Merge branch 'main' into cxymds/fix-1925-heal-state-fence

.github/actions/schedule-failure-issue/action.yml

@@ -14,10 +14,9 @@
 
 name: "Schedule Failure Issue"
 description: >-
-  Open (or update) a tracking issue when a scheduled workflow run fails or
-  does not complete normally.
+  Open (or update) a tracking issue when a scheduled workflow run fails.
   Dedupes by workflow name: if an open issue titled
-  "[scheduled-failure] <workflow name>" already exists, the result is
+  "[scheduled-failure] <workflow name>" already exists, the failure is
   appended as a comment; otherwise a new issue is created. This is the
   single alerting mechanism for all scheduled pipelines (backlog#1149 ci-8).
 
@@ -39,26 +38,6 @@ inputs:
       Set to an empty string to skip labeling.
     required: false
     default: "infrastructure"
-  source-run-id:
-    description: "Run ID to report. Defaults to the current workflow run."
-    required: false
-    default: ${{ github.run_id }}
-  source-run-attempt:
-    description: "Run attempt to report. Defaults to the current attempt."
-    required: false
-    default: ${{ github.run_attempt }}
-  source-event:
-    description: "Trigger event of the run being reported."
-    required: false
-    default: ${{ github.event_name }}
-  source-ref-name:
-    description: "Ref name of the run being reported."
-    required: false
-    default: ${{ github.ref_name }}
-  source-sha:
-    description: "Commit SHA of the run being reported."
-    required: false
-    default: ${{ github.sha }}
 
 runs:
   using: "composite"
@@ -69,21 +48,17 @@ runs:
         GH_TOKEN: ${{ inputs.github-token }}
         WORKFLOW_NAME: ${{ inputs.workflow-name }}
         ISSUE_LABEL: ${{ inputs.label }}
-        SOURCE_RUN_ID: ${{ inputs.source-run-id }}
-        SOURCE_RUN_ATTEMPT: ${{ inputs.source-run-attempt }}
-        SOURCE_EVENT: ${{ inputs.source-event }}
-        SOURCE_REF_NAME: ${{ inputs.source-ref-name }}
-        SOURCE_SHA: ${{ inputs.source-sha }}
       run: |
         set -euo pipefail
 
         title="[scheduled-failure] ${WORKFLOW_NAME}"
-        run_url="${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}/actions/runs/${SOURCE_RUN_ID}"
+        run_url="${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}/actions/runs/${GITHUB_RUN_ID}"
 
-        # Inspect the reported run attempt. It can be the current in-workflow
-        # failure or a completed run observed by the external watchdog.
+        # Failed job names for this run attempt. The alert job runs while the
+        # run as a whole is still in progress, so inspect the jobs that have
+        # already completed with a non-success conclusion.
         failed_jobs="$(gh api \
-          "repos/${GITHUB_REPOSITORY}/actions/runs/${SOURCE_RUN_ID}/attempts/${SOURCE_RUN_ATTEMPT}/jobs" \
+          "repos/${GITHUB_REPOSITORY}/actions/runs/${GITHUB_RUN_ID}/attempts/${GITHUB_RUN_ATTEMPT}/jobs" \
           --paginate \
           --jq '.jobs[]
                 | select(.conclusion == "failure" or .conclusion == "timed_out" or .conclusion == "cancelled")
@@ -93,13 +68,13 @@ runs:
         fi
 
         body="$(cat <<EOF
-        Scheduled run of **${WORKFLOW_NAME}** did not complete successfully.
+        Scheduled run of **${WORKFLOW_NAME}** failed.
 
-        - Run: ${run_url} (attempt ${SOURCE_RUN_ATTEMPT})
-        - Event: \`${SOURCE_EVENT}\`
-        - Ref: \`${SOURCE_REF_NAME}\` @ \`${SOURCE_SHA}\`
+        - Run: ${run_url} (attempt ${GITHUB_RUN_ATTEMPT})
+        - Event: \`${GITHUB_EVENT_NAME}\`
+        - Ref: \`${GITHUB_REF_NAME}\` @ \`${GITHUB_SHA}\`
 
-        Non-success jobs:
+        Failed jobs:
         ${failed_jobs}
         EOF
         )"

.github/workflows/build.yml

@@ -1032,23 +1032,3 @@ jobs:
 
           echo "🎉 Released $TAG successfully!"
           echo "📄 Release URL: ${{ needs.create-release.outputs.release_url }}"
-
-  alert-on-failure:
-    name: Alert on scheduled failure
-    needs: [build-check, prepare-platform-matrix, build-rustfs, build-summary]
-    if: >-
-      always() && github.event_name == 'schedule' &&
-      (contains(needs.*.result, 'failure') || contains(needs.*.result, 'cancelled'))
-    runs-on: ubuntu-latest
-    timeout-minutes: 10
-    permissions:
-      contents: read
-      issues: write
-    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-        with:
-          persist-credentials: false
-      - name: Open or update failure-tracking issue
-        uses: ./.github/actions/schedule-failure-issue
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/ci.yml

