to evaluate backend availability independently of lifecycle state.
    pub async fn health_check(&self) -> Result<bool> {
        let checked_state = self.state.load_full();
        match checked_state.current_service.as_ref() {
            Some(service_version) => {
                let manager = service_version.manager.clone();
                let checked_version = service_version.version;
                // Perform health check on the backend
                match manager.health_check().await {
                    Ok(healthy) => {
@@ -597,6 +594,8 @@ impl KmsServiceManager {
                    }
                    Err(e) => {
                        error!("KMS health check error: {}", e);
                        let _guard = self.lifecycle_mutex.lock().await;
                        self.mark_health_error_if_current(checked_version, &e);
                        Err(e)
                    }
                }
@@ -740,6 +739,17 @@ impl KmsServiceManager {
            task: std::sync::Mutex::new(Some(task)),
        }))
    }
    fn mark_health_error_if_current(&self, checked_version: u64, error: &KmsError) {
        let current = self.state.load_full();
        if current.current_service.as_ref().map(|version| version.version) == Some(checked_version) {
            self.state.store(Arc::new(RuntimeState {
                config: current.config.clone(),
                status: KmsServiceStatus::Error(format!("Health check failed: {error}")),
                current_service: current.current_service.clone(),
            }));
        }
    }
 }
 impl Default for KmsServiceManager {
@@ -994,6 +1004,19 @@ mod tests {
        assert!(manager.get_service_version().await.expect("restarted version") > first_version);
    }
    #[tokio::test]
    async fn stale_health_failure_cannot_poison_new_service_status() {
        let manager = KmsServiceManager::new();
        manager.configure(static_config("key-a", 0x11)).await.expect("configure");
        manager.start().await.expect("start");
        let old_version = manager.get_service_version().await.expect("old version");
        manager.restart().await.expect("restart");
        manager.mark_health_error_if_current(old_version, &KmsError::backend_error("stale failure"));
        assert_eq!(manager.get_status().await, KmsServiceStatus::Running);
    }
    #[tokio::test]
    async fn forbidden_local_master_key_change_preserves_running_config_and_service() {
        use crate::types::{CreateKeyRequest, KeyUsage};
@@ -75,44 +75,6 @@ fn unreachable_vault_config() -> KmsConfig {
    }
 }
 #[tokio::test]
 async fn transient_health_failure_does_not_latch_the_service_status() {
    let kms = TestKms::local().await;
    let manager = kms.manager();
    let service = manager.get_encryption_service().await.expect("running service");
    let version = manager.get_service_version().await.expect("running version");
    assert!(manager.health_check().await.expect("initial backend health"));
    // Move only this test's keys out of reach, then restore the same backend.
    let key_dir = kms.key_dir().expect("local key directory");
    let outage = tempfile::TempDir::new().expect("temporary outage directory");
    let hidden_keys = outage.path().join("keys");
    tokio::fs::rename(&key_dir, &hidden_keys)
        .await
        .expect("make backend unavailable");
    let failure = manager.health_check().await;
    let outage_status = manager.get_status().await;
    tokio::fs::rename(&hidden_keys, &key_dir).await.expect("restore backend");
    assert!(failure.is_err(), "the outage must surface as a health-check error");
    assert!(manager.health_check().await.expect("backend recovers without restart"));
    assert!(Arc::ptr_eq(
        &service,
        &manager.get_encryption_service().await.expect("service survives the outage")
    ));
    assert_eq!(manager.get_service_version().await, Some(version));
    assert_eq!(
        manager.get_status().await,
        KmsServiceStatus::Running,
        "a recovered backend must not leave service-status and readiness latched in Error"
    );
    assert_eq!(
        outage_status,
        KmsServiceStatus::Running,
        "backend health does not change the running service's lifecycle state"
    );
 }
 #[tokio::test]
 async fn starting_against_an_unreachable_backend_fails_without_publishing_a_service() {
    let manager = KmsServiceManager::new();
@@ -263,6 +263,14 @@ pub struct ScannerUsageRecoveryIntentResponse {
    pub mode: String,
    pub intent_id: String,
    pub state: String,
    #[serde(default)]
    pub actor_sha256: Option<String>,
    #[serde(default)]
    pub idempotency_key_sha256: Option<String>,
    #[serde(default)]
    pub request_sha256: Option<String>,
    #[serde(default)]
    pub accepted_at_unix_secs: Option<u64>,
    #[serde(flatten)]
    pub extra: serde_json::Map<String, serde_json::Value>,
 }
@@ -869,6 +877,10 @@ mod tests {
            "mode": "full-rebuild",
            "intent_id": "0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef",
            "state": "accepted",
            "actor_sha256": "1111111111111111111111111111111111111111111111111111111111111111",
            "idempotency_key_sha256": "2222222222222222222222222222222222222222222222222222222222222222",
            "request_sha256": "3333333333333333333333333333333333333333333333333333333333333333",
            "accepted_at_unix_secs": 7,
            "future": {"worker": "pending"}
        }))
        .unwrap();
@@ -877,7 +889,33 @@ mod tests {
        assert_eq!(intent.mode, "full-rebuild");
        assert_eq!(intent.intent_id, "0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef");
        assert_eq!(intent.state, "accepted");
        assert_eq!(
            intent.actor_sha256.as_deref(),
            Some("1111111111111111111111111111111111111111111111111111111111111111")
        );
        assert_eq!(
            intent.idempotency_key_sha256.as_deref(),
            Some("2222222222222222222222222222222222222222222222222222222222222222")
        );
        assert_eq!(
            intent.request_sha256.as_deref(),
            Some("3333333333333333333333333333333333333333333333333333333333333333")
        );
        assert_eq!(intent.accepted_at_unix_secs, Some(7));
        assert_eq!(intent.extra["future"]["worker"], "pending");
        let legacy_intent: ScannerUsageRecoveryIntentResponse = serde_json::from_value(json!({
            "status": "accepted",
            "action": "usage-full-rebuild",
            "mode": "full-rebuild",
            "intent_id": "0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef",
            "state": "accepted"
        }))
        .unwrap();
        assert!(legacy_intent.actor_sha256.is_none());
        assert!(legacy_intent.idempotency_key_sha256.is_none());
        assert!(legacy_intent.request_sha256.is_none());
        assert!(legacy_intent.accepted_at_unix_secs.is_none());
    }
    #[test]
@@ -971,7 +1009,7 @@ mod tests {
    #[tokio::test]
    async fn scanner_usage_async_reset_posts_explicit_intent_contract() {
        let server = TestServer::spawn(
-            r#"{"status":"accepted","action":"usage-full-rebuild","mode":"full-rebuild","intent_id":"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa","state":"accepted"}"#,
+            r#"{"status":"accepted","action":"usage-full-rebuild","mode":"full-rebuild","intent_id":"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa","state":"accepted","actor_sha256":"1111111111111111111111111111111111111111111111111111111111111111","idempotency_key_sha256":"2222222222222222222222222222222222222222222222222222222222222222","request_sha256":"3333333333333333333333333333333333333333333333333333333333333333","accepted_at_unix_secs":7}"#,
            202,
        )
        .await;
@@ -985,6 +1023,7 @@ mod tests {
        assert_eq!(accepted.status, "accepted");
        assert_eq!(accepted.mode, "full-rebuild");
        assert_eq!(accepted.state, "accepted");
        assert_eq!(accepted.accepted_at_unix_secs, Some(7));
        let request = server.recorded();
        assert_eq!(request.method, "POST");
        assert_eq!(request.path, "/rustfs/admin/v3/scanner/usage-state/reset");
@@ -166,24 +166,6 @@ fn rule_replicates(rule: &ReplicationRule, obj: &ObjectOpts) -> bool {
    }
 }
 fn replication_filter_tags_match(filter: &s3s::dto::ReplicationRuleFilter, object_tags: &HashMap<String, String>) -> bool {
    let tag_matches = |tag: &s3s::dto::Tag| match (&tag.key, &tag.value) {
        (None, None) => true,
        (Some(key), _) if key.is_empty() => true,
        (Some(key), Some(value)) => object_tags.get(key) == Some(value),
        _ => false,
    };
    filter
        .and
        .as_ref()
        .and_then(|and| and.tags.as_deref())
        .into_iter()
        .flatten()
        .chain(filter.tag.iter())
        .all(tag_matches)
 }
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
 pub enum ReplicationTargetValidationError {
    RoleWithMultipleDestinations,
@@ -722,7 +704,7 @@ impl ReplicationConfigurationExt for ReplicationConfiguration {
            if let Some(filter) = &rule.filter {
                let object_tags = ReplicationTagFilter::decode_tags_to_map(&obj.user_tags);
-                if replication_filter_tags_match(filter, &object_tags) {
+                if filter.test_tags(&object_tags) {
                    rules.push(rule.clone());
                }
            } else {
@@ -1157,47 +1139,6 @@ mod tests {
        assert_eq!(validate_replication_config_structure(&structure_config(vec![rule])), Ok(()));
    }
    #[test]
    fn actionable_rules_require_every_and_tag_to_match() {
        let mut rule = replication_rule("rule-1", "arn:target:a");
        rule.filter = Some(s3s::dto::ReplicationRuleFilter {
            and: Some(s3s::dto::ReplicationRuleAndOperator {
                prefix: None,
                tags: Some(vec![
                    s3s::dto::Tag {
                        key: Some("env".to_string()),
                        value: Some("prod".to_string()),
                    },
                    s3s::dto::Tag {
                        key: Some("tier".to_string()),
                        value: Some("gold".to_string()),
                    },
                ]),
            }),
            ..Default::default()
        });
        let config = structure_config(vec![rule]);
        let object = |user_tags: &str| ObjectOpts {
            name: "object".to_string(),
            user_tags: user_tags.to_string(),
            ..Default::default()
        };
        assert!(config.filter_target_arns(&object("env=prod")).is_empty());
        assert_eq!(config.filter_target_arns(&object("env=prod&tier=gold")), vec!["arn:target:a"]);
        assert!(config.filter_target_arns(&object("")).is_empty());
        let mut malformed = config;
        malformed.rules[0].filter.as_mut().unwrap().and.as_mut().unwrap().tags = Some(vec![s3s::dto::Tag {
            key: Some("env".to_string()),
            value: None,
        }]);
        assert!(
            malformed.filter_target_arns(&object("env=prod")).is_empty(),
            "a malformed tag filter must fail closed"
        );
    }
    #[test]
    fn structure_validation_allows_tag_filter_when_delete_marker_replication_disabled() {
        let mut rule = replication_rule("rule-1", "arn:target:a");
@@ -580,30 +580,6 @@ impl FailStats {
        FailedMetric { count, size }
    }
    /// Both rolling windows from one walk of the samples. `short` must be the
    /// narrower window; the walk stops at `long`. Callers that need both (the
    /// per-node site snapshot) would otherwise scan the deque twice while
    /// holding the bucket-stats read lock, and the deque is only bounded by
    /// the one-hour window - an unreachable target under load fills it.
    pub fn recent_windows(&self, short: Duration, long: Duration) -> (FailedMetric, FailedMetric) {
        let now = Instant::now();
        let mut short_metric = FailedMetric::default();
        let mut long_metric = FailedMetric::default();
        for sample in self.recent.iter().rev() {
            let age = now.duration_since(sample.observed_at);
            if age > long {
                break;
            }
            if age <= short {
                short_metric.count += 1;
                short_metric.size += sample.size;
            }
            long_metric.count += 1;
            long_metric.size += sample.size;
        }
        (short_metric, long_metric)
    }
    pub fn merge(&self, other: &FailStats) -> Self {
        Self {
            count: self.count.saturating_add(other.count),
@@ -936,26 +912,6 @@ mod tests {
        assert_eq!(last_hour.size, 96);
    }
    #[test]
    fn fail_stats_recent_windows_matches_two_separate_scans() {
        let mut stats = FailStats::default();
        stats.add_size(64, None::<&()>);
        stats.add_size(32, None::<&()>);
        let (minute, hour) = stats.recent_windows(Duration::from_secs(60), Duration::from_secs(60 * 60));
        let expected_minute = stats.recent_since(Duration::from_secs(60));
        let expected_hour = stats.recent_since(Duration::from_secs(60 * 60));
        assert_eq!((minute.count, minute.size), (expected_minute.count, expected_minute.size));
        assert_eq!((hour.count, hour.size), (expected_hour.count, expected_hour.size));
        assert_eq!(minute.count, 2);
        assert_eq!(hour.size, 96);
        let empty = FailStats::default();
        let (minute, hour) = empty.recent_windows(Duration::from_secs(60), Duration::from_secs(60 * 60));
        assert_eq!((minute.count, minute.size, hour.count, hour.size), (0, 0, 0, 0));
    }
    #[test]
    fn fail_stats_saturate_instead_of_wrapping() {
        let mut stats = FailStats {
@@ -97,9 +97,11 @@ pub use scanner::{
 pub use scanner_io::{
    ScannerDirtyUsageAckError, ScannerDirtyUsageBucket, ScannerDirtyUsageSnapshot, ScannerDirtyUsageState,
    acknowledge_dirty_usage_generation, acknowledge_scoped_dirty_usage, clear_dirty_usage_bucket, record_dirty_usage_bucket,
-    record_dirty_usage_object, record_scanner_maintenance_change, scanner_activity_epoch, scanner_dirty_usage_snapshot,
+    record_dirty_usage_bucket_from_producer, record_dirty_usage_object, record_dirty_usage_object_from_producer,
-    scanner_dirty_usage_state, scanner_maintenance_generation,
+    record_scanner_maintenance_change, scanner_activity_epoch, scanner_dirty_usage_snapshot, scanner_dirty_usage_state,
    scanner_maintenance_generation,
 };
 pub use segment_invalidation::SegmentInvalidationProducerIdentity;
 pub use sleeper::{DynamicSleeper, SCANNER_IDLE_MODE, SCANNER_SLEEPER};
 use std::sync::atomic::{AtomicU64, Ordering};
 pub use storage_api::ScannerReplicationConfig as ReplicationConfig;
@@ -3,7 +3,8 @@
 use super::*;
 use crate::segment_invalidation::{
    MAX_SEGMENT_INVALIDATION_BYTES, MAX_SEGMENT_INVALIDATION_ENTRIES, SegmentInvalidationDomain, SegmentInvalidationEnvelope,
-    SegmentInvalidationError, SegmentInvalidationProducer, SegmentInvalidationProof, admit_segment_invalidation,
+    SegmentInvalidationError, SegmentInvalidationProducer, SegmentInvalidationProducerIdentity, SegmentInvalidationProof,
    admit_segment_invalidation, complete_segment_invalidation_producers,
 };
 use std::collections::BTreeSet;
@@ -11,7 +12,8 @@ const MAX_WALK_SAMPLES: usize = 32;
 const MAX_WALK_BYTES: usize = 1024;
 fn segment_producers() -> BTreeSet<SegmentInvalidationProducer> {
-    SegmentInvalidationProducer::REQUIRED.into_iter().collect()
+    complete_segment_invalidation_producers(SegmentInvalidationProducerIdentity::REQUIRED_PRODUCTION)
        .expect("fixture should enumerate the complete production producer matrix")
 }
 fn segment_envelope() -> SegmentInvalidationEnvelope {
@@ -39,7 +39,7 @@ use s3s::dto::{
    BucketLifecycleConfiguration, ObjectLockConfiguration, ObjectLockEnabled, ReplicationConfiguration, VersioningConfiguration,
 };
 use sha2::{Digest as _, Sha256};
-use std::collections::{HashMap, HashSet};
+use std::collections::{BTreeSet, HashMap, HashSet};
 use std::future::Future;
 use std::path::Path;
 use std::pin::Pin;
@@ -1224,8 +1224,9 @@ pub(crate) use cache::{
 pub use dirty_usage::{
    ScannerDirtyUsageAckError, ScannerDirtyUsageBucket, ScannerDirtyUsageSnapshot, ScannerDirtyUsageState,
    acknowledge_dirty_usage_generation, acknowledge_scoped_dirty_usage, clear_dirty_usage_bucket, record_dirty_usage_bucket,
-    record_dirty_usage_object, record_scanner_maintenance_change, scanner_activity_epoch, scanner_dirty_usage_snapshot,
+    record_dirty_usage_bucket_from_producer, record_dirty_usage_object, record_dirty_usage_object_from_producer,
-    scanner_dirty_usage_state, scanner_maintenance_generation,
+    record_scanner_maintenance_change, scanner_activity_epoch, scanner_dirty_usage_snapshot, scanner_dirty_usage_state,
    scanner_maintenance_generation,
 };
 #[cfg(test)]
 pub(crate) use dirty_usage::{clear_dirty_usage_buckets_for_tests, dirty_usage_buckets_for_tests};
@@ -22,6 +22,11 @@ pub(super) static DIRTY_USAGE_BUCKETS: LazyLock<StdMutex<DirtyUsageBuckets>> = L
 // matching scope.
 pub(super) static DIRTY_USAGE_BUCKET_SCOPES: LazyLock<StdMutex<DirtyUsageBucketScopes>> =
    LazyLock::new(|| StdMutex::new(HashMap::new()));
 // Non-authoritative process-local producer coverage. Any future segment reuse
 // activation must bind this to the exact generation window and durable proof.
 pub(super) static DIRTY_USAGE_PRODUCER_IDENTITIES: LazyLock<
    StdMutex<BTreeSet<crate::segment_invalidation::SegmentInvalidationProducerIdentity>>,
 > = LazyLock::new(|| StdMutex::new(BTreeSet::new()));
 pub(super) static DIRTY_USAGE_BUCKET_NOTIFY: LazyLock<Notify> = LazyLock::new(Notify::new);
 pub(super) static SCANNER_ACTIVITY_EPOCH: LazyLock<String> = LazyLock::new(|| format!("{:032x}", rand::random::<u128>()));
 pub(super) static SCANNER_MAINTENANCE_GENERATION: AtomicU64 = AtomicU64::new(0);
@@ -153,6 +158,7 @@ fn apply_scoped_dirty_usage_ack(
 #[cfg(test)]
 mod scoped_dirty_usage_tests {
    use super::*;
    use crate::segment_invalidation::SegmentInvalidationProducerIdentity;
    #[test]
    fn scoped_dirty_usage_preserves_uncovered_newer_and_replayed_generations() {
@@ -213,6 +219,30 @@ mod scoped_dirty_usage_tests {
            assert_eq!(scopes, original_scopes);
        }
    }
    #[test]
    fn dirty_usage_tracks_known_segment_producer_identities_without_authorizing_unknown_sources() {
        clear_dirty_usage_buckets_for_tests();
        record_dirty_usage_object_from_producer("photos", "hot/object", SegmentInvalidationProducerIdentity::PutObject);
        record_dirty_usage_object_from_producer("photos", "archive/object", SegmentInvalidationProducerIdentity::DeleteObject);
        record_dirty_usage_bucket_from_producer("photos", SegmentInvalidationProducerIdentity::Unknown);
        assert_eq!(
            dirty_usage_producer_identities_for_tests(),
            BTreeSet::from([
                SegmentInvalidationProducerIdentity::PutObject,
                SegmentInvalidationProducerIdentity::DeleteObject
            ])
        );
        assert_eq!(
            dirty_usage_bucket_scopes_for_tests().get("photos"),
            Some(&DirtyUsageBucketScope::WholeBucket),
            "an unknown producer keeps the bucket dirty but must not count as producer coverage"
        );
        clear_dirty_usage_buckets_for_tests();
        assert!(dirty_usage_producer_identities_for_tests().is_empty());
    }
 }
 pub(super) fn dirty_usage_buckets() -> MutexGuard<'static, DirtyUsageBuckets> {
@@ -225,6 +255,13 @@ fn dirty_usage_bucket_scopes() -> MutexGuard<'static, DirtyUsageBucketScopes> {
        .unwrap_or_else(|poisoned| poisoned.into_inner())
 }
 fn dirty_usage_producer_identities()
 -> MutexGuard<'static, BTreeSet<crate::segment_invalidation::SegmentInvalidationProducerIdentity>> {
    DIRTY_USAGE_PRODUCER_IDENTITIES
        .lock()
        .unwrap_or_else(|poisoned| poisoned.into_inner())
 }
 pub(super) fn usize_to_u64_saturated(value: usize) -> u64 {
    u64::try_from(value).unwrap_or(u64::MAX)
 }
@@ -240,6 +277,22 @@ pub fn record_dirty_usage_bucket(bucket: &str) {
        return;
    }
    record_dirty_usage_bucket_inner(bucket);
 }
 pub fn record_dirty_usage_bucket_from_producer(
    bucket: &str,
    producer: crate::segment_invalidation::SegmentInvalidationProducerIdentity,
 ) {
    if bucket.is_empty() {
        return;
    }
    record_segment_invalidation_producer_identity(producer);
    record_dirty_usage_bucket_inner(bucket);
 }
 fn record_dirty_usage_bucket_inner(bucket: &str) {
    let pending_buckets = {
        let mut dirty_buckets = dirty_usage_buckets();
        let mut dirty_scopes = dirty_usage_bucket_scopes();
@@ -263,6 +316,23 @@ pub fn record_dirty_usage_bucket(bucket: &str) {
 /// local: after restart or any unverified distributed path the scanner falls
 /// back to its ordinary bucket scan.
 pub fn record_dirty_usage_object(bucket: &str, object: &str) {
    record_dirty_usage_object_inner(bucket, object);
 }
 pub fn record_dirty_usage_object_from_producer(
    bucket: &str,
    object: &str,
    producer: crate::segment_invalidation::SegmentInvalidationProducerIdentity,
 ) {
    if bucket.is_empty() {
        return;
    }
    record_segment_invalidation_producer_identity(producer);
    record_dirty_usage_object_inner(bucket, object);
 }
 fn record_dirty_usage_object_inner(bucket: &str, object: &str) {
    let Some(top_level_entry) = dirty_usage_top_level_entry(object) else {
        record_dirty_usage_bucket(bucket);
        return;
@@ -296,6 +366,17 @@ pub fn record_dirty_usage_object(bucket: &str, object: &str) {
    DIRTY_USAGE_BUCKET_NOTIFY.notify_one();
 }
 fn record_segment_invalidation_producer_identity(producer: crate::segment_invalidation::SegmentInvalidationProducerIdentity) {
    if producer.producer().is_some() {
        dirty_usage_producer_identities().insert(producer);
    }
 }
 #[cfg(test)]
 fn dirty_usage_producer_identities_for_tests() -> BTreeSet<crate::segment_invalidation::SegmentInvalidationProducerIdentity> {
    dirty_usage_producer_identities().clone()
 }
 fn dirty_usage_top_level_entry(object: &str) -> Option<String> {
    let (top_level_entry, _) = object.split_once('/').unwrap_or((object, ""));
    (!top_level_entry.is_empty()
@@ -577,6 +658,7 @@ pub(super) fn dirty_usage_bucket_count() -> usize {
 pub(crate) fn clear_dirty_usage_buckets_for_tests() {
    dirty_usage_buckets().clear();
    dirty_usage_bucket_scopes().clear();
    dirty_usage_producer_identities().clear();
 }
 #[cfg(test)]
@@ -16,23 +16,29 @@ use super::*;
 use crate::data_usage_define::{DATA_USAGE_OBJ_NAME_PATH, read_config_with_revision};
 async fn create_cohort_bucket(store: &ECStore, bucket: &str) {
    create_cohort_bucket_objects(store, bucket, 1).await;
 }
 async fn create_cohort_bucket_objects(store: &ECStore, bucket: &str, objects: usize) {
    store
        .make_bucket(bucket, &MakeBucketOptions::default())
        .await
        .expect("fixture bucket");
    for set in store.all_set_disks() {
-        let mut reader = ScannerPutObjReader::from_vec(b"cohort".to_vec());
+        for index in 0..objects {
-        set.put_object(
+            let mut reader = ScannerPutObjReader::from_vec(b"cohort".to_vec());
-            bucket,
+            set.put_object(
-            "initial",
+                bucket,
-            &mut reader,
+                &format!("object-{index:04}"),
-            &ScannerObjectOptions {
+                &mut reader,
-                no_lock: true,
+                &ScannerObjectOptions {
-                ..Default::default()
+                    no_lock: true,
-            },
+                    ..Default::default()
-        )
+                },
-        .await
+            )
-        .expect("fixture object and all rename tails should persist");
+            .await
            .expect("fixture object and all rename tails should persist");
        }
    }
 }
@@ -168,6 +174,92 @@ async fn service_cohort_production_dispatch_services_waiters_across_sources() {
    clear_dirty_usage_buckets_for_tests();
 }
 #[tokio::test]
 #[serial]
 async fn service_cohort_flat_bucket_budget_does_not_publish_unscanned_small_bucket() {
    let (_dir, store) = setup_two_pool_scanner_store().await;
    clear_dirty_usage_buckets_for_tests();
    let flat = format!("a-flat-{}", Uuid::new_v4().simple());
    let small = format!("z-small-{}", Uuid::new_v4().simple());
    create_cohort_bucket_objects(&store, &flat, 6).await;
    create_cohort_bucket(&store, &small).await;
    let cohort = Arc::new(StdMutex::new(ScannerServiceCohort::default()));
    let expected_flat = store
        .all_set_disks()
        .iter()
        .map(|set| (DataUsageCacheSource::new(set.pool_index, set.set_index), flat.clone()))
        .collect::<HashSet<_>>();
    let expected_small = store
        .all_set_disks()
        .iter()
        .map(|set| (DataUsageCacheSource::new(set.pool_index, set.set_index), small.clone()))
        .collect::<HashSet<_>>();
    let ctx = CancellationToken::new();
    let budget = ScannerCycleBudget::new_with_progress_tracking(
        &ctx,
        ScannerCycleBudgetConfig {
            max_objects: Some(1),
            ..Default::default()
        },
    );
    let (result, usage) = run_cohort_cycle(&store, cohort.clone(), 1, budget.clone()).await;
    assert_eq!(result.status, ScannerCycleStatus::Incomplete);
    assert!(usage.is_none(), "wide-bucket budget exhaustion must not publish a partial aggregate");
    assert!(budget.budget_elapsed());
    let first_round_admitted = cohort
        .lock()
        .expect("cohort lock")
        .admitted_members()
        .into_iter()
        .collect::<HashSet<_>>();
    assert!(
        !first_round_admitted.is_disjoint(&expected_flat),
        "the first fixed budget round should exercise the wide flat bucket"
    );
    assert!(
        first_round_admitted.is_disjoint(&expected_small),
        "a small bucket not yet reached by the real scanner must not be marked admitted"
    );
    for cycle in 2..=4 {
        let ctx = CancellationToken::new();
        let (result, usage) = run_cohort_cycle(
            &store,
            cohort.clone(),
            cycle,
            ScannerCycleBudget::new_with_progress_tracking(
                &ctx,
                ScannerCycleBudgetConfig {
                    max_objects: Some(1),
                    ..Default::default()
                },
            ),
        )
        .await;
        assert_eq!(result.status, ScannerCycleStatus::Incomplete);
        assert!(usage.is_none(), "mixed partial coverage still cannot publish the set root");
    }
    let admitted_after_budgeted_rounds = cohort
        .lock()
        .expect("cohort lock")
        .admitted_members()
        .into_iter()
        .collect::<HashSet<_>>();
    assert!(
        expected_small.is_subset(&admitted_after_budgeted_rounds),
        "tracked small buckets must receive real execution opportunities within their fixed service-round bound"
    );
    let ctx = CancellationToken::new();
    let (result, usage) =
        run_cohort_cycle(&store, cohort, 5, ScannerCycleBudget::new(&ctx, ScannerCycleBudgetConfig::default())).await;
    assert_eq!(result.status, ScannerCycleStatus::Complete);
    assert_eq!(usage.expect("final complete aggregate").objects_total_count, 14);
    clear_dirty_usage_buckets_for_tests();
 }
 #[tokio::test]
 #[serial]
 async fn service_cohort_fresh_complete_aggregate_preserves_reordered_sources() {
@@ -25,6 +25,7 @@ pub enum SegmentInvalidationError {
    ByteLimit,
    InvalidProof,
    InvalidKey,
    UnknownProducer,
 }
 #[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)]
@@ -50,6 +51,74 @@ impl SegmentInvalidationProducer {
    ];
 }
 #[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)]
 pub enum SegmentInvalidationProducerIdentity {
    PutObject,
    DeleteObject,
    DeleteMarker,
    CompleteMultipartUpload,
    Replication,
    TierTransition,
    TierExpiration,
    DirectoryObject,
    Unknown,
    TestFixture,
 }
 impl SegmentInvalidationProducerIdentity {
    pub const REQUIRED_PRODUCTION: [Self; 8] = [
        Self::PutObject,
        Self::DeleteObject,
        Self::DeleteMarker,
        Self::CompleteMultipartUpload,
        Self::Replication,
        Self::TierTransition,
        Self::TierExpiration,
        Self::DirectoryObject,
    ];
    pub fn producer(self) -> Option<SegmentInvalidationProducer> {
        match self {
            Self::PutObject => Some(SegmentInvalidationProducer::Put),
            Self::DeleteObject => Some(SegmentInvalidationProducer::Delete),
            Self::DeleteMarker => Some(SegmentInvalidationProducer::DeleteMarker),
            Self::CompleteMultipartUpload => Some(SegmentInvalidationProducer::Multipart),
            Self::Replication => Some(SegmentInvalidationProducer::Replication),
            Self::TierTransition | Self::TierExpiration => Some(SegmentInvalidationProducer::Tier),
            Self::DirectoryObject => Some(SegmentInvalidationProducer::DirectoryObject),
            Self::Unknown | Self::TestFixture => None,
        }
    }
 }
 pub fn complete_segment_invalidation_producers<I>(
    identities: I,
 ) -> Result<BTreeSet<SegmentInvalidationProducer>, SegmentInvalidationError>
 where
    I: IntoIterator<Item = SegmentInvalidationProducerIdentity>,
 {
    let mut covered_identities = BTreeSet::new();
    let mut producers = BTreeSet::new();
    for identity in identities {
        let Some(producer) = identity.producer() else {
            return Err(SegmentInvalidationError::UnknownProducer);
        };
        covered_identities.insert(identity);
        producers.insert(producer);
    }
    if SegmentInvalidationProducerIdentity::REQUIRED_PRODUCTION
        .iter()
        .all(|identity| covered_identities.contains(identity))
        && SegmentInvalidationProducer::REQUIRED
            .iter()
            .all(|producer| producers.contains(producer))
    {
        Ok(producers)
    } else {
        Err(SegmentInvalidationError::InvalidProof)
    }
 }
 #[derive(Clone, Copy, Debug, PartialEq, Eq)]
 pub enum SegmentInvalidationDomain {
    LocalSingleSet,
@@ -170,7 +239,8 @@ mod tests {
    use super::*;
    fn producers() -> BTreeSet<SegmentInvalidationProducer> {
-        SegmentInvalidationProducer::REQUIRED.into_iter().collect()
+        complete_segment_invalidation_producers(SegmentInvalidationProducerIdentity::REQUIRED_PRODUCTION)
            .expect("production producer matrix should be complete")
    }
    fn envelope() -> SegmentInvalidationEnvelope {