@@ -1032,37 +1032,3 @@ jobs:
           path: artifacts/s3tests-single/**
           if-no-files-found: ignore
           retention-days: 3
-
-  alert-on-failure:
-    name: Alert on scheduled failure
-    needs:
-      - typos
-      - quick-checks
-      - test-and-lint
-      - test-ilm-integration-serial
-      - test-and-lint-rio-v2
-      - test-and-lint-protocols
-      - build-rustfs-debug-binary
-      - build-rustfs-debug-binary-rio-v2
-      - uring-integration
-      - e2e-tests
-      - e2e-full
-      - e2e-tests-rio-v2
-      - s3-implemented-tests
-      - s3-lifecycle-behavior-tests
-    if: >-
-      always() && github.event_name == 'schedule' &&
-      (contains(needs.*.result, 'failure') || contains(needs.*.result, 'cancelled'))
-    runs-on: ubuntu-latest
-    timeout-minutes: 10
-    permissions:
-      contents: read
-      issues: write
-    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-        with:
-          persist-credentials: false
-      - name: Open or update failure-tracking issue
-        uses: ./.github/actions/schedule-failure-issue
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/minio-interop.yml

@@ -121,21 +121,3 @@ jobs:
           cargo nextest run --run-ignored ignored-only --no-tests=fail \
             -p "$INTEROP_PACKAGE" --features "$INTEROP_FEATURES" \
             -E "$INTEROP_FILTER"
-
-  alert-on-failure:
-    name: Alert on scheduled failure
-    needs: [minio-interop]
-    if: always() && github.event_name == 'schedule' && contains(needs.*.result, 'failure')
-    runs-on: ubuntu-latest
-    timeout-minutes: 10
-    permissions:
-      contents: read
-      issues: write
-    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-        with:
-          persist-credentials: false
-      - name: Open or update failure-tracking issue
-        uses: ./.github/actions/schedule-failure-issue
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/nightly-gnu.yml

@@ -194,23 +194,3 @@ jobs:
 
       - name: Run HA leader failover live checks (three-node Raft cluster in Docker)
         run: bash scripts/test/vault_ha_kms_live.sh
-
-  alert-on-failure:
-    name: Alert on scheduled failure
-    needs: [build, kms-vault-lane, kms-vault-ha-failover]
-    if: >-
-      always() && github.event_name == 'schedule' &&
-      (contains(needs.*.result, 'failure') || contains(needs.*.result, 'cancelled'))
-    runs-on: ubuntu-latest
-    timeout-minutes: 10
-    permissions:
-      contents: read
-      issues: write
-    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-        with:
-          persist-credentials: false
-      - name: Open or update failure-tracking issue
-        uses: ./.github/actions/schedule-failure-issue
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/scheduled-validation-watchdog.yml

@@ -1,63 +0,0 @@
-# Copyright 2024 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.
-
-name: Scheduled Validation Watchdog
-
-on:
-  workflow_run:
-    workflows:
-      - "Security Audit"
-      - "Build and Release"
-      - "Continuous Integration"
-      - "coverage"
-      - "e2e-nightly"
-      - "e2e-s3tests"
-      - "Fuzz"
-      - "mint"
-      - "minio-interop"
-      - "Nightly GNU Build"
-      - "Performance A/B"
-      - "Runner Hygiene"
-    types: [completed]
-
-permissions:
-  contents: read
-
-jobs:
-  alert-on-incomplete-run:
-    name: Alert on incomplete scheduled run
-    if: >-
-      github.event.workflow_run.event == 'schedule' &&
-      github.event.workflow_run.conclusion != 'success' &&
-      github.event.workflow_run.conclusion != 'failure'
-    runs-on: ubuntu-latest
-    timeout-minutes: 10
-    permissions:
-      actions: read
-      contents: read
-      issues: write
-    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-        with:
-          persist-credentials: false
-      - name: Open or update incomplete-run issue
-        uses: ./.github/actions/schedule-failure-issue
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-          workflow-name: ${{ github.event.workflow_run.name }}
-          source-run-id: ${{ github.event.workflow_run.id }}
-          source-run-attempt: ${{ github.event.workflow_run.run_attempt }}
-          source-event: ${{ github.event.workflow_run.event }}
-          source-ref-name: ${{ github.event.workflow_run.head_branch }}
-          source-sha: ${{ github.event.workflow_run.head_sha }}

crates/config/src/constants/runtime.rs

@@ -57,6 +57,13 @@ pub const DEFAULT_MAX_IO_EVENTS_PER_TICK: usize = 1024;
 pub const DEFAULT_EVENT_INTERVAL: u32 = 61;
 pub const DEFAULT_RNG_SEED: Option<u64> = None; // None means random
 
+/// Dedicated blocking thread pool for fsync/fdatasync operations.
+/// When > 1, fsync operations are isolated from the main blocking pool to
+/// prevent device-bound fsync from starving read operations (pread/stat/open).
+/// Default 0 means auto (no isolation, use main runtime).
+pub const ENV_FSYNC_BLOCKING_THREADS: &str = "RUSTFS_RUNTIME_FSYNC_BLOCKING_THREADS";
+pub const DEFAULT_FSYNC_BLOCKING_THREADS: usize = 0;
+
 // Dial9 Tokio Telemetry Default values
 pub const DEFAULT_RUNTIME_DIAL9_ENABLED: bool = false; // Disabled by default
 pub const DEFAULT_RUNTIME_DIAL9_OUTPUT_DIR: &str = "/var/log/rustfs/telemetry";

crates/ecstore/src/core/pools.rs

@@ -1675,7 +1675,7 @@ impl PoolMeta {
         self.load_no_lock(pool).await
     }
 