@@ -320,6 +390,63 @@ mod tests {
        );
    }
    #[test]
    fn segment_invalidation_producer_identities_must_be_known_and_complete() {
        assert_eq!(
            complete_segment_invalidation_producers(SegmentInvalidationProducerIdentity::REQUIRED_PRODUCTION),
            Ok(SegmentInvalidationProducer::REQUIRED.into_iter().collect())
        );
        assert_eq!(
            complete_segment_invalidation_producers([
                SegmentInvalidationProducerIdentity::PutObject,
                SegmentInvalidationProducerIdentity::DeleteObject,
                SegmentInvalidationProducerIdentity::DeleteMarker,
                SegmentInvalidationProducerIdentity::CompleteMultipartUpload,
                SegmentInvalidationProducerIdentity::Replication,
                SegmentInvalidationProducerIdentity::TierTransition,
                SegmentInvalidationProducerIdentity::DirectoryObject,
                SegmentInvalidationProducerIdentity::Unknown,
            ]),
            Err(SegmentInvalidationError::UnknownProducer)
        );
        assert_eq!(
            complete_segment_invalidation_producers([
                SegmentInvalidationProducerIdentity::PutObject,
                SegmentInvalidationProducerIdentity::DeleteObject,
                SegmentInvalidationProducerIdentity::DeleteMarker,
                SegmentInvalidationProducerIdentity::CompleteMultipartUpload,
                SegmentInvalidationProducerIdentity::Replication,
                SegmentInvalidationProducerIdentity::TierTransition,
                SegmentInvalidationProducerIdentity::DirectoryObject,
                SegmentInvalidationProducerIdentity::TestFixture,
            ]),
            Err(SegmentInvalidationError::UnknownProducer)
        );
        assert_eq!(
            complete_segment_invalidation_producers([
                SegmentInvalidationProducerIdentity::PutObject,
                SegmentInvalidationProducerIdentity::DeleteObject,
                SegmentInvalidationProducerIdentity::DeleteMarker,
                SegmentInvalidationProducerIdentity::Replication,
                SegmentInvalidationProducerIdentity::TierTransition,
                SegmentInvalidationProducerIdentity::DirectoryObject,
            ]),
            Err(SegmentInvalidationError::InvalidProof)
        );
        assert_eq!(
            complete_segment_invalidation_producers([
                SegmentInvalidationProducerIdentity::PutObject,
                SegmentInvalidationProducerIdentity::DeleteObject,
                SegmentInvalidationProducerIdentity::DeleteMarker,
                SegmentInvalidationProducerIdentity::CompleteMultipartUpload,
                SegmentInvalidationProducerIdentity::Replication,
                SegmentInvalidationProducerIdentity::TierTransition,
                SegmentInvalidationProducerIdentity::DirectoryObject,
            ]),
            Err(SegmentInvalidationError::InvalidProof)
        );
    }
    #[test]
    fn segment_invalidation_entries_are_bounded_and_key_checked() {
        let envelope = envelope();
@@ -919,22 +919,6 @@ where
            }
        }
        // The filename's item count is untrusted. Reject a payload that contains
        // more items than advertised instead of returning success and allowing the
        // caller to delete the entry with trailing events still in the file.
        match deserializer.next() {
            None => {}
            Some(Ok(_)) => {
                return Err(StoreError::Deserialization(format!(
                    "Batch for key {key} contains more than {} items",
                    key.item_count
                )));
            }
            Some(Err(e)) => {
                return Err(StoreError::Deserialization(format!("Failed to deserialize trailing batch item: {e}")));
            }
        }
        if items.is_empty() && key.item_count > 0 {
            return Err(StoreError::Deserialization("No items found".to_string()));
        }
@@ -1397,39 +1381,6 @@ mod tests {
        let _ = store.delete();
    }
    #[test]
    fn get_multiple_errors_on_batch_with_trailing_items_instead_of_partial_success() {
        let dir = temp_store_dir("trailing-batch-items");
        let store = QueueStore::<String>::new_with_compression(&dir, 8, ".test", false);
        store.open().unwrap();
        let items = vec!["aa".to_string(), "bb".to_string(), "cc".to_string()];
        let original_key = store.put_multiple(items).unwrap();
        assert_eq!(original_key.item_count, 3);
        // Keep the three-item payload but make its filename claim that it contains
        // only two items, simulating a corrupt or otherwise untrusted queue key.
        let original_path = store.file_path(&original_key);
        let advertised_key = Key {
            item_count: 2,
            ..original_key
        };
        let advertised_path = store.file_path(&advertised_key);
        std::fs::rename(&original_path, &advertised_path).unwrap();
        let err = store.get_multiple(&advertised_key).unwrap_err();
        assert!(
            matches!(err, StoreError::Deserialization(_)),
            "expected Deserialization error, got {err:?}"
        );
        // Because get_multiple failed, the batch entry remains available for
        // inspection or recovery instead of being silently discarded.
        assert!(advertised_path.exists());
        let _ = store.delete();
    }
    #[test]
    fn concurrent_put_raw_respects_entry_limit() {
        let dir = temp_store_dir("concurrent-limit");
@@ -1,23 +0,0 @@
 # Documentation
 Use the focused indexes rather than treating this directory as an unordered
 collection:
 - [Architecture knowledge base](architecture/README.md)
 - [Testing references](testing/README.md)
 ## Operations
 Operational runbooks live under [`operations/`](operations/). Replication
 operators should start with:
 | Runbook | Use it for |
 |---|---|
 | [Site replication operations](operations/site-replication-operations.md) | Health fields, pending operations, outage recovery, re-pair admission, IAM/SSE boundaries, and upgrades. |
 | [Replication target check](operations/replication-check.md) | Validating an S3 destination and version fidelity before enabling replication. |
 | [Replication object size limits](operations/replication-object-size-limits.md) | Multipart routing, large-object limits, and retry characteristics. |
 | [Replication outbound transport](operations/replication-outbound-transport.md) | Integrity headers, generic target behavior, and transport knobs. |
 Other runbooks remain grouped by filename in [`operations/`](operations/);
 architecture pages link to the relevant runbook where a cross-boundary
 procedure is required.
@@ -60,8 +60,6 @@ Required headings and strings in these files are asserted by `scripts/check_arch
 | [minio-rustfs-router-compatibility.md](minio-rustfs-router-compatibility.md) | a client or `mc` call that works against MinIO fails against RustFS and you need to know whether the endpoint is missing, stubbed, or deliberately different |
 | [minio-file-format-compat.md](minio-file-format-compat.md) | deciding whether a MinIO drive set, bucket-metadata blob, or SSE object can be read or imported by a given RustFS build, or before touching a listed version anchor |
-Operations runbooks are registered in the [documentation operations index](../README.md#operations), and testing references live in [../testing/README.md](../testing/README.md).
+Operations runbooks live in [../operations/](../operations/) and testing references in [../testing/README.md](../testing/README.md).
 For replication operations, start with [site replication operations](../operations/site-replication-operations.md), [replication target check](../operations/replication-check.md), [replication object size limits](../operations/replication-object-size-limits.md), and [replication outbound transport](../operations/replication-outbound-transport.md).
 For per-node HTTP failure ratios and cached storage probe provenance, see [S3 write failure diagnostics](../operations/s3-write-failure-diagnostics.md).
@@ -38,42 +38,6 @@ Counts ignore blank lines and comments; compute them from the files. The lifecyc
 "Supported" for the SSE row means RustFS encrypts and decrypts its own objects. MinIO SSE objects (SSE-S3, SSE-KMS, SSE-C) are not readable in default builds; see [minio-file-format-compat.md Part C](minio-file-format-compat.md#part-c--server-side-encryption-sse) for the `rio-v2` migration build.
 ## Replication Support Boundary
 Site replication and bucket replication are not the same compatibility claim.
 Site replication requires RustFS-compatible peer admin APIs and coordinates
 IAM, topology, buckets, and metadata. A generic S3-compatible service can only
 be a bucket-replication data target.
 For a generic S3 target, RustFS supports object PUT/HEAD/DELETE, multipart
 uploads, tags, version deletes, and Object Lock mutations when the target
 implements the corresponding S3 APIs and has versioning enabled. Targets that
 mint their own version IDs are supported through a per-target version ledger;
 pre-ledger replicas are adopted only when exact key and ETag identify one
 unambiguous target version. `NoSuchVersion` for an already absent addressed
 replica is treated as converged.
 The following are capability boundaries, not universal S3 claims:
 - `GET /BUCKET?replication-check` must pass the phases required by the intended
  workload. `VersionFidelity` may report a minting target as mismatched even
  though ledger-addressed delete and Object Lock phases succeed.
 - A target that rejects standard multipart constraints, required Object Lock
  integrity headers, or the configured checksum framing is unsupported until
  its transport settings are made compatible.
 - SSE-S3 and SSE-KMS are decrypted at the source and re-encrypted by the
  destination's KMS. SSE-C uses ciphertext passthrough and requires target
  evidence. Unsupported or ambiguous encryption metadata fails closed.
 - ACL authorization is intentionally unsupported, and generic targets never
  receive RustFS IAM/site-control-plane state.
 - RustFS does not guess between multiple target versions with the same key and
  ETag. The mutation remains failed and retryable until repair establishes an
  unambiguous mapping.
 See [site replication operations](../operations/site-replication-operations.md)
 for health, recovery, and upgrade rules and [replication outbound transport](../operations/replication-outbound-transport.md)
 for the tested target classes and knobs.
 ## Not Yet Passing
 Standard S3 areas that must not be described as complete:
@@ -66,6 +66,7 @@ The manifest has the following JSON contract (all fields are required):
 | `rounds`, `duration_seconds`, `min_free_bytes` | 3..10 groups, 900..86400 seconds for measured runs, and the independently estimated free-space reservation in bytes. Synthetic runs may use 1 second. |
 | `baseline`, `candidate` | Each contains executable `binary`, full 40-character `revision`, and verified `sha256`. The runner rehashes binaries before every leg. |
 | `fixed` | `config_sha256`, `dataset_sha256`, `release_flags`, `durability`, `disk_type`, `cache_state`, `load_command`, `resource_isolation`, `topology` (`EC8+4`), and positive `offered_load_ops`. Hashes use 64 lowercase hexadecimal characters. |
 | `release_evidence` | Required for `measured` runs. It binds the 3x4 EC8+4 topology, multi-pool/multi-set coverage, per-node metrics endpoints, same-window distributed sampling, process restart and crash-restart fault modes, mixed-version reader/writer/rollback participation, and allocation/flamegraph/RSS/save-frequency profile artifact requirements. Synthetic runs do not need this field and still cannot approve release evidence. |
 | `oracles` | A map with all five scenario names. Each value contains positive integer `objects`, `versions`, `bytes`, and `sha256` of the independently prepared canonical object/version/content manifest. |
 | `expected_healed_objects` | A map with all five scenario names and independently seeded repair counts. Running-heal and MRF-replay require a positive count. |
@@ -75,6 +76,14 @@ object/version/content result. Fix the foreground arrival rate (offered load),
 cache preparation procedure, configuration, and hardware across every leg.
 Do not include credentials in the manifest, adapter output, or saved commands;
 the collector reads `RUSTFS_ACCESS_KEY` and `RUSTFS_SECRET_KEY` from its environment.
 The adapter must echo the measured run's `release_evidence` object in every
 measurement response. A mismatch fails the cell because it means the deployment,
 mixed-version set, crash mode, or profiler contract no longer matches the
 operator-reviewed manifest. This echo is provenance binding only; it does not
 replace the independent correctness oracle, distributed metrics samples, profile
 artifacts, or ABBA comparison thresholds. The summary tool revalidates the same
 manifest contract before it can print a measured PASS result, so hand-built or
 trimmed reports without this provenance fail closed.
 #### Deployment Adapter Contract
@@ -204,6 +213,14 @@ The command prints only `PASS scanner_heal_perf ...` for measured passing ABBA
 evidence, otherwise `FAIL scanner_heal_perf ...`. The JSON and Markdown outputs
 carry the key p99/throughput/P1/P2/cache-cost fields and artifact provenance
 hashes; raw per-cell logs remain in the original artifact tree for audit.