-    async fn load_no_lock<S>(&mut self, pool: Arc<S>) -> Result<()>
+    pub(crate) async fn load_no_lock<S>(&mut self, pool: Arc<S>) -> Result<()>
     where
         S: EcstoreObjectIO,
     {

crates/ecstore/src/disk/os.rs

@@ -315,7 +315,7 @@ pub async fn fsync_dir(dir: impl AsRef<Path>) -> io::Result<()> {
     #[cfg(unix)]
     {
         let dir = dir.as_ref().to_path_buf();
-        tokio::task::spawn_blocking(move || fsync_dir_std(dir)).await?
+        fsync_spawn_blocking(move || fsync_dir_std(dir)).await?
     }
 
     #[cfg(not(unix))]
@@ -683,7 +683,7 @@ async fn fsync_open_dst_dir_group(group: &DstDirFsyncGroup) -> io::Result<()> {
     #[cfg(test)]
     let dir = group.dir.clone();
     let dir_file = group.dir_file.clone();
-    tokio::task::spawn_blocking(move || {
+    fsync_spawn_blocking(move || {
         #[cfg(test)]
         {
             if let Some(kind) = fsync_dir_recorder::take_grouped_failure(&dir) {
@@ -1080,6 +1080,44 @@ const TEST_GLOBAL_FILE_SYNCS: usize = 64;
 
 static FILE_SYNC_PERMITS: LazyLock<Semaphore> = LazyLock::new(|| Semaphore::new(global_file_sync_limit()));
 static DISK_FILE_SYNC_LIMITERS: LazyLock<Mutex<HashMap<PathBuf, Weak<Semaphore>>>> = LazyLock::new(|| Mutex::new(HashMap::new()));
+
+/// Dedicated tokio runtime for fsync/fdatasync blocking operations. When
+/// configured with >1 threads, isolates device-bound fsync from the main
+/// blocking pool so reads (pread/stat/open) are not starved. `None` means
+/// fall back to the main runtime (zero behavior change).
+static FSYNC_RUNTIME: LazyLock<Option<tokio::runtime::Runtime>> = LazyLock::new(|| {
+    let threads =
+        rustfs_utils::get_env_usize(rustfs_config::ENV_FSYNC_BLOCKING_THREADS, rustfs_config::DEFAULT_FSYNC_BLOCKING_THREADS);
+    if threads <= 1 {
+        return None;
+    }
+    let mut builder = tokio::runtime::Builder::new_multi_thread();
+    builder
+        .worker_threads(num_cpus::get().min(8))
+        .max_blocking_threads(threads)
+        .thread_name("rustfs-fsync")
+        .thread_stack_size(512 * 1024)
+        .enable_all();
+    match builder.build() {
+        Ok(rt) => {
+            tracing::info!(threads, "fsync dedicated blocking pool enabled");
+            Some(rt)
+        }
+        Err(err) => {
+            tracing::warn!(%err, "failed to build fsync runtime, falling back to main pool");
+            None
+        }
+    }
+});
+
+/// Spawn a blocking task on the fsync-dedicated runtime if configured,
+/// otherwise fall back to the main tokio blocking pool.
+fn fsync_spawn_blocking<T: Send + 'static>(f: impl FnOnce() -> T + Send + 'static) -> tokio::task::JoinHandle<T> {
+    match FSYNC_RUNTIME.as_ref() {
+        Some(rt) => rt.spawn_blocking(f),
+        None => tokio::task::spawn_blocking(f),
+    }
+}
 static DISK_VOLUME_MUTATION_LOCKS: LazyLock<Mutex<HashMap<PathBuf, Weak<RwLock<()>>>>> =
     LazyLock::new(|| Mutex::new(HashMap::new()));
 type NamespaceMutationLock = AsyncMutex<()>;
@@ -1217,7 +1255,7 @@ where
     F: FnOnce() -> io::Result<T> + Send + 'static,
 {
     let (disk_permit, global_permit) = acquire_file_sync_permits(disk_permits).await?;
-    let result = tokio::task::spawn_blocking(move || {
+    let result = fsync_spawn_blocking(move || {
         let _disk_permit = disk_permit;
         work()
     })
@@ -2146,7 +2184,7 @@ async fn run_blocking_namespace_file_sync_operation_with_global<T: Send + 'stati
         wait_started,
     );
     let disk_permit = admission.disk_permit.clone();
-    let result = tokio::task::spawn_blocking(move || {
+    let result = fsync_spawn_blocking(move || {
         let _lease = lease;
         let _disk_permit = disk_permit;
         operation()

crates/ecstore/src/store/heal.rs

@@ -13,7 +13,12 @@
 // limitations under the License.
 
 use super::*;
+use crate::core::pools::POOL_META_NAME;
+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_lock::NamespaceLockGuard;
 use tracing::trace;
 
 const LOG_COMPONENT_ECSTORE: &str = "ecstore";
@@ -30,7 +35,119 @@ fn invalid_heal_pool_index(pool_idx: usize, pool_count: usize) -> Error {
     )
 }
 