 Failed or interrupted ABBA reports that contain only `status`, `performance`,
 `completed_cells`, and `error` also summarize as `FAIL`; they do not become
 performance evidence, and a missing comparison matrix is accepted only for a
 non-passing report.
 Measured passing reports must also retain the W10/W11 foreground-pressure,
 heal-lock-wait, and heal-attempt-cost fields emitted by the ABBA evaluator. If
 those fields are removed, empty, malformed, or length-mismatched, the quiet
 summary fails closed instead of treating the report as performance evidence.
 They cover the complete 120-cell schedule, data isolation, missing builds and
 oracles, zero samples/requests, swallowed request errors, offered-load drift,
@@ -1,258 +0,0 @@
 # Site Replication Operations
 **Use this when:** operating a site-replication deployment, diagnosing a peer
 outage or incomplete topology change, pairing sites that already contain data,
 or planning an upgrade.
 **Source of truth:** `rustfs/src/admin/handlers/site_replication.rs`,
 `rustfs/src/site_replication/`, and the bucket-replication worker under
 `crates/ecstore/src/bucket/replication/`.
 Site replication combines two different convergence paths:
 - the control plane replicates buckets, bucket metadata, IAM, and topology;
 - ordinary bucket replication moves object versions and delete operations.
 An `enabled: true` response only says that a site has more than one configured
 peer. It does not prove that every peer is reachable or caught up. Always read
 `pendingOperation`, `retryStats`, `PeerErrors`, and `Metrics` as well.
 ## Routine checks
 Run these commands from an admin workstation with one alias per site:
 ```console
 mc admin replicate info site-a
 mc admin replicate status site-a
 ```
 Check more than one site. A partition can leave each side with a different but
 locally valid view.
 `replicate info` is the compact control-plane view:
 | Field | Interpretation |
 |---|---|
 | `enabled` | More than one site is configured; this is not a health verdict. |
 | `sites` | The locally persisted topology. Compare deployment IDs and endpoints on every site. |
 | `retryStats.pending` | Collapsed peer deliveries waiting to be retried. |
 | `retryStats.failed` | Deliveries that crossed the escalation threshold and require attention. |
 | `retryStats.lastError` | A redacted summary of the most recent delivery failure. |
 | `pendingOperation` | A durable multi-step topology operation described below. Absence is the healthy steady state. |
 `replicate status` adds detailed convergence state:
 | Field | Interpretation |
 |---|---|
 | `Sites` / `PeerStates` | Configured peers and derived reachability/configuration state. |
 | `PeerErrors` | A peer could not be queried. Its detailed counters may be absent; do not read zeros as success. |
 | `BucketStats` | Per-bucket presence and versioning, replication, lifecycle, Object Lock, and metadata mismatches. |
 | `PolicyStats`, `UserStats`, `GroupStats` | IAM inventory mismatches. |
 | `RetryStats` | Durable control-plane retry backlog and escalation count. |
 | `Metrics.replMetrics` | Per-destination online state, downtime, replicated counts/bytes, and `failed` totals/windows. |
 | `Metrics.queued` / `Metrics.inProgress` | Object work waiting or active on the responding node. |
 | `Metrics.errors` | Node-level object-replication failures. When only queue statistics are available, RustFS synthesizes a node entry and preserves this counter rather than reporting zero. |
 | `Metrics.retries` | Redeliveries. Always zero today: a failed object is not retried by an event, it waits for the scanner pass described below. Read `errors` instead. |
 Healthy means: the same topology is visible on all sites, no pending operation,
 no peer error, no failed retry escalation, required bucket/IAM state is in sync,
 and queue/error counters are stable or falling. Counters are cumulative; alert on
 their rate and on a backlog that does not drain, not merely on a non-zero total.
 ## Pending operations and recovery
 `pendingOperation` contains `operation`, an opaque `id`, `pendingPeers`, and
 `ackedPeers`. Do not edit the site-replication state object by hand. The marker
 is the crash-recovery journal and removing it can make a partially applied
 operation look complete.
 The heavyweight reconciler runs once at startup and every 600 seconds. The
 lightweight retry drain runs every 30 seconds. A restart is therefore a valid
 way to cause an immediate heavyweight pass after the underlying fault has been
 fixed, but it is not a substitute for fixing connectivity, credentials, TLS,
 or the remote endpoint.
 ### `remove`
 The original topology and each peer acknowledgement are persisted before the
 operation finalizes. While peers remain in `pendingPeers`, restore access to
 them and wait for reconciliation. If a peer is permanently gone, a new remove
 request may remove all currently active unacknowledged peers; RustFS permits
 that request and then finalizes against the remaining topology. Removing the
 local site or all sites is also an explicit completion path.
 Do not re-add a site merely to hide this marker. First compare the topology on
 all reachable peers. If the same operation ID makes no progress for more than
 one heavyweight interval, collect `PeerErrors`, `RetryStats`, and the
 site-replication logs before retrying the remove.
 ### `rotate-svc-acct`
 Service-account rotation keeps the candidate secrets and peer acknowledgements
 until every current remote peer accepts the rotation. Restore the failing peer
 and allow the reconciler to resume it. Do not manually delete either candidate
 credential during this window: doing so can remove the only credential that a
 not-yet-acknowledged peer accepts.
 After the marker clears, verify `replicate status` from every site, then retire
 any separately retained old credential material according to local policy.
 ### `endpoint-refresh`
 An endpoint, CA, or TLS-verification edit first refreshes the replication
 target on every active peer and records acknowledgements. On startup and every
 heavyweight pass, RustFS probes peer capability, uses the endpoint-refresh API
 when supported (or the legacy peer-edit fallback), refreshes local bucket
 targets, and commits the edit only after every still-active peer acknowledges.
 If this marker is stuck:
 1. Confirm that the proposed endpoint and CA are correct and reachable from
   every site, not only from the admin workstation.
 2. Restore the site-replication service account and TLS trust path.
 3. Wait for one 600-second pass or restart one healthy node to trigger the
   startup pass.
 4. Re-run the identical edit only if the operation remains visible; a different
   endpoint edit is rejected while the existing refresh is pending. The journal
   pins the edit's payload, so a re-run without `--replicate-ilm-expiry` keeps
   the value the first attempt recorded, and a re-run asking for a different
   value is rejected. Finish or remove the pending refresh before changing it.
 A peer removed from the topology no longer blocks completion. A remove request
 is accepted when it removes every active unacknowledged peer.
 While this marker is present, control-plane retry replay to the other peers
 keeps running, but bucket wiring reconciliation waits: it rewrites the same
 targets the refresh is changing. Expect bucket-level drift on this site to
 persist until the refresh settles.
 ## Outage recovery and convergence time
 Control-plane retry begins on the 30-second drain, while heavyweight snapshots,
 pending topology operations, and bucket wiring are revisited on the 600-second
 pass. Object MRF entries are persisted every 10 seconds by default and target
 health is probed every 5 seconds. These are scheduling bounds, not delivery
 SLAs: network timeouts and the amount of queued work add to them.
 Objects that must be rediscovered by the scanner have this conservative upper
 bound before discovery:
 ```text
 RUSTFS_DATA_USAGE_UPDATE_DIR_CYCLES
    × max(RUSTFS_SCANNER_CYCLE, actual duration of one scanner cycle)
 ```
 The defaults re-descend a compacted directory every 16 cycles. A practical
 production starting point for a tighter recovery objective is
 `RUSTFS_DATA_USAGE_UPDATE_DIR_CYCLES=4`; `1` forces re-descent every cycle.
 Measure the additional disk and metadata load before lowering it further or
 tuning the scanner cadence. For an immediate operator-driven recovery, start a
 site resync with `mc admin replicate resync start` and monitor its status.
 Transfer time after discovery remains proportional to backlog size, bandwidth,
 worker capacity, and target latency. Use queue depth and the rate of
 `Metrics.errors` rather than the formula alone to decide whether convergence is
 progressing.
 ## Pairing sites that already contain data
 When more than one requested site is non-empty, preflight considers each bucket
 name held by more than one site:
 - versioning must be `Enabled` on every site holding the shared bucket;
 - Object Lock enablement must be identical on every holder.
 A bucket present on only one site is safe: post-add backfill creates it on the
 other peers. A shared unversioned bucket is rejected because merging can
 overwrite the only copy of an object. An Object Lock mismatch is rejected
 because lock enablement cannot be changed after bucket creation and convergence
 could otherwise strip a WORM guarantee.
 If preflight rejects the pair, keep the authoritative copy, delete the
 conflicting bucket (or its contents) from all other sites, run `replicate add`
 again, and then start `replicate resync` from the surviving site. Back up and
 validate the authoritative data before deleting anything.
 ## IAM convergence and repair boundary
 Ordinary IAM changes are delivered to each peer. A successful bulk IAM import
 also schedules one collapsed full-IAM snapshot per remote peer. A failed IAM
 deletion is replayed before that snapshot so the snapshot cannot re-create a
 principal or grant that was already revoked.
 The safety state has two bounds:
 - deletion high-water marks are retained for 30 days;
 - deletion replay bodies are capped at 256 distinct entities per peer.
 Repeated deletion of the same entity replaces its saved body. When the per-peer
 cap is exceeded or the body cannot be serialized, the retry entry remains
 escalated rather than pretending the deletion is replayable. An item from an
 older sender without a source timestamp cannot install the 30-day high-water
 mark, so verify it explicitly after a prolonged split. A successful drain
 clears replay bodies; removing the peer prunes its bodies. For an escalated IAM
 retry, use the site-replication repair workflow for the affected peer and IAM
 family, then verify users, service accounts, groups, policies, and mappings on
 both sides. Repair is the operator's explicit accountability transfer and
 clears the saved deletion bodies only after the IAM repair succeeds.
 A group's status converges in one direction. An explicit disable is applied
 everywhere, including through a snapshot, but a membership change never
 carries an enable - it would otherwise re-enable a group frozen on the
 receiving site. If a group ended up disabled on one site only, re-enable it
 there explicitly with `mc admin group enable`; a snapshot or repair will not
 do it.
 Treat IAM divergence as a security incident: a user deleted on one site can
 remain usable on an unreachable peer until replay or repair completes. A peer
 whose IAM entry is escalated does not receive scheduled snapshots either -
 including the one a bulk import schedules - until the repair settles it.
 ## Encrypted objects
 | Source form | Replication behavior | Fail-closed condition |
 |---|---|---|
 | SSE-S3 | The source decrypts the object; the request sends only `AES256` intent; the destination encrypts with its own KMS. Source envelope material never leaves the site. | The destination cannot satisfy the encryption request, or the source metadata is incomplete/unsupported. The replica is `FAILED`; plaintext is not silently stored. |
 | SSE-KMS | The source decrypts the object; the request sends `aws:kms` intent without the source-local key ID; the destination selects its own configured KMS key. | Either side cannot decrypt/encrypt, or the metadata mixes incompatible encryption evidence. |
 | SSE-C | Stored ciphertext and the required SSE-C replication transport metadata pass through. RustFS verifies target evidence before accepting the replica. | The target does not echo the customer-algorithm evidence, required material/layout is absent, or the metadata is ambiguous. |
 Unknown MinIO/RustFS encryption markers are never forwarded as ordinary user
 metadata. They fail replication so an operator must migrate or repair the
 object with a supported format.
 ## Rolling upgrades and rollback
 Keep every node in one site on the same version whenever possible. Upgrade all
 nodes of one site consecutively, verify its startup reconciliation and status,
 then move to the next site. Do not intentionally leave a site mixed-version:
 admin requests can land on different nodes, and an older node may not resume a
 new pending-operation shape or expose its health fields.
 Current state additions are optional and defaulted, so older readers ignore
 them. The target-version ledger is stored as dual-prefixed internal object
 metadata and is also ignored by older readers; rollback does not corrupt the
 object format, but older code loses the assigned-version routing improvement.
 Before rolling back across the fix that retains the data directory of a version
 awaiting purge replication (rustfs/rustfs#7307), ensure no version purge is
 pending. Older code can free that retained version's data directory before the
 remote purge is acknowledged, leaving unreadable metadata and blocking bucket
 deletion. Drain or repair replication and take a metadata/data backup first.
 ## Runtime knobs
 These values are read when the owning background task starts. Restart the
 server after changing them. The millisecond intervals have a 10 ms floor;
 invalid values fall back to the default with a warning.
 | Variable | Default | Effect |
 |---|---:|---|
 | `RUSTFS_REPL_HEALTH_CHECK_INTERVAL_MS` | `5000` | Remote-target health probe interval. Lowering it increases outbound probes. |
 | `RUSTFS_REPL_MRF_FLUSH_INTERVAL_MS` | `10000` | Maximum periodic interval between MRF persistence flushes; 1,000 new entries also trigger a flush. |
 | `RUSTFS_REPL_RESYNC_POLL_MAX_MS` | `60000` | Upper bound for randomized resync retry-poll sleep. |
 | `RUSTFS_REPL_RESYNC_MAX_JOBS` | `2` | Concurrent resync jobs; values are bounded to `1..=32`. |
 Transport-specific controls and target behavior are documented in
 [Replication outbound transport](replication-outbound-transport.md). Validate a
 new destination with [Replication target check](replication-check.md), and read
 [Replication object size limits](replication-object-size-limits.md) before
 moving large objects.
@@ -3,7 +3,7 @@
 **Use this when:** a check is red and you need to know whether it blocks the merge, which workflow and job produced it, and how to reproduce it locally.
 **Source of truth:** the live `main` ruleset (command below) for required status; `.github/workflows/<file>.yml` for triggers, `paths`, `timeout-minutes`, and cron; `.config/nextest.toml` for e2e profile filters; `.github/scheduled-validations.json` for the freshness-watchdog list.
-A job blocks a merge when its exact check name is required by the live `main` ruleset, or when its result is required by the `Test and Lint` aggregate. A workflow name, a `merge_group` trigger, or an unrelated red PR check does not make a job required by itself.
+A job blocks a merge only when its exact check name is in the live `main` ruleset. A workflow name, a `merge_group` trigger, or a red PR check does not make a job required by itself.