+#[derive(Debug, Clone, Copy)]
+enum HealFormatPoolSkip {
+    Completed,
+    Retryable,
+}
+
+fn classify_heal_format_pool(
+    pool_idx: usize,
+    pool_cmd_line: &str,
+    pool_meta: &PoolMeta,
+    rebalance_meta: Option<&RebalanceMeta>,
+) -> Option<HealFormatPoolSkip> {
+    let Some(pool) = pool_meta.pools.get(pool_idx) else {
+        return Some(HealFormatPoolSkip::Retryable);
+    };
+
+    if pool.id != pool_idx || pool_cmd_line.is_empty() || pool.cmd_line.is_empty() || pool.cmd_line != pool_cmd_line {
+        return Some(HealFormatPoolSkip::Retryable);
+    }
+
+    if let Some(decommission) = pool.decommission.as_ref() {
+        if decommission.complete {
+            return Some(HealFormatPoolSkip::Completed);
+        }
+        if decommission.failed || decommission.canceled || decommission.queued || pool_meta.is_suspended(pool_idx) {
+            return Some(HealFormatPoolSkip::Retryable);
+        }
+    }
+
+    if let Some(meta) = rebalance_meta {
+        let Some(pool_stats) = meta.pool_stats.get(pool_idx) else {
+            return Some(HealFormatPoolSkip::Retryable);
+        };
+        if pool_stats.info.stopping || (pool_stats.participating && pool_stats.info.status == RebalStatus::Started) {
+            return Some(HealFormatPoolSkip::Retryable);
+        }
+    }
+
+    None
+}
+
+fn heal_format_pool_skip_error(skip: HealFormatPoolSkip) -> Error {
+    match skip {
+        HealFormatPoolSkip::Completed => StorageError::NoHealRequired,
+        HealFormatPoolSkip::Retryable => StorageError::SlowDown,
+    }
+}
+
+fn heal_format_fence_lost_error() -> Error {
+    StorageError::SlowDown
+}
+
 impl ECStore {
+    async fn acquire_heal_format_fence(
+        &self,
+    ) -> Result<(NamespaceLockGuard, NamespaceLockGuard, PoolMeta, Option<RebalanceMeta>)> {
+        let metadata_pool = self
+            .pools
+            .first()
+            .cloned()
+            .ok_or_else(|| Error::other("heal format requires at least one storage pool"))?;
+
+        // Metadata fence order is part of the decommission/rebalance protocol:
+        // pool.bin must always be acquired before rebalance.bin.
+        let pool_lock = metadata_pool.new_ns_lock(RUSTFS_META_BUCKET, POOL_META_NAME).await?;
+        let pool_guard = pool_lock.get_write_lock(get_lock_acquire_timeout()).await?;
+        let rebalance_lock = metadata_pool.new_ns_lock(RUSTFS_META_BUCKET, REBAL_META_NAME).await?;
+        let rebalance_guard = rebalance_lock.get_write_lock(get_lock_acquire_timeout()).await?;
+
+        if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
+            return Err(heal_format_fence_lost_error());
+        }
+
+        let mut pool_meta = PoolMeta::default();
+        pool_meta.load_no_lock(metadata_pool.clone()).await?;
+        if pool_meta.pools.len() != self.pools.len()
+            || pool_meta.pools.iter().enumerate().any(|(pool_idx, pool)| {
+                pool.id != pool_idx || pool.cmd_line.is_empty() || pool.cmd_line != self.pools[pool_idx].endpoints.cmd_line
+            })
+        {
+            return Err(heal_format_fence_lost_error());
+        }
+
+        let mut rebalance_meta = RebalanceMeta::new();
+        let rebalance_meta = match rebalance_meta
+            .load_with_opts(
+                metadata_pool,
+                ObjectOptions {
+                    no_lock: true,
+                    ..Default::default()
+                },
+            )
+            .await
+        {
+            Ok(()) => Some(rebalance_meta),
+            Err(Error::ConfigNotFound) => None,
+            Err(err) => return Err(err),
+        };
+
+        if rebalance_meta
+            .as_ref()
+            .is_some_and(|meta| meta.pool_stats.len() != self.pools.len())
+        {
+            return Err(heal_format_fence_lost_error());
+        }
+
+        if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
+            return Err(heal_format_fence_lost_error());
+        }
+
+        Ok((pool_guard, rebalance_guard, pool_meta, rebalance_meta))
+    }
+
     fn get_pools_for_heal_object(&self, opts: &HealOpts) -> Result<Vec<Arc<Sets>>> {
         match opts.pool {
             Some(pool_idx) => Ok(vec![
@@ -52,8 +169,24 @@ impl ECStore {
         };
 
         let mut count_no_heal = 0;
+        let mut count_completed = 0;
         let mut first_error = None;
-        for pool in self.pools.iter() {
+        for (pool_idx, pool) in self.pools.iter().enumerate() {
+            let (pool_guard, rebalance_guard, pool_meta, rebalance_meta) = self.acquire_heal_format_fence().await?;
+            if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
+                first_error.get_or_insert(heal_format_fence_lost_error());
+                break;
+            }
+            if let Some(skip) = classify_heal_format_pool(pool_idx, &pool.endpoints.cmd_line, &pool_meta, rebalance_meta.as_ref())
+            {
+                if matches!(skip, HealFormatPoolSkip::Completed) {
+                    count_completed += 1;
+                } else {
+                    first_error.get_or_insert(heal_format_pool_skip_error(skip));
+                }
+                continue;
+            }
+
             let (mut result, err) = pool.heal_format(dry_run).await?;
             if let Some(err) = err {
                 match err {
@@ -69,11 +202,18 @@ impl ECStore {
             r.set_count += result.set_count;
             r.before.drives.append(&mut result.before.drives);
             r.after.drives.append(&mut result.after.drives);
+
+            // Sets::heal_format cannot observe this guard before each disk write;
+            // fail closed after the call if the lease was lost during format IO.
+            if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
+                first_error.get_or_insert(heal_format_fence_lost_error());
+                break;
+            }
         }
         if let Some(err) = first_error {
             return Ok((r, Some(err)));
         }
-        if count_no_heal == self.pools.len() {
+        if count_no_heal + count_completed == self.pools.len() {
             info!(
                 event = EVENT_HEAL_FORMAT_COMPLETED,
                 component = LOG_COMPONENT_ECSTORE,
@@ -300,6 +440,7 @@ mod tests {
     use crate::core::pools::{PoolDecommissionInfo, PoolStatus};
     use crate::disk::{DiskOption, format::FormatV3, new_disk};
     use crate::layout::endpoints::{Endpoints, PoolEndpoints};
+    use crate::services::rebalance::{RebalanceInfo, RebalanceStats};
     use crate::store::init_format::{load_format_erasure, save_format_file};
 
     async fn minimal_heal_pool(pool_idx: usize) -> Arc<Sets> {
@@ -347,6 +488,164 @@ mod tests {
         }
     }
 
+    fn pool_meta_with_decommission(info: PoolDecommissionInfo) -> PoolMeta {
+        PoolMeta {
+            pools: vec![PoolStatus {
+                id: 0,
+                cmd_line: "pool-0".to_string(),
+                last_update: OffsetDateTime::UNIX_EPOCH,
+                decommission: Some(info),
+            }],
+            ..Default::default()
+        }
+    }
+
+    #[test]
+    fn heal_format_pool_state_barriers_are_classified() {
+        let active = pool_meta_with_decommission(PoolDecommissionInfo {
+            start_time: Some(OffsetDateTime::UNIX_EPOCH),
+            ..Default::default()
+        });
+        assert!(matches!(
+            classify_heal_format_pool(0, "pool-0", &active, None),
+            Some(HealFormatPoolSkip::Retryable)
+        ));
+
+        for info in [
+            PoolDecommissionInfo {
+                failed: true,
+                ..Default::default()
+            },
+            PoolDecommissionInfo {
+                canceled: true,
+                ..Default::default()
+            },
+        ] {
+            assert!(matches!(
+                classify_heal_format_pool(0, "pool-0", &pool_meta_with_decommission(info), None),
+                Some(HealFormatPoolSkip::Retryable)
+            ));
+        }
+
+        let completed = pool_meta_with_decommission(PoolDecommissionInfo {
+            complete: true,
+            ..Default::default()
+        });
+        assert!(matches!(
+            classify_heal_format_pool(0, "pool-0", &completed, None),
+            Some(HealFormatPoolSkip::Completed)
+        ));
+    }
+
+    #[test]
+    fn heal_format_pool_rebalance_barriers_and_identity_are_fail_closed() {
+        let identity_meta = pool_meta_with_decommission(PoolDecommissionInfo::default());
+        let rebalance = RebalanceMeta {
+            pool_stats: vec![RebalanceStats {
+                participating: true,
+                info: RebalanceInfo {
+                    status: RebalStatus::Started,
+                    ..Default::default()
+                },
+                ..Default::default()
+            }],
+            ..Default::default()
+        };
+        assert!(matches!(
+            classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&rebalance)),
+            Some(HealFormatPoolSkip::Retryable)
+        ));
+
+        let stopping = RebalanceMeta {
+            pool_stats: vec![RebalanceStats {
+                info: RebalanceInfo {
+                    stopping: true,
+                    ..Default::default()
+                },
+                ..Default::default()
+            }],
+            ..Default::default()
+        };
+        assert!(matches!(
+            classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&stopping)),
+            Some(HealFormatPoolSkip::Retryable)
+        ));
+
+        let identity = pool_meta_with_decommission(PoolDecommissionInfo::default());
+        assert!(matches!(
+            classify_heal_format_pool(0, "pool-new", &identity, None),
+            Some(HealFormatPoolSkip::Retryable)
+        ));
+
+        let identity_without_decommission = PoolMeta {
+            pools: vec![PoolStatus {
+                id: 0,
+                cmd_line: "pool-0".to_string(),
+                last_update: OffsetDateTime::UNIX_EPOCH,
+                decommission: None,
+            }],
+            ..Default::default()
+        };
+        assert!(matches!(
+            classify_heal_format_pool(0, "pool-new", &identity_without_decommission, None),
+            Some(HealFormatPoolSkip::Retryable)
+        ));
+
+        assert!(matches!(
+            classify_heal_format_pool(0, "", &identity_meta, None),
+            Some(HealFormatPoolSkip::Retryable)
+        ));
+
+        assert!(matches!(
+            classify_heal_format_pool(0, "pool-0", &PoolMeta::default(), None),
+            Some(HealFormatPoolSkip::Retryable)
+        ));
+
+        let stopped = RebalanceMeta {
+            stopped_at: Some(OffsetDateTime::UNIX_EPOCH),
+            pool_stats: vec![RebalanceStats {
+                participating: true,
+                info: RebalanceInfo {
+                    status: RebalStatus::Stopped,
+                    ..Default::default()
+                },
+                ..Default::default()
+            }],
+            ..Default::default()
+        };
+        assert!(classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&stopped)).is_none());
+
+        let stopping_after_stop = RebalanceMeta {
+            stopped_at: Some(OffsetDateTime::UNIX_EPOCH),
+            pool_stats: vec![RebalanceStats {
+                participating: true,
+                info: RebalanceInfo {
+                    status: RebalStatus::Started,
+                    stopping: true,
+                    ..Default::default()
+                },
+                ..Default::default()
+            }],
+            ..Default::default()
+        };
+        assert!(matches!(
+            classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&stopping_after_stop)),
+            Some(HealFormatPoolSkip::Retryable)
+        ));
+    }
+
+    #[test]
+    fn skipped_heal_format_pool_is_never_reported_as_success() {
+        assert!(matches!(
+            heal_format_pool_skip_error(HealFormatPoolSkip::Retryable),
+            StorageError::SlowDown
+        ));
+        assert!(matches!(
+            heal_format_pool_skip_error(HealFormatPoolSkip::Completed),
+            StorageError::NoHealRequired
+        ));
+    }
+
     #[tokio::test]
     async fn heal_object_pool_scope_selects_only_requested_pool() {
         let store = minimal_heal_store().await;
@@ -615,6 +914,18 @@ mod tests {
             bucket_fence_registry: std::sync::Arc::default(),
         };
 