 ## Required merge checks
@@ -13,11 +13,9 @@ The `main` ruleset (`6436880`) requires exactly these contexts, with `strict_req
 |---|---|---|
 | `CLA Check` | `cla.yml` | Contributor agreement |
 | `Quick Checks` | `ci.yml` job `quick-checks` | Formatting and repository guard scripts |
-| `Test and Lint` | `ci.yml` job `required-checks` | Exact expected results for every CI validation job, including workspace checks, critical E2E, feature lanes, and event-specific full suites |
+| `Test and Lint` | `ci.yml` job `test-and-lint` | Clippy, workspace nextest (`ci` profile, excluding `e2e_test`), doctests, migration-gate count (`scripts/check_migration_gate_count.sh`) |
-Every PR enters `ci.yml`. The `classify-changes` job uses the base revision of `scripts/ci_gate.py` to select a conservative documentation-only path: root Markdown/licenses, `AGENTS.md`, Markdown under `docs/` or `.agents/skills/`, and documentation images. Unknown paths, unavailable Git history, an empty diff, or a missing base policy select the full matrix. Renames include their deleted source path. Documentation-only PRs still run Quick Checks and Typos; the aggregate requires the expensive jobs to be skipped exactly as selected.
+For PRs limited to the `paths-ignore` list in `ci.yml`, `ci-docs-only.yml` reports `Quick Checks` and `Test and Lint` under the same names; it runs the quick checks and `scripts/check_no_planning_docs.sh`, not a Rust build or tests. `scripts/check_ci_paths_sync.sh` keeps the two path lists aligned.
 `required-checks` runs even after failed or skipped dependencies. `scripts/ci_gate.py verify` rejects missing jobs, unexpected jobs, failure, cancellation, and unexpected skips; optional lanes are required only on their declared events. `Workspace Test and Lint` is the ordinary Rust job, while `Test and Lint` uniquely names the aggregate. New validation jobs must update both its direct dependencies and the script contract. Test this wiring and its failure cases with `python3 scripts/ci_gate.py --self-test`.
 Verify the live rule before changing merge policy:
@@ -26,36 +24,37 @@ gh api repos/rustfs/rustfs/rulesets/6436880 \
  --jq '.rules[] | select(.type == "required_status_checks") | .parameters'
 ```
-The aggregate requires the validation lanes already selected by `ci.yml`; this closes the gap where a failing critical lane left the required workspace check green. Independent workflows remain report-only unless separately required. Before adding a new expensive lane or moving existing PR coverage to a schedule, collect representative execution and regression evidence, establish ownership and a working scheduled replacement, and update this reference with the resulting policy.
+Promotion rule: never promote a report-only lane to required from one green run. Require at least 14 days and 30 representative PRs with at least 99% complete execution, then update the ruleset and this file together.
 ## Pull request and merge matrix
-"Via aggregate" means a wrong result fails the required `Test and Lint` check. "Report-only" means visible and actionable but outside both the required list and aggregate. Budgets are each job's `timeout-minutes` in the named workflow and are not copied here.
+"Report-only" means visible and actionable but not in the required list. Budgets are each job's `timeout-minutes` in the named workflow and are not copied here.
 | Event | Check name | Workflow / job | Merge status | Reproduce |
 |---|---|---|---|---|
 | PR, non-doc change | `Quick Checks` | `ci.yml` `quick-checks` | Required | `make pre-commit` |
-| PR, non-doc change | `Workspace Test and Lint` | `ci.yml` `test-and-lint` | Via aggregate | `cargo clippy --all-targets -- -D warnings`; `cargo nextest run --profile ci --all --exclude e2e_test`; `cargo test --all --doc`; `scripts/check_migration_gate_count.sh` |
+| PR, non-doc change | `Test and Lint` | `ci.yml` `test-and-lint` | Required | `cargo clippy --all-targets -- -D warnings`; `cargo nextest run --profile ci --all --exclude e2e_test`; `cargo test --all --doc`; `scripts/check_migration_gate_count.sh` |
-| PR, non-doc change | `Typos` | `ci.yml` `typos` | Via aggregate | `typos` |
+| PR, non-doc change | `Typos` | `ci.yml` `typos` | Report-only | `typos` |
-| PR, non-doc change | `ILM Integration (serial)` | `ci.yml` `test-ilm-integration-serial` | Via aggregate | exact command in the job |
+| PR, non-doc change | `ILM Integration (serial)` | `ci.yml` `test-ilm-integration-serial` | Report-only | exact command in the job |
-| PR, non-doc change | `Test and Lint (rio-v2)`, `Test and Lint (swift)`, `Test and Lint (sftp)` | `ci.yml` `test-and-lint-rio-v2`, `test-and-lint-protocols` | Via aggregate | `cargo nextest run` with the job's `--features` |
+| PR, non-doc change | `Test and Lint (rio-v2)`, `Test and Lint (swift)`, `Test and Lint (sftp)` | `ci.yml` `test-and-lint-rio-v2`, `test-and-lint-protocols` | Report-only | `cargo nextest run` with the job's `--features` |
-| PR, non-doc change | `Connect Short Credential Boundary` | `ci.yml` `connect-short-credential-boundary` | Via aggregate | `cargo test -p rustfs --test connect_registration --features connect-e2e-short-credentials`; `cargo check -p rustfs --release --features connect-e2e-short-credentials` must fail |
+| PR, non-doc change | `Connect Short Credential Boundary` | `ci.yml` `connect-short-credential-boundary` | Report-only | `cargo test -p rustfs --test connect_registration --features connect-e2e-short-credentials`; `cargo check -p rustfs --release --features connect-e2e-short-credentials` must fail |
-| PR, non-doc change | `Build RustFS Debug Binary` | `ci.yml` `build-rustfs-debug-binary` | Via aggregate; prerequisite for black-box jobs | `cargo build -p rustfs --bins --features e2e-test-hooks` |
+| PR, non-doc change | `Build RustFS Debug Binary` | `ci.yml` `build-rustfs-debug-binary` | Report-only; prerequisite for the black-box jobs | `cargo build -p rustfs --bins` |
-| PR, non-doc change | `io_uring Integration (real)` | `ci.yml` `uring-integration` | Via aggregate | `cargo test -p rustfs-ecstore --lib uring_ -- --test-threads=1 --nocapture` |
+| PR, non-doc change | `io_uring Integration (real)` | `ci.yml` `uring-integration` | Report-only | `cargo test -p rustfs-ecstore --lib uring_ -- --test-threads=1 --nocapture` |
-| PR, non-doc change | `End-to-End Tests` | `ci.yml` `e2e-tests` | Via aggregate | `cargo nextest run --profile e2e-smoke -p e2e_test`, then `./scripts/e2e-run.sh ./target/debug/rustfs <data-dir>`; membership guards `scripts/check_test_wiring.py --check-profile e2e-smoke <listing.json>` and `scripts/check_security_smoke_count.sh check <listing.json>` |
+| PR, non-doc change | `End-to-End Tests` | `ci.yml` `e2e-tests` | Report-only | `cargo nextest run --profile e2e-smoke -p e2e_test`, then `./scripts/e2e-run.sh ./target/debug/rustfs <data-dir>`; membership guards `scripts/check_test_wiring.py --check-profile e2e-smoke <listing.json>` and `scripts/check_security_smoke_count.sh check <listing.json>` |
-| PR, non-doc change | `S3 Implemented Tests` | `ci.yml` `s3-implemented-tests` | Via aggregate | build `rustfs`, then `scripts/s3-tests/run.sh` with the job's `DEPLOY_MODE` / `TEST_MODE` / `MAXFAIL` env |
+| PR, non-doc change | `S3 Implemented Tests` | `ci.yml` `s3-implemented-tests` | Report-only | build `rustfs`, then `scripts/s3-tests/run.sh` with the job's `DEPLOY_MODE` / `TEST_MODE` / `MAXFAIL` env |
-| PR, non-doc change | `S3 Lifecycle Behavior Tests` | `ci.yml` `s3-lifecycle-behavior-tests` | Via aggregate | `scripts/s3-tests/run.sh` with the job's accelerated-scanner env |
+| PR, non-doc change | `S3 Lifecycle Behavior Tests` | `ci.yml` `s3-lifecycle-behavior-tests` | Report-only | `scripts/s3-tests/run.sh` with the job's accelerated-scanner env |
-| PR touching `paths` in `audit.yml` | `Cargo Deny`, `Workflow Pin Report`, `Dependency Review` | `audit.yml` `cargo-deny`, `workflow-pin-report`, `dependency-review` | Report-only | `cargo deny check`; `scripts/security/check_workflow_pins.sh` |
+| PR to `main` or `release` touching `paths` in `audit.yml` | `Cargo Deny`, `Workflow Pin Report`, `Dependency Review` | `audit.yml` `cargo-deny`, `workflow-pin-report`, `dependency-review` | Report-only | `cargo deny check`; `scripts/security/check_workflow_pins.sh` |
 | Push to `main` or `release` touching `paths` in `audit.yml` | `Cargo Deny`, `Workflow Pin Report` | `audit.yml` `cargo-deny`, `workflow-pin-report` | Report-only | `cargo deny check`; `scripts/security/check_workflow_pins.sh` |
 | PR touching `paths` in `architecture-migration-rules.yml` | `Architecture Migration Rules` | `architecture-migration-rules.yml` `architecture-migration-rules` | Report-only | `scripts/check_architecture_migration_rules.sh` |
 | PR touching `paths` in `nix.yml` | `Nix Build & Check` | `nix.yml` `nix-validation` | Report-only | `nix flake check` |
 | PR touching `paths` in `fuzz.yml` | `Build Fuzz Harness`, `Smoke / <target>` | `fuzz.yml` `fuzz-build`, `pr-fuzz-smoke` | Report-only | `MAX_TOTAL_TIME=60 ./scripts/fuzz/run.sh` |
 | PR touching `paths` in `windows-filesystem.yml` | `Rename Safety` | `windows-filesystem.yml` `rename-safety` | Report-only | the `cargo test -p rustfs-ecstore --lib <filter>` commands in the job, on Windows |
 | PR touching `paths` in `coverage.yml` | `Workspace line coverage` | `coverage.yml` `coverage` | Report-only | `make coverage`; `python3 scripts/check_security_coverage.py target/llvm-cov/coverage.json` |
-| PR touching `paths` in `e2e-upgrade.yml` | `Direct upgrade from the previous release`, `Mixed-version rolling upgrade from the previous release`, `Bucket configuration survives the upgrade`, `Rollback reads current bucket metadata` | `e2e-upgrade.yml` `upgrade` matrix | Report-only | the `cargo test --locked -p e2e_test` command in the job with `RUSTFS_UPGRADE_SOURCE_BINARY` pointing at the pinned previous release (`UPGRADE_SOURCE_VERSION`) |
+| PR touching `paths` in `e2e-upgrade.yml` | `Direct upgrade from the previous release`, `Mixed-version rolling upgrade from the previous release`, `Bucket configuration survives the upgrade`, `Rollback reads current bucket metadata`, `ODM configuration recovery after rc.5 rollback`, `Multipart layouts survive the rc.5 upgrade`, `rc.5 multipart replication baseline` | `e2e-upgrade.yml` `upgrade` matrix | Report-only | the `cargo test --locked -p e2e_test` command in the job with `RUSTFS_UPGRADE_SOURCE_BINARY` pointing at the pinned previous release (`UPGRADE_SOURCE_VERSION`) |
 | PR touching `paths` in `oidc-keycloak.yml` | `OIDC Keycloak live gate` | `oidc-keycloak.yml` `oidc-keycloak-live` | Report-only | `cargo build --locked -p rustfs --bin rustfs`, then `bash scripts/test/oidc_keycloak_live.sh ./target/debug/rustfs` |
 | PR touching `paths` in `targets-integration.yml` | `PostgreSQL, MySQL, AMQP, and NATS` | `targets-integration.yml` `targets-live` | Report-only | start the containers as in the job, export the `RUSTFS_TEST_*` DSNs, then the job's `cargo test --locked -p rustfs-targets --test <name> -- --ignored --test-threads=1` commands |
-| PR, documentation-only selection | `Quick Checks`, `Typos`, `Test and Lint` | `ci.yml` `quick-checks`, `typos`, `required-checks` | Required directly or via aggregate | Quick Checks commands; `python3 scripts/ci_gate.py --self-test` |
+| PR limited to main-CI-excluded paths | `Quick Checks`, `Test and Lint` | `ci-docs-only.yml` `quick-checks`, `test-and-lint` | Required | `git diff --check`; `make doc-paths-check`; `scripts/check_no_planning_docs.sh` |
-| `merge_group`; push to `main` | `End-to-End Tests (full merge gate)` | `ci.yml` `e2e-full` | Via aggregate on these events | `cargo nextest run --profile e2e-full -p e2e_test` |
+| `merge_group`; push to `main` or `release` | `End-to-End Tests (full merge gate)` | `ci.yml` `e2e-full` | Report-only | `cargo nextest run --profile e2e-full -p e2e_test` |
 e2e filters live in `.config/nextest.toml`; extend a profile instead of adding a second selector. Before a profile runs, `scripts/check_test_wiring.py` compares its listing to the committed digest in `.config/e2e-<profile>-selection.txt`, so a silent test drop fails closed.
@@ -64,16 +63,16 @@ cost. `data_usage_test` runs in the PR `e2e-smoke` lane so changes that affect
 authoritative scanner usage publication, quota-visible usage, or admin usage
 snapshots get an end-to-end signal before merge review. `heal_erasure_disk_rebuild_test`
 runs in `e2e-full` so core erasure heal rebuild regressions are caught no later
-than the merge queue or `main` push lane; it also remains in `e2e-nightly` with
+than the merge queue or `main`/`release` push lane; it also remains in `e2e-nightly` with
 the serialized cluster fault-domain suites for scheduled soak signal.
 ## Scheduled validation
-Scheduled lanes never block a PR. Their workflow-local gate fails the run, scheduled failures route to the shared failure-issue action, and `scheduled-validation-freshness.yml` fails when a workflow listed in `.github/scheduled-validations.json` has no recent attempt or completed successful scheduled run within its `max_age_hours` (a `never_ran_grace_until` entry covers the window before a newly enabled cron's first slot). Cadence is qualitative here; the cron lives in each workflow's `on.schedule`.