+        let err = store
+            .handle_heal_format(false)
+            .await
+            .expect_err("missing pool metadata must fail closed before format writes");
+        assert!(matches!(err, StorageError::SlowDown));
+
+        let pool_meta = PoolMeta::new(&store.pools, &PoolMeta::default());
+        pool_meta
+            .save(store.pools.clone())
+            .await
+            .expect("pool metadata should be persisted before format heal");
+
         let (result, err) = store
             .handle_heal_format(false)
             .await
@@ -628,5 +939,22 @@ mod tests {
             .await
             .expect("the later pool should be healed despite the first pool error");
         assert_eq!(healed.erasure.this, recoverable_format.erasure.sets[0][2]);
+
+        let mut completed_meta = PoolMeta::new(&store.pools, &PoolMeta::default());
+        for status in &mut completed_meta.pools {
+            status.decommission = Some(PoolDecommissionInfo {
+                complete: true,
+                ..Default::default()
+            });
+        }
+        completed_meta
+            .save(store.pools.clone())
+            .await
+            .expect("completed pool metadata should be persisted");
+        let (_, err) = store
+            .handle_heal_format(false)
+            .await
+            .expect("completed pools should be reported as a no-op");
+        assert!(matches!(err, Some(StorageError::NoHealRequired)));
     }
 }

crates/heal/src/heal/task/heal_erasure_set.rs

@@ -231,6 +231,10 @@ impl HealTask {
                             "Heal erasure set format repair skipped because no format heal was required"
                         );
                     } else {
+                        let error = e;
+                        if error.is_recoverable_heal() {
+                            return Err(error);
+                        }
                         error!(
                             target: "rustfs::heal::task",
                             event = EVENT_HEAL_ERASURE_SET_RESULT,
@@ -239,7 +243,7 @@ impl HealTask {
                             task_id = %self.id,
                             set_disk_id,
                             result = "format_failed",
-                            error = %e,
+                            error = %error,
                             "Heal erasure set failed"
                         );
                         {
@@ -247,7 +251,7 @@ impl HealTask {
                             progress.update_progress(4, 4, 0, 0);
                         }
                         return Err(Error::TaskExecutionFailed {
-                            message: format!("Failed to heal disk format for {set_disk_id}: {e}"),
+                            message: format!("Failed to heal disk format for {set_disk_id}: {error}"),
                         });
                     }
                 } else {
@@ -284,6 +288,9 @@ impl HealTask {
             Err(Error::TaskCancelled) => return Err(Error::TaskCancelled),
             Err(Error::TaskTimeout) => return Err(Error::TaskTimeout),
             Err(e) => {
+                if e.is_recoverable_heal() {
+                    return Err(e);
+                }
                 error!(
                     target: "rustfs::heal::task",
                     event = EVENT_HEAL_ERASURE_SET_RESULT,

crates/heal/src/heal/task/tests.rs

@@ -547,6 +547,7 @@ struct MockStorage {
     heal_object_outcome: Mutex<Option<MockHealObjectOutcome>>,
     heal_object_outcomes: Mutex<HashMap<String, VecDeque<MockHealObjectOutcome>>>,
     format_no_heal_required: Mutex<bool>,
+    format_error: Mutex<Option<Error>>,
     global_format_calls: Mutex<u32>,
     replacement_format_calls: Mutex<Vec<(usize, usize, Vec<String>)>>,
     replacement_targets_ready: Mutex<bool>,
@@ -867,6 +868,9 @@ impl HealStorageAPI for MockStorage {
 
     async fn heal_format(&self, _dry_run: bool) -> Result<(HealResultItem, Option<Error>)> {
         *self.global_format_calls.lock().unwrap() += 1;
+        if let Some(error) = self.format_error.lock().unwrap().take() {
+            return Err(error);
+        }
         let no_heal_required = *self.format_no_heal_required.lock().unwrap();
         if no_heal_required {
             Ok((HealResultItem::default(), Some(Error::Storage(EcstoreError::NoHealRequired))))
@@ -2052,6 +2056,30 @@ async fn test_erasure_set_heal_continues_after_format_no_heal_required() {
     );
 }
 
+#[tokio::test]
+async fn erasure_set_format_slowdown_is_propagated() {
+    let storage = Arc::new(MockStorage {
+        format_error: Mutex::new(Some(Error::Storage(EcstoreError::SlowDown))),
+        ..Default::default()
+    });
+    let request = HealRequest::new(
+        HealType::ErasureSet {
+            buckets: Vec::new(),
+            set_disk_id: "pool_0_set_0".to_string(),
+        },
+        HealOptions::default(),
+        HealPriority::Normal,
+    );
+    let task = HealTask::from_request(request, storage);
+
+    let error = task
+        .execute()
+        .await
+        .expect_err("format SlowDown must remain recoverable for the task manager");
+
+    assert!(matches!(error, Error::Storage(EcstoreError::SlowDown)));
+}
+
 #[tokio::test]
 async fn erasure_set_bucket_prepass_failure_stops_before_object_heal() {
     let temp = TempDir::new().expect("temporary directory should be created");

crates/test-utils/src/lib.rs

@@ -245,6 +245,18 @@ impl TestECStoreEnvBuilder {
             .await
             .expect("build test ECStore");
 
+        // The production bootstrap only persists pool.bin from the elected
+        // first cluster node.  Test stores intentionally have no cluster
+        // election, but heal-format still requires that durable fence before
+        // it can write any disk format.  Materialize the validated topology
+        // here so the shared fixture models a ready single-node store.
+        let mut pool_meta = ecstore.pool_meta.read().await.clone();
+        pool_meta.dont_save = false;
+        pool_meta
+            .save(ecstore.pools.clone())
+            .await
+            .expect("persist test pool metadata");
+
         if self.init_bucket_metadata {
             let buckets_list = ecstore
                 .list_bucket(&BucketOptions {

scripts/check_test_wiring.py

@@ -15,20 +15,6 @@ from pathlib import Path
 
 
 ROOT = Path(__file__).resolve().parents[1]
-SCHEDULED_ALERT_WORKFLOWS = (
-    ".github/workflows/audit.yml",
-    ".github/workflows/build.yml",
-    ".github/workflows/ci.yml",
-    ".github/workflows/coverage.yml",
-    ".github/workflows/e2e-replication-nightly.yml",
-    ".github/workflows/e2e-s3tests.yml",
-    ".github/workflows/fuzz.yml",
-    ".github/workflows/mint.yml",
-    ".github/workflows/minio-interop.yml",
-    ".github/workflows/nightly-gnu.yml",
-    ".github/workflows/performance-ab.yml",
-    ".github/workflows/runner-hygiene.yml",
-)
 
 
 def words(value: str) -> set[str]:
@@ -266,79 +252,6 @@ def check_profile_definitions(root: Path) -> list[str]:
     return errors
 
 
-def check_scheduled_alerts(root: Path) -> list[str]:
-    errors: list[str] = []
-    for relative in SCHEDULED_ALERT_WORKFLOWS:
-        path = root / relative
-        try:
-            lines = path.read_text().splitlines()
-        except FileNotFoundError:
-            errors.append(f"{relative}: missing scheduled validation workflow")
-            continue
-
-        try:
-            start = lines.index("  alert-on-failure:") + 1
-        except ValueError:
-            errors.append(f"{relative}: missing alert-on-failure job")
-            continue
-        end = next(
-            (index for index in range(start, len(lines)) if re.fullmatch(r"  [A-Za-z0-9_-]+:", lines[index])),
-            len(lines),
-        )
-        job = "\n".join(line.split("#", 1)[0] for line in lines[start:end])
-        required = (
-            "always()",
-            "github.event_name == 'schedule'",
-            "contains(needs.*.result, 'failure')",
-            "issues: write",
-            "uses: ./.github/actions/schedule-failure-issue",
-            "github-token: ${{ secrets.GITHUB_TOKEN }}",
-        )
-        missing = [token for token in required if token not in job]
-        if missing:
-            errors.append(f"{relative}: alert-on-failure missing {', '.join(missing)}")
-
-    watchdog_path = root / ".github/workflows/scheduled-validation-watchdog.yml"
-    try:
-        watchdog = "\n".join(
-            line.split("#", 1)[0] for line in watchdog_path.read_text().splitlines()
-        )
-    except FileNotFoundError:
-        errors.append(".github/workflows/scheduled-validation-watchdog.yml: missing completion watchdog")
-        return errors
-    for relative in SCHEDULED_ALERT_WORKFLOWS:
-        path = root / relative
-        if not path.is_file():
-            continue
-        source = path.read_text()
-        match = re.search(r"^name:\s*[\"']?([^\"'\n]+)", source, re.MULTILINE)
-        if not match:
-            errors.append(f"{relative}: missing workflow name")
-        elif f'- "{match.group(1).strip()}"' not in watchdog:
-            errors.append(f"{relative}: missing from scheduled completion watchdog")
-    required = (
-        "github.event.workflow_run.event == 'schedule'",
-        "github.event.workflow_run.conclusion != 'success'",
-        "github.event.workflow_run.conclusion != 'failure'",
-        "actions: read",
-        "issues: write",
-        "uses: ./.github/actions/schedule-failure-issue",
-        "github-token: ${{ secrets.GITHUB_TOKEN }}",
-        "workflow-name: ${{ github.event.workflow_run.name }}",
-        "source-event: ${{ github.event.workflow_run.event }}",
-        "source-ref-name: ${{ github.event.workflow_run.head_branch }}",
-        "source-sha: ${{ github.event.workflow_run.head_sha }}",
-        "source-run-id: ${{ github.event.workflow_run.id }}",
-        "source-run-attempt: ${{ github.event.workflow_run.run_attempt }}",
-    )
-    missing = [token for token in required if token not in watchdog]
-    if missing:
-        errors.append(
-            ".github/workflows/scheduled-validation-watchdog.yml: missing " + ", ".join(missing)
-        )
-    return errors
-
-
 def check_profile_listing(root: Path, profile: str, listing: Path) -> list[str]:
     try:
         expected_digest = profile_selection(root, profile)
@@ -368,7 +281,6 @@ def validate(root: Path) -> list[str]:
     errors.extend(check_runner_selection(root))
     errors.extend(check_s3_tests_runner(root))
     errors.extend(check_profile_definitions(root))
-    errors.extend(check_scheduled_alerts(root))
     return errors
 