+Scheduled lanes never block a PR. Their workflow-local gate fails the run, scheduled failures route to the shared failure-issue action, and `scheduled-validation-freshness.yml` fails when a workflow listed in `.github/scheduled-validations.json` has not run within its `max_age_hours` (a `never_ran_grace_until` entry covers the window before a newly enabled cron's first slot). Cadence is qualitative here; the cron lives in each workflow's `on.schedule`.
 | Workflow (cadence) | Jobs | Verdict and artifacts | In freshness list | Reproduce |
 |---|---|---|---|---|
-| `ci.yml` (weekly) | full matrix, including the schedule/dispatch-only rio-v2 jobs `build-rustfs-debug-binary-rio-v2` and `e2e-tests-rio-v2` | strict aggregate; the full E2E lane runs on dispatch, merge groups, and main pushes | yes | dispatch `ci.yml` |
+| `ci.yml` (weekly) | full matrix, including the schedule/dispatch-only rio-v2 jobs `build-rustfs-debug-binary-rio-v2` and `e2e-tests-rio-v2` | per-job | yes | dispatch `ci.yml` |
 | `build.yml` (weekly) | `build-rustfs` over the six-target platform matrix in `prepare-platform-matrix` (four Linux, macOS aarch64, Windows x86_64) | build/package integrity | yes | dispatch `build.yml` with an exact platform set |
 | `e2e-replication-nightly.yml` (nightly) | `repl-nightly`, `cluster-nightly`, `protocols-nightly` | three independent gates; JUnit, membership listing, server logs | yes | `cargo nextest run --profile e2e-repl-nightly -p e2e_test`; `--profile e2e-nightly`; `-j 1 --profile e2e-protocols` |
 | `e2e-distributed.yml` (storage-sensitive PRs + nightly) | `distributed` | fail-closed 4-node 4-disk S3, durability, replication, movement, fault, and direct/rolling upgrade gate; JUnit, membership listing, per-node server logs | yes, with `never_ran_grace_until` | download the pinned previous release as in the workflow, export `RUSTFS_UPGRADE_SOURCE_BINARY`, then `cargo nextest run --profile e2e-distributed -p e2e_test` |
@@ -87,10 +86,10 @@ Scheduled lanes never block a PR. Their workflow-local gate fails the run, sched
 | `mint.yml` (weekly) | `mint` | report-only by design; per-suite PASS/FAIL/NA and raw `log.json` | yes | pinned Docker sequence in the workflow |
 | `coverage.yml` (weekly) | `coverage` | report-only trend; lcov and JSON artifact | yes | `make coverage` |
 | `runner-hygiene.yml` (monthly) | `check-ephemerality` | runner ephemerality | yes | dispatch |
-| `e2e-upgrade.yml` (weekly) | `upgrade` (4-case matrix) | upgrade and rollback gate; server logs | no | see the PR row |
+| `e2e-upgrade.yml` (weekly) | `upgrade` (7-case matrix) | upgrade and rollback gate; server logs | no | see the PR row |
 | `oidc-keycloak.yml` (weekly) | `oidc-keycloak-live` | live OIDC gate | no | see the PR row |
 | `targets-integration.yml` (nightly) | `targets-live` | live target gate; container logs | no | see the PR row |
-| `scheduled-validation-freshness.yml` (nightly) | `check-freshness` | fails on missing or stale attempts or completed successes | n/a | dispatch |
+| `scheduled-validation-freshness.yml` (nightly) | `check-freshness` | fails on a never-created or stale schedule | n/a | dispatch |
 Manual `workflow_dispatch` runs are debugging evidence and do not open scheduled-failure issues. A manual performance run may explicitly allow a known regression; that override is not a passing baseline.
@@ -150,11 +149,21 @@ and exact S3 content.
 This case is a **four-node, one-drive-per-node process-restart test**. It is not
 power-loss validation, a 3x4 EC8+4 experiment, an all-version inventory, or proof
 of scanner enumeration, exact MRF disposition, legacy migration, or rollback.
-The registry keeps all G01-G14/P1-P4 and R-E/R-D/R-L release requirements pending
+The schema 2 registry separates the implemented single-set restart lane from
-until their actual feature-specific oracles and required topologies exist.
+structured release lanes for authority coverage, checkpoint/crash, status and
-Missing cases cannot be supplied by synthetic W20 results. W20's bounded JSON
+outcome, MRF responsibility, mixed-version rollback, scheduler pressure,
-and file-hash helpers are reused; its ABBA performance contracts remain in
+maintenance producers, and EC8+4 multi-set coverage. All G01-G14/P1-P4 and
 R-E/R-D/R-L release requirements stay `pending` until their actual
 feature-specific oracles, measurements and required topologies exist. Missing
 cases cannot be supplied by synthetic W20 results. W20's bounded JSON and
 file-hash helpers are reused; its ABBA performance contracts remain in
 `docs/operations/scanner-benchmark-runbook.md`.
 Measured ABBA manifests must also carry the runbook's `release_evidence`
 contract. The runner rejects reports that cannot bind the exact 3x4 EC8+4
 topology, multi-pool/multi-set shape, distributed same-window metrics endpoints,
 restart/crash modes, mixed-version reader/writer/rollback participation, and
 allocation/flamegraph/RSS/save-frequency profile artifact plan. Synthetic runs
 and manifests missing that contract remain harness-only evidence.
 ### Recording One Case
@@ -234,14 +243,64 @@ For automation, `--check-scanner-heal-release "$RUN_DIR"` emits one compact
 JSON decision and exits nonzero while blocked. `verified_cases` contains only
 cases that pass the complete receipt, build provenance, nextest/JUnit and real
 oracle checks; `rejected_cases` names registered cases that do not, and
-`pending_gates` names the unimplemented release requirements. Approval requires
+`pending_gates` names the unimplemented release requirements and
-every registered case to verify, `pending_gates` to be empty, and a future
+`pending_lanes` names the structured release lanes that still need real
-registry schema capable of representing the complete release matrix. Schema 1
+evidence. Schema 1 is deliberately marked `release_schema_capable: false`
-is deliberately marked `release_schema_capable: false`: it models only the
+because it models only the single-version, unversioned-object restart/crash
-single-version, unversioned-object restart/crash cases and cannot represent
+cases. Schema 2 can describe the wider release matrix, but approval still
-mixed-version, rollback, EC8+4 or performance evidence. A focused run,
+requires every registered case to verify and every required gate to leave
-synthetic harness, compile-only result, skipped/retried test, ordinary CI
+`pending` only after a future checker can bind it to real feature-specific
-success, or removal of pending text therefore cannot become a release approval.
+evidence. The current checker hard-rejects missing structured requirements and
 pending gates mapped to an implemented lane, so clearing pending text cannot
 become approval. A focused run, synthetic harness, compile-only result,
 skipped/retried test, ordinary CI success, or unregistered mixed-version,
 rollback, EC8+4 or performance claim therefore cannot become a release approval.
 For high-risk rollback gates, `evidence_fields` records the specific proof
 fields that a future real-evidence checker must bind before a pending gate can
 move out of the blocked set. G03 keeps scoped ACK tied to durable root
 publication, ACK request identity, participating peer capability snapshots, and
 mixed-peer fallback oracles; G09 keeps mixed-version reader, writer, and rollback
 payload evidence explicit. These fields are part of the release contract, not
 evidence by themselves.
 When the real release lanes have produced their dedicated artifacts, validate
 the complete hard-gate bundle with:
 ```bash
 scripts/python_bin.sh scripts/check_test_wiring.py \
  --check-scanner-heal-release-bundle /path/to/release-evidence.json
 ```
 The bundle checker is intentionally stricter than the case checker. It requires
 schema 2 registry metadata, `evidence: measured`, the current checkout revision,
 all G01-G14/P1-P4/R-E/R-D/R-L gates, per-gate `status: pass`, lane identity,
 relative artifact paths, matching SHA256 hashes, and non-empty summaries. It
 also binds each evidence field to its own run provenance: `source_revision`,
 `run_id`, `measurement_window_id`, timezone-qualified `started_at` and
 `finished_at`, command arguments, and artifact format. The field
 `source_revision` must match the bundle revision, and measured performance
 duration cannot exceed the recorded run window.
 The hard evidence shape remains claim-specific: mixed-version gates must name at
 least two participating versions, crash/durable replay gates must include
 crash-boundary evidence, G14 must record EC8+4 with at least three nodes and four
 drives per node plus multi-set and multi-pool evidence, performance gates need
 measured durations, P3's pressure run needs at least two hours, and P1 needs a
 symbolized profile summary with resolved samples. Every G14 field and every
 performance gate's fields must also share one `measurement_window_id`, so EC8+4,
 multi-set/multi-pool, ABBA, throughput, and profiling artifacts cannot be
 stitched together from unrelated runs. P1 `profile_evidence` must bind every
 required profile artifact kind (`allocation-profile`, `flamegraph`,
 `rss-samples`, and `save-frequency`) with a relative path, artifact format,
 non-empty file, matching SHA256, and the same measurement window when a
 per-artifact window is declared. Missing, synthetic, stale, tampered, undersized,
 or topology-mismatched evidence returns a compact blocked or invalid JSON result
 and a nonzero exit.
 This command validates the evidence package; it does not create evidence. A
 handwritten JSON file, a synthetic harness pass, a single focused case, or a
 local unit fixture still cannot satisfy the distributed, mixed-version,
 crash-restart, durable MRF replay, EC8+4, ABBA, or profiling gates.
 Run parser/receipt regressions with
 `scripts/python_bin.sh scripts/check_test_wiring.py --self-test`. Those fixtures
@@ -12,7 +12,52 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 use std::path::Path;
 use std::process::Command;
 fn git(root: &Path, args: &[&str]) -> Option<String> {
    let output = Command::new("git").args(args).current_dir(root).output().ok()?;
    output
        .status
        .success()
        .then(|| String::from_utf8_lossy(&output.stdout).trim().to_owned())
 }
 fn main() -> shadow_rs::SdResult<()> {
    let manifest = std::env::var_os("CARGO_MANIFEST_DIR").ok_or("missing CARGO_MANIFEST_DIR")?;
    let root = Path::new(&manifest).join("..");
    println!("cargo:rerun-if-changed=build.rs");
    // Match the E2E build identity's source and configuration inputs.
    for path in [
        "crates",
        "rustfs",
        "Cargo.toml",
        "Cargo.lock",
        "rust-toolchain.toml",
        ".cargo",
        ".config",
    ] {
        let path = root.join(path);
        if path.exists() {
            println!("cargo:rerun-if-changed={}", path.display());
        }
    }
    // Resolve worktree-local HEAD/index and common refs through Git. Watching
    // refs also covers tags and a new loose ref created from packed refs.
    for name in ["HEAD", "index", "refs", "packed-refs"] {
        if let Some(path) = git(&root, &["rev-parse", "--git-path", name]) {
            let path = Path::new(&path);
            let path = if path.is_absolute() {
                path.to_owned()
            } else {
                root.join(path)
            };
            if path.exists() {
                println!("cargo:rerun-if-changed={}", path.display());
            }
        }
    }
    println!("cargo:rerun-if-env-changed=RUSTFS_BUILD_VERSION");
    if let Ok(version) = std::env::var("RUSTFS_BUILD_VERSION")
        && !version.is_empty()
@@ -627,32 +627,6 @@ pub(crate) async fn cluster_replication_stats(bucket: &str, context: Option<Arc<
        .await
 }
 /// Reload the bucket's metadata on every peer so a follow-up
 /// `put-bucket-replication` on another node does not read a stale target.
 ///
 /// Best effort, like every S3 bucket-config write path
 /// (`app::bucket_usecase::notify_bucket_metadata_reload`): the target is
 /// already persisted and live on this node, and the 15-minute refresh closes
 /// the gap, so a peer that cannot be reached must not turn a completed write
 /// into a failed request.
 async fn notify_remote_target_metadata_reload(bucket: &str, context: Option<Arc<AppContext>>, action: &'static str) {
    let Some(notification_system) = current_notification_system_for_context(context.as_deref()) else {
        return;
    };
    if let Err(err) = notification_system.load_bucket_metadata(bucket).await {
        warn!(
            event = EVENT_ADMIN_REMOTE_TARGET_STATE,
            component = LOG_COMPONENT_ADMIN,
            subsystem = LOG_SUBSYSTEM_REPLICATION,
            action = action,
            result = "peer_metadata_reload_failed",
            bucket = %bucket,
            error = ?err,
            "admin remote target state"
        );
    }
 }
 fn unique_replication_peers(peer_clients: &[Option<PeerRestClient>]) -> (Vec<&PeerRestClient>, u32) {
    let mut seen_grid_hosts = HashSet::new();
    let peers: Vec<_> = peer_clients
@@ -725,7 +699,6 @@ pub struct SetRemoteTargetHandler {}
 impl Operation for SetRemoteTargetHandler {
    async fn call(&self, req: S3Request<Body>, _params: Params<'_, '_>) -> S3Result<S3Response<(StatusCode, Body)>> {
        let cred = validate_replication_admin_request(&req, AdminAction::SetBucketTargetAction).await?;
        let app_context = app_context_from_req(&req);
        let queries = extract_query_params(&req.uri);
@@ -953,8 +926,6 @@ impl Operation for SetRemoteTargetHandler {
            .map_err(map_bucket_target_error)?;
        let _targets_guard = lock_bucket_targets_metadata(bucket).await;
        let arn = persist_remote_target_write(bucket, remote_target, incarnation, mode).await?;
        drop(_targets_guard);
        notify_remote_target_metadata_reload(bucket, app_context, "set_remote_target").await;
        let arn_str = serde_json::to_string(&arn)
            .map_err(|_| S3Error::with_message(S3ErrorCode::InternalError, "Failed to serialize target ARN"))?;
@@ -1035,7 +1006,6 @@ pub struct RemoveRemoteTargetHandler {}
 impl Operation for RemoveRemoteTargetHandler {
    async fn call(&self, req: S3Request<Body>, _params: Params<'_, '_>) -> S3Result<S3Response<(StatusCode, Body)>> {
        validate_replication_admin_request(&req, AdminAction::SetBucketTargetAction).await?;
        let app_context = app_context_from_req(&req);
        debug!("remove remote target called");
        let queries = extract_query_params(&req.uri);
@@ -1111,7 +1081,6 @@ impl Operation for RemoveRemoteTargetHandler {
        }
        let json_targets = serde_json::to_vec(&targets)
            .map_err(|_| S3Error::with_message(S3ErrorCode::InternalError, "Failed to serialize targets"))?;
        let notification_bucket = bucket.clone();
        let bucket = bucket.clone();
        let arn = arn_str.clone();
        // The pool cancellation owns a detached task. Both outer guards must
@@ -1132,8 +1101,6 @@ impl Operation for RemoveRemoteTargetHandler {
            S3Error::with_message(S3ErrorCode::InternalError, format!("remote target removal task failed: {error}"))
        })??;
        notify_remote_target_metadata_reload(&notification_bucket, app_context, "remove_remote_target").await;
        Ok(S3Response::new((StatusCode::NO_CONTENT, Body::from("".to_string()))))
    }
 }
@@ -1820,25 +1787,6 @@ mod tests {
        pairs.iter().map(|(k, v)| (k.to_string(), v.to_string())).collect()
    }
    #[test]
    fn remote_target_writes_notify_peer_metadata_caches() {
        let source = include_str!("replication.rs");
        for (start, end) in [
            ("impl Operation for SetRemoteTargetHandler", "pub struct ListRemoteTargetHandler"),
            ("impl Operation for RemoveRemoteTargetHandler", "async fn cancel_active_resync_intent"),
        ] {
            let body = source
                .split(start)
                .nth(1)
                .and_then(|rest| rest.split(end).next())
                .expect(start);
            assert!(
                body.contains("notify_remote_target_metadata_reload"),
                "{start} must notify every node before returning success"
            );
        }
    }
    #[test]
    fn update_ops_parse_minio_query_contract() {
        let ops = parse_remote_target_update_ops(&query_map(&[
@@ -75,6 +75,10 @@ struct ScannerRecoveryIntentResponse {
    mode: String,
    intent_id: String,
    state: String,
    actor_sha256: String,
    idempotency_key_sha256: String,
    request_sha256: String,
    accepted_at_unix_secs: u64,
 }
 #[derive(Debug, Serialize)]
@@ -332,6 +336,10 @@ fn scanner_recovery_intent_record_response(
        mode: record.mode,
        intent_id: record.intent_id,
        state: record.state,
        actor_sha256: record.actor_sha256,
        idempotency_key_sha256: record.idempotency_key_sha256,
        request_sha256: record.request_sha256,
        accepted_at_unix_secs: record.accepted_at_unix_secs,
    };
    let body = serde_json::to_vec(&response).map_err(|err| {
        S3Error::with_message(
@@ -371,10 +379,7 @@ fn scanner_recovery_intent_accept_response(
 fn scanner_recovery_intent_executor_id(result: &rustfs_scanner::ScannerRecoveryIntentAcceptResult) -> Option<String> {
    match result {
-        rustfs_scanner::ScannerRecoveryIntentAcceptResult::Accepted { record }
+        rustfs_scanner::ScannerRecoveryIntentAcceptResult::Accepted { record } if record.state == "accepted" => {
        | rustfs_scanner::ScannerRecoveryIntentAcceptResult::Replayed { record }
            if matches!(record.state.as_str(), "accepted" | "running") =>
        {
            Some(record.intent_id.clone())
        }
        _ => None,
@@ -673,7 +678,7 @@ mod tests {
    }
    #[test]
-    fn scanner_recovery_intent_executor_only_starts_non_terminal_work() {
+    fn scanner_recovery_intent_executor_starts_only_newly_accepted_work() {
        let mut record = rustfs_scanner::ScannerRecoveryIntentRecord {
            schema_version: 1,
            intent_id: "0".repeat(64),
@@ -699,7 +704,16 @@ mod tests {
                record: record.clone(),
            })
            .as_deref(),
-            Some(record.intent_id.as_str())
+            None,
            "a lost-response retry must not start a duplicate executor"
        );
        record.state = "accepted".to_string();
        assert!(
            scanner_recovery_intent_executor_id(&rustfs_scanner::ScannerRecoveryIntentAcceptResult::Replayed {
                record: record.clone(),
            })
            .is_none(),
            "replayed accepted records remain durable for startup/control recovery instead of duplicating work"
        );
        record.state = "completed".to_string();
        assert!(
@@ -1318,24 +1318,6 @@ impl Operation for ImportIam {
            failed,
        };
        // The entities are already imported locally. A snapshot that cannot be
        // scheduled is a convergence delay the reconcile pass still closes, so
        // it must not turn a completed import into a failed request - the same
        // best-effort contract every other site-replication hook here follows.
        if let Err(err) =
            crate::site_replication::enqueue_site_replication_iam_snapshot("iam import scheduled a full snapshot").await
        {
            warn!(
                component = LOG_COMPONENT_ADMIN,
                subsystem = LOG_SUBSYSTEM_USER,
                event = EVENT_ADMIN_USER_STATE,
                action = "import_iam",
                result = "site_replication_snapshot_not_scheduled",
                error = ?err,
                "admin user state"
            );
        }
        let body = serde_json::to_vec(&ret).map_err(|e| S3Error::with_message(S3ErrorCode::InternalError, e.to_string()))?;
        let mut header = HeaderMap::new();
@@ -1442,16 +1424,6 @@ mod tests {
        assert!(include_str!("user.rs").contains(mapper_call));
    }
    #[test]
    fn import_iam_enqueues_a_site_replication_snapshot() {
        let body = source_block(include_str!("user.rs"), "impl Operation for ImportIam");
        assert!(
            body.contains("enqueue_site_replication_iam_snapshot"),
            "a successful IAM import must schedule a full IAM snapshot for every remote site"
        );
    }
    #[test]
    fn test_should_check_deny_only_for_regular_self_request() {
        let cred = Credentials {
@@ -409,26 +409,6 @@ fn transfer_summaries(stats: &InternalReplicationStats) -> (XferSummaryWire, Tar
    (summary, per_target)
 }
 /// Node-level failure counters for `errors`. The sibling `retries` field
 /// stays zero on purpose: it means redeliveries in the minio-go shape, and a
 /// failed object is not retried by an event today (it waits for the scanner's
 /// heal pass), so reporting failures there would claim a redelivery that
 /// never happened.
 fn failure_counters(stats: &InternalReplicationStats) -> CounterSummaryWire {
    let (total, last1m, last1hr) = stats.stats.values().fold((0i64, 0i64, 0i64), |acc, stat| {
        (
            acc.0.saturating_add(stat.fail_stats.count),
            acc.1.saturating_add(stat.fail_stats.last_minute.count),
            acc.2.saturating_add(stat.fail_stats.last_hour.count),
        )
    });
    CounterSummaryWire {
        total: u64::try_from(total.max(0)).unwrap_or_default(),
        last1m: u64::try_from(last1m.max(0)).unwrap_or_default(),
        last1hr: u64::try_from(last1hr.max(0)).unwrap_or_default(),
    }
 }
 impl MetricsV2Wire {
    /// Project the aggregated internal stats onto the `MetricsV2` shape.
    ///
@@ -438,7 +418,6 @@ impl MetricsV2Wire {
    /// `queueStats.nodes` and treats an empty list as "no data".
    pub(crate) fn from_stats(bucket_stats: &BucketStats, node_name: &str) -> Self {
        let (xfer_stats, tgt_xfer_stats) = transfer_summaries(&bucket_stats.replication_stats);
        let failed = failure_counters(&bucket_stats.replication_stats);
        let mut nodes: Vec<ReplQNodeStatsWire> = bucket_stats
            .queue_stats
            .nodes
@@ -457,7 +436,6 @@ impl MetricsV2Wire {
                q_stats: InQueueMetricWire::from(&bucket_stats.replication_stats.q_stat),
                xfer_stats: xfer_stats.clone(),
                tgt_xfer_stats: tgt_xfer_stats.clone(),
                errors: failed,
                ..Default::default()
            });
        } else {
@@ -466,7 +444,6 @@ impl MetricsV2Wire {
            if let Some(first) = nodes.first_mut() {
                first.xfer_stats = xfer_stats.clone();
                first.tgt_xfer_stats = tgt_xfer_stats.clone();
                first.errors = failed;
            }
        }
@@ -501,12 +478,6 @@ mod tests {
        target.replicated_size = 4096;
        target.failed.count = 3;
        target.failed.size = 900;
        target.fail_stats.count = 3;
        target.fail_stats.size = 900;
        target.fail_stats.last_minute.count = 2;
        target.fail_stats.last_minute.size = 600;
        target.fail_stats.last_hour.count = 3;
        target.fail_stats.last_hour.size = 900;
        target.bandwidth_limit_bytes_per_sec = 1024;
        target.current_bandwidth_bytes_per_sec = 512.5;
        stats
@@ -566,10 +537,6 @@ mod tests {
        assert_eq!(node["queueStats"]["peak"], node["queueStats"]["max"]);
        assert!(node["activeWorkers"].get("curr").is_some());
        assert!(node["transferSummary"].get("Total").is_some());
        assert_eq!(node["errors"]["total"], 3);
        assert_eq!(node["errors"]["last1m"], 2);
        assert_eq!(node["errors"]["last1hr"], 3);
        assert_eq!(node["retries"]["total"], 0, "failures are not redeliveries; retries must not claim one");
        assert_eq!(json["downtimeInfo"], serde_json::json!({}));
    }
@@ -814,7 +814,11 @@ impl DefaultObjectUsecase {
                    }
                }
-                rustfs_scanner::record_dirty_usage_object(&bucket, &key);
+                rustfs_scanner::record_dirty_usage_object_from_producer(
                    &bucket,
                    &key,
                    rustfs_scanner::SegmentInvalidationProducerIdentity::PutObject,
                );
                Ok::<_, S3Error>((oi, dest_versioned))
            }
        });
@@ -1175,7 +1175,12 @@ impl DefaultObjectUsecase {
        let manager = get_capacity_manager();
        manager.record_write_operation().await;
        let _ = helper.complete(&result);
-        rustfs_scanner::record_dirty_usage_object(&bucket, &key);
+        let producer = if delete_marker && version_id_clone.is_none() {
            rustfs_scanner::SegmentInvalidationProducerIdentity::DeleteMarker
        } else {
            rustfs_scanner::SegmentInvalidationProducerIdentity::DeleteObject
        };
        rustfs_scanner::record_dirty_usage_object_from_producer(&bucket, &key, producer);
        result
    }
 }
@@ -689,7 +689,11 @@ impl DefaultObjectUsecase {
                    schedule_object_replication(obj_info.clone(), store, completion_replication_decision).await;
                }
-                rustfs_scanner::record_dirty_usage_object(&bucket, &key);
+                rustfs_scanner::record_dirty_usage_object_from_producer(
                    &bucket,
                    &key,
                    rustfs_scanner::SegmentInvalidationProducerIdentity::CompleteMultipartUpload,
                );
                Ok::<_, ApiError>(obj_info)
            }
        });
@@ -2054,7 +2054,11 @@ impl DefaultObjectUsecase {
                    schedule_object_replication(obj_info.clone(), store, dsc).await;
                }
-                rustfs_scanner::record_dirty_usage_object(&bucket, &key);
+                rustfs_scanner::record_dirty_usage_object_from_producer(
                    &bucket,
                    &key,
                    rustfs_scanner::SegmentInvalidationProducerIdentity::PutObject,
                );
                rustfs_io_metrics::record_put_object_stage_duration_from("app_post_store_bookkeeping", post_store_stage_start);
                let capacity_update_stage_start = put_stage_metrics_enabled.then(Instant::now);
@@ -217,26 +217,6 @@ pub(crate) fn settle_observed_site_replication_retry_event(
    before.saturating_sub(queue.len())
 }
 /// Make sure `peer` has a collapsed entry for `path` without counting the
 /// call as a delivery failure. A bulk local mutation (`import-iam`) needs the
 /// entry to exist so the next drain sends the snapshot; routing it through
 /// [`upsert_site_replication_retry_event`] would raise `retry_count` on every
 /// import and escalate a healthy peer to `failed` after
 /// [`SITE_REPLICATION_RETRY_FAILED_AFTER`] of them, with the scheduling note
 /// shown to operators as `lastError`.
 pub(crate) fn ensure_site_replication_retry_event(
    queue: &mut Vec<SiteReplicationRetryEvent>,
    peer: &PeerInfo,
    path: &str,
    reason: &str,
 ) -> S3Result<Vec<SiteReplicationRetryEvent>> {
    let path = collapsed_retry_queue_path(path).unwrap_or(path);
    if queue.iter().any(|event| retry_event_matches(event, peer, path)) {
        return Ok(Vec::new());
    }
    push_site_replication_retry_event(queue, peer, path, summarize_peer_error_detail(reason), false, None)
 }
 pub(crate) fn upsert_site_replication_retry_event(
    queue: &mut Vec<SiteReplicationRetryEvent>,
    peer: &PeerInfo,
@@ -264,17 +244,6 @@ pub(crate) fn upsert_site_replication_retry_event(
        return Ok(Vec::new());
    }
    push_site_replication_retry_event(queue, peer, path, detail, peer_unreachable, generation)
 }
 fn push_site_replication_retry_event(
    queue: &mut Vec<SiteReplicationRetryEvent>,
    peer: &PeerInfo,
    path: &str,
    detail: String,
    peer_unreachable: bool,
    generation: Option<u64>,
 ) -> S3Result<Vec<SiteReplicationRetryEvent>> {
    let slots_needed = queue
        .len()
        .saturating_add(1)
@@ -305,7 +274,7 @@ fn push_site_replication_retry_event(
        retry_count: 1,
        failed: false,
        last_error: detail,
-        updated_at: Some(OffsetDateTime::now_utc()),
+        updated_at: Some(now),
        edit_generation: generation,
        peer_unreachable,
        deletions_recorded: false,
@@ -396,60 +365,6 @@ pub(crate) async fn enqueue_site_replication_retry_event_for_generation(
    }
 }
 /// Returns the number of peers whose snapshot entry is escalated and therefore
 /// will not carry this scheduling: the marker records a deletion that a
 /// snapshot cannot replay, and only a repair settles it, so clearing it to make
 /// the entry drainable again would drop that liability.
 pub(crate) fn record_iam_snapshot_retries(
    state: &mut SiteReplicationState,
    local_peer: &PeerInfo,
    reason: &str,
 ) -> S3Result<usize> {
    let peers = state
        .peers
        .values()
        .filter(|peer| {
            peer.deployment_id != local_peer.deployment_id && !same_identity_endpoint(&peer.endpoint, &local_peer.endpoint)
        })
        .cloned()
        .collect::<Vec<_>>();
    let mut escalated = 0usize;
    for peer in peers {
        if state.retry_queue.iter().any(|event| {
            retry_event_matches(event, &peer, SITE_REPLICATION_RETRY_IAM_SNAPSHOT_PATH)
                && event.last_error == SITE_REPLICATION_RETRY_SNAPSHOT_REPLAYED_MARKER
        }) {
            escalated += 1;
            continue;
        }
        ensure_site_replication_retry_event(&mut state.retry_queue, &peer, SITE_REPLICATION_RETRY_IAM_SNAPSHOT_PATH, reason)?;
    }
    Ok(escalated)
 }
 /// Schedule one collapsed full-IAM snapshot per remote peer after a bulk
 /// local mutation such as `import-iam`.