 
@@ -451,7 +363,6 @@ class SelfTests(unittest.TestCase):
                 mock.patch(__name__ + ".check_fuzz_targets", return_value=[]),
                 mock.patch(__name__ + ".check_runner_selection", return_value=[]),
                 mock.patch(__name__ + ".check_profile_definitions", return_value=[]),
-                mock.patch(__name__ + ".check_scheduled_alerts", return_value=[]),
             ):
                 self.assertEqual(len(validate(root)), 1)
 
@@ -502,75 +413,6 @@ class SelfTests(unittest.TestCase):
             with mock.patch.object(sys, "platform", "linux"):
                 self.assertEqual(len(check_profile_listing(root, "e2e-full", listing)), 1)
 
-    def test_scheduled_alerts_require_completion_watchdog(self) -> None:
-        with tempfile.TemporaryDirectory() as tmp:
-            root = Path(tmp)
-            alert = (
-                "  alert-on-failure:\n"
-                "    if: always() && github.event_name == 'schedule' && "
-                "contains(needs.*.result, 'failure')\n"
-                "    permissions:\n"
-                "      issues: write\n"
-                "    steps:\n"
-                "      - uses: ./.github/actions/schedule-failure-issue\n"
-                "        with:\n"
-                "          github-token: ${{ secrets.GITHUB_TOKEN }}\n"
-            )
-            names: list[str] = []
-            for relative in SCHEDULED_ALERT_WORKFLOWS:
-                path = root / relative
-                path.parent.mkdir(parents=True, exist_ok=True)
-                names.append(path.stem)
-                path.write_text(f'name: "{path.stem}"\n{alert}')
-            watchdog = root / ".github/workflows/scheduled-validation-watchdog.yml"
-            watchdog_source = (
-                "\n".join(f'- "{name}"' for name in names)
-                + "\ngithub.event.workflow_run.event == 'schedule'\n"
-                + "github.event.workflow_run.conclusion != 'success'\n"
-                + "github.event.workflow_run.conclusion != 'failure'\n"
-                + "actions: read\n"
-                + "issues: write\n"
-                + "uses: ./.github/actions/schedule-failure-issue\n"
-                + "github-token: ${{ secrets.GITHUB_TOKEN }}\n"
-                + "workflow-name: ${{ github.event.workflow_run.name }}\n"
-                + "source-event: ${{ github.event.workflow_run.event }}\n"
-                + "source-ref-name: ${{ github.event.workflow_run.head_branch }}\n"
-                + "source-sha: ${{ github.event.workflow_run.head_sha }}\n"
-                + "source-run-id: ${{ github.event.workflow_run.id }}\n"
-                + "source-run-attempt: ${{ github.event.workflow_run.run_attempt }}\n"
-            )
-            watchdog.write_text(watchdog_source)
-            self.assertEqual(check_scheduled_alerts(root), [])
-
-            first = root / SCHEDULED_ALERT_WORKFLOWS[0]
-            mutations = (
-                ("contains(needs.*.result, 'failure')", "false"),
-                ("issues: write", "issues: read"),
-                ("uses: ./.github/actions/schedule-failure-issue", "uses: actions/checkout@v7"),
-                ("github-token: ${{ secrets.GITHUB_TOKEN }}", "github-token: missing"),
-            )
-            for required, replacement in mutations:
-                original = first.read_text()
-                first.write_text(original.replace(required, replacement))
-                self.assertEqual(len(check_scheduled_alerts(root)), 1)
-                first.write_text(original)
-
-            watchdog.write_text(watchdog_source.replace(f'- "{names[0]}"\n', ""))
-            self.assertEqual(len(check_scheduled_alerts(root)), 1)
-
-            watchdog_mutations = (
-                ("actions: read", "actions: none"),
-                ("issues: write", "issues: read"),
-                ("uses: ./.github/actions/schedule-failure-issue", "uses: actions/checkout@v7"),
-                ("github-token: ${{ secrets.GITHUB_TOKEN }}", "github-token: missing"),
-                ("source-event: ${{ github.event.workflow_run.event }}", "source-event: wrong"),
-                ("source-ref-name: ${{ github.event.workflow_run.head_branch }}", "source-ref-name: wrong"),
-                ("source-sha: ${{ github.event.workflow_run.head_sha }}", "source-sha: wrong"),
-            )
-            for required, replacement in watchdog_mutations:
-                watchdog.write_text(watchdog_source.replace(required, replacement))
-                self.assertEqual(len(check_scheduled_alerts(root)), 1)
-
 def main() -> int:
     if sys.argv[1:] == ["--self-test"]:
         suite = unittest.defaultTestLoader.loadTestsFromTestCase(SelfTests)
@@ -594,7 +436,7 @@ def main() -> int:
         for error in errors:
             print(f"ERROR: {error}", file=sys.stderr)
         return 1
-    print("OK: e2e modules, runner selection, fuzz matrices, profiles, scheduled alerts, and bounded diagnostics are wired")
+    print("OK: e2e modules, runner selection, fuzz matrices, profiles, and bounded diagnostics are wired")
     return 0