 pub(crate) async fn enqueue_site_replication_iam_snapshot(reason: &str) -> S3Result<()> {
    let state = load_site_replication_state().await?;
    if !state.enabled() {
        return Ok(());
    }
    let local_peer = current_local_runtime_peer(&state);
    let reason = reason.to_string();
    let escalated = update_site_replication_state(move |state| record_iam_snapshot_retries(state, &local_peer, &reason)).await?;
    if escalated > 0 {
        warn!(
            component = LOG_COMPONENT_ADMIN,
            subsystem = LOG_SUBSYSTEM_SITE_REPLICATION,
            event = EVENT_ADMIN_SITE_REPLICATION_STATE,
            escalated,
            result = "iam_snapshot_not_scheduled_for_escalated_peer",
            "site replication peers hold an escalated IAM entry; the snapshot waits for a repair"
        );
    }
    Ok(())
 }
 pub(crate) const SITE_REPLICATION_PEER_IAM_ITEM_WIRE_PATH: &str = "/rustfs/admin/v3/site-replication/peer/iam-item";
 /// Per-peer cap on recorded deletion bodies. Beyond it the peer's collapsed
@@ -168,8 +168,6 @@ mod rfc3339_map {
 pub(crate) struct PendingEndpointRefresh {
    pub(crate) id: String,
    pub(crate) peer: PeerInfo,
    #[serde(default, skip_serializing_if = "Option::is_none")]
    pub(crate) ilm_expiry_override: Option<bool>,
    #[serde(default, skip_serializing_if = "BTreeMap::is_empty")]
    pub(crate) remote_peers: BTreeMap<String, PeerInfo>,
    #[serde(default, skip_serializing_if = "BTreeSet::is_empty")]
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
overtrue	81038fc4e2	test(e2e): synchronize disk restore and heal log assertions	2026-09-09 13:27:18 +08:00
overtrue	6920abfe29	ci: retain full e2e recovery diagnostics	2026-09-09 12:31:04 +08:00
overtrue	f9913bbe72	fix(build): refresh cached git provenance for server builds	2026-09-09 12:31:03 +08:00
overtrue	d0a6260002	fix(tier): start config refresh after initial reload failure	2026-09-09 11:12:59 +08:00
overtrue	4b4360a1ca	test(e2e): preserve disk ownership through volume proxy	2026-09-09 10:23:59 +08:00
overtrue	9ae11bd3b5	merge: integrate release recovery and evidence updates	2026-09-09 10:23:19 +08:00
overtrue	3e64697d03	merge: integrate main into release validation # Conflicts: # .config/e2e-nightly-selection.txt # crates/e2e_test/src/inline_fast_path_cluster_test.rs # crates/ecstore/src/bucket/replication/replication_resyncer.rs # crates/kms/src/backends/local.rs	2026-09-08 20:52:53 +08:00
overtrue	9fa1d3f58f	fix(heal): recover pool metadata during ordinary healing	2026-09-08 20:36:34 +08:00
houseme	9aebcefa9c	test(scanner): harden measured ABBA evidence claims (#7491 ) Reject measured Scanner/Heal ABBA manifests whose mixed-version evidence uses the same baseline and candidate source revision or binary hash. Require crash fault modes and profile artifact names to match the exact supported sets, rejecting missing, duplicate, and unknown values. Update harness fixtures and regression coverage for same-build mixed-version claims and exact-set release evidence fields. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 20:32:52 +08:00
houseme	1748814bbf	test(scanner): bind profile artifacts in release evidence (#7487 ) Require Scanner/Heal release bundles to attach every required profiling artifact to P1 profile evidence with relative paths, artifact formats, non-empty files, hashes, and optional per-artifact measurement-window checks. Document the tightened release bundle profile contract and cover missing, tampered, and mismatched-window profile artifact regressions in the existing checker self-test. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 20:32:37 +08:00
houseme	ee5f76c180	fix(heal): publish committed MRF runtime checkpoints (#7490 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 20:32:22 +08:00
houseme	4d68c32b75	test(scanner): require mixed-version evidence roles (#7485 ) Tighten the Scanner/Heal release bundle checker so mixed-version, scoped-ACK, and rollback fields cannot reuse a generic versions list without proving the expected evidence role. Require version lists to use source revision identities and include the tested source revision. Also require profile evidence fields to name the core profiling artifacts before release approval. The release gate remains blocked until measured field evidence is present. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 20:07:11 +08:00
houseme	8c6689ff13	test(scanner): derive evidence runner profile from registry (#7484 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 19:27:23 +08:00
houseme	595f9f662d	test(ecstore): bind MRF manifest CAS dirsync recovery (#7482 ) Cover the committed MRF manifest path through LocalDisk conditional CAS when the metadata directory fsync fails. The fixture proves the previous manifest anchor survives rollback, an unanchored first successor is removed, and the legacy MRF journal remains readable even while global durability is relaxed. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 18:53:58 +08:00
houseme	4a2b15cb82	fix(heal): harden MRF replay boundaries (#7483 ) Reject journal records with unknown version-presence flags even when their CRC is valid, so rollback/future payloads cannot be accepted as known records. Gate committed checkpoint cleanup by the writer owner captured from the replay source, preserving retained manifests from other owners inside the same sequence window. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 18:53:43 +08:00
houseme	2ba7f95547	test(heal): write EC84 distributed restart oracle (#7481 ) Bind the distributed EC8+4 restart evidence lane to a scanner/heal oracle artifact so release validation can consume the real nextest run instead of accepting only a passing test. Require the registry to assert 8+4 erasure geometry for the three-node, four-drive case. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 18:15:46 +08:00
houseme	7b3dad6bae	test(heal): cover MRF snapshot torn successor recovery Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 17:28:09 +08:00
houseme	c0754f5b1c	test(heal): preserve EC84 restart semantics (#7478 ) Keep the EC8+4 background restart lane on the graceful-stop path and assert clean-restart marker absence only for restart scenarios. This prevents the hard evidence gate from silently exercising the crash path when it claims restart coverage. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 17:26:11 +08:00
overtrue	c03d3cdd59	fix: close replacement and protocol validation gaps	2026-09-08 17:23:53 +08:00
houseme	5dc0e3b402	heal: preserve merged result for duplicate submits Keep same-request-id replay receipts accepted for the receipt API, but preserve the legacy submit_heal_request duplicate admission result as Merged. Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 16:59:58 +08:00
houseme	cc5487e7de	heal: verify admin recreate pool metadata (#7474 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 16:57:13 +08:00
overtrue	590fab5c7e	test: stabilize release recovery and transition gates	2026-09-08 16:57:04 +08:00
overtrue	c02967baf6	fix: address confirmed release validation regressions	2026-09-08 16:42:41 +08:00
houseme	0b05b6c6ff	test(heal): cover quorum and mixed repair receipts Add focused oracles for transient quorum results that carry a matching receipt and for mixed grace plus repaired receipt batches. Only the repaired object may produce positive proof. Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 15:02:19 +08:00
houseme	c507da8f75	heal: verify replacement pool metadata repair (#7471 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 14:56:41 +08:00
houseme	550dabeffd	test(scanner): add EC8+4 heal restart evidence (#7469 ) Register and wire Scanner/Heal background target restart and crash evidence cases for the 3x4 EC8+4 topology. Validate the observed data/parity geometry in scanner-heal evidence receipts so multi-drive runs cannot satisfy the gate without proving EC8+4 metadata. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 14:52:07 +08:00
houseme	4d7f0344d3	test(heal): cover bucket object repair receipts (#7468 ) Cover bucket and root heal sweeps recording authoritative object outcomes only when storage receipts match the latched bucket incarnation. Verify unavailable or stale receipt ownership keeps object repair execution intact while leaving canonical outcome proof as Unknown. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 14:25:43 +08:00
houseme	084338add6	test(heal): reject failed object repair receipts Cover the receipt consumer path where a storage repair result carries both an error and a matching positive receipt. The failure may be recorded, but the receipt must not create repaired, healthy, or absent proof. Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 14:20:12 +08:00
houseme	8d339da706	heal: reject cancelled object repair receipts Do not record positive storage repair receipts once an object heal task has been cancelled, even if the receipt still matches the requested owner and object identity. Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 14:11:27 +08:00
houseme	817ad0a682	heal: reject dry-run object repair receipts Do not record positive storage repair receipts for dry-run object heal tasks, even if a producer accidentally returns a matching receipt. Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 14:04:15 +08:00
houseme	80321e5bb4	test(scanner): bind hard evidence bundle provenance (#7467 ) Require Scanner/Heal release bundle fields to carry source, run, window, timestamp, command, and artifact format provenance before a measured gate can pass. Keep EC8+4 and performance gate fields tied to a single measurement window so unrelated artifacts cannot be stitched into a release approval. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 14:03:37 +08:00
houseme	e7475cfa4d	heal: replay committed MRF checkpoints durably (#7465 ) Prefer committed MRF checkpoints during startup replay, retain accepted replay responsibilities until exact verified repair proofs arrive, and reclaim committed manifests only after discharge. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 14:02:24 +08:00
houseme	8c15025a5a	heal: latch object receipt owner before repair Capture the expected bucket incarnation before invoking object repair so a post-repair owner change cannot rewrite the responsibility that a storage receipt is allowed to prove. Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 13:48:29 +08:00
houseme	6f2ec66263	heal: require exact owner for object receipts Require object heal receipts to match the expected bucket incarnation before they can be recorded as positive repair evidence. This prevents stale or cross-incarnation receipts from clearing the wrong heal responsibility. Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 13:19:09 +08:00
houseme	62542ddc57	heal: reclaim superseded MRF snapshots after readback Add a committed snapshot predecessor cleanup primitive that deletes only an older slot after the successor is read back as the current committed snapshot. Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 13:09:39 +08:00
houseme	a8bb53218d	heal: retain MRF replay journal after accepted/merged replays Keep startup replay journal until a durable successor snapshot exists after Accepted/Merged admission. Co-Authored-By: heihutu <heihutu@gmail.com> Co-Authored-By: zhi22915 <qiuzgang@gmail.com>	2026-09-08 12:54:01 +08:00
houseme	5355d9f8f8	test(scanner): require complete ABBA summary matrix (#7462 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 12:16:38 +08:00
houseme	50d7a049ee	test(scanner): echo measured release evidence in ABBA fake adapter (#7460 ) Keep the synthetic ABBA test adapter aligned with the measured release evidence contract so result validation covers release_evidence drift. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 12:03:21 +08:00
houseme	3149c87cf2	test(heal): cover EC8+4 restart shard rebuild (#7461 ) Retry heal-control RPCs once after transport auth rejects a stale replay-scope epoch, and add a distributed EC8+4 restart heal evidence case that rebuilds a replaced drive with exact shard/body assertions. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 12:02:58 +08:00
houseme	b7fa6a4615	feat(heal): add MRF committed snapshot writer (#7458 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 11:33:09 +08:00
houseme	6fe83f87a4	test(scanner): require hard evidence ABBA manifest (#7459 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 11:32:34 +08:00
houseme	df6981d88e	test(scanner): gate release evidence bundles (#7457 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 11:32:04 +08:00
houseme	590adad5ae	fix(scanner): tag dirty usage producer identities (#7454 ) Record production-facing segment invalidation producer identities when existing object-level dirty usage hooks observe PUT, CopyObject, DeleteObject/DeleteMarker, and CompleteMultipartUpload mutations. Keep the data non-authoritative and process-local so segment reuse activation still requires durable generation-window proof. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 10:51:13 +08:00
houseme	be5d14c985	test(scanner): bind release evidence fields (#7452 ) Require explicit scanner/heal release evidence field contracts for the scoped ACK and mixed-version rollback gates. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 10:14:57 +08:00
houseme	703086d71d	test(scanner): require perf summary evidence fields (#7453 ) Fail closed when measured passing Scanner/Heal ABBA summaries omit W10/W11 foreground pressure, lock wait, or attempt-cost evidence. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 10:14:41 +08:00
houseme	a159f312f0	fix(heal): reuse same admission request id (#7451 ) Keep a retried heal start with the same request id from bypassing admission deduplication when the transport replay cache is unavailable but the manager still owns the task. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 10:14:34 +08:00
houseme	2f6f095298	test(scanner): cover flat-bucket quantum fairness (#7449 ) Add a production-entry scanner cohort regression that combines a wide flat bucket with a small bucket under a fixed object budget. The test keeps partial budgeted rounds unpublished, verifies small buckets are only marked after real execution, and confirms a later unbudgeted round can publish the complete aggregate. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 09:25:12 +08:00
cxymds	efd8ef005f	fix(ci): route replication read plan through boundary (#7448 )	2026-09-08 09:24:56 +08:00
houseme	60fa33773a	fix(common): retain unmatched MRF repair proofs (#7447 ) Add a recorded verified-repair consumer that discharges only exact retained anchors for the requested bucket while leaving unmatched proofs in the event ring. This gives the future durable successor writer a fail-closed primitive before any tombstone or GC path is enabled. Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 09:16:47 +08:00
houseme	ee752b0b03	test(scanner): summarize failed heal perf reports (#7446 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 09:16:39 +08:00
houseme	99c1f4418b	fix(scanner): expose recovery intent identity (#7445 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 09:16:33 +08:00
houseme	7ce0ac72cf	fix(scanner): require segment producer identities (#7444 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 09:15:53 +08:00
houseme	944e26d432	test(scanner): structure heal release evidence lanes (#7442 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 08:43:43 +08:00
Zhengchao An	f0b0a99260	fix(ci): enable release branch checks and repair test imports (#7441 )	2026-09-08 08:28:46 +08:00
houseme	33ddc10ffd	test(heal): cover MRF manifest CAS legacy transition (#7443 ) Co-authored-by: zhi22915 <qiuzgang@gmail.com>	2026-09-08 08:28:08 +08:00
`@@ -1,2 +1,2 @@`
	`sha256-darwin=f0c78fdb93471575d9a64c5c46eae6c806bdd0bc10a6e33d7fb574aabd8db5a3`	`sha256-darwin=f0d15f2d1183be319d9d977d20b48bb05a29a424c266eed910c67aaa6f1ff955`
	`sha256-linux=03ed7016cab672de9320e31375a0358eceacb4408b0e79cf063614fa7c878b87`	`sha256-linux=340aa702576ebed5266b7c47e591f267a11178fa04bf3828d64f53fe92eb0907`
`@@ -1,2 +1,2 @@`
	`sha256-darwin=364f2329a7b72eb9f1608dbe1a3af37af4095354014f3cbe23ca448492d89961`	`sha256-darwin=12d30fff5ed48fe95bbfb310dd507048f81782b2954e00eb72435708b1133f9c`
	`sha256-linux=60983f1ebe7068cf660d473c5f76c76a650410ccc99d71934ddca7fd67607987`	`sha256-linux=e917b2fdb303d01e6008ac4f1836c7b268698afe8383bd2aa4e697061df85cf6`