`GET /v3/tier-stats` answered from whichever process received the
request, returning that node's rolling 24-hour transition counters as
if they were cluster totals, and the `TierRequestsSuccess` and
`TierRequestsFailure` metric names had no producer at all.
The body now separates the two quantities a tier carries. Stored
inventory comes from the persisted scanner usage snapshot, which is
already cluster-wide; rolling activity is summed over every member
through a new read-only `TierDailyStats` peer RPC. Rings are merged
rather than added, so an idle node's expired hours age out, and each
node counts only its own committed transitions, so a retry is counted
once. Coverage travels with the numbers: `activity.status` names the
reporting members and the ones that could not be asked, timed out, or
answered with a ring this build refuses to merge, and per-tier
inventory is absent rather than zero when the snapshot has no
accounting. The version 1 body stays reachable at `?format=legacy`.
Tier request counters are recorded at the two seams every remote
request passes through, so a new provider is counted by construction,
with a closed operation/outcome label set that can never grow a tier
name, endpoint or object key.
Closesrustfs/backlog#2207
Co-authored-by: cxymds <cxymds@gmail.com>
Also replace deprecated Atomic::fetch_update calls with try_update so the
current Rust toolchain keeps lint and CI jobs warning-clean.
Co-authored-by: heihutu <heihutu@gmail.com>
TierAzure.storage_class and .sp_auth round-trip faithfully through the
admin API and on-disk config (ExternalTierAzure encode/decode in
tier.rs), so an operator can configure them, read them back via
ListTier, and never learn they do nothing. They are dropped only at the
WarmBackendAzure construction boundary: the Azure warm backend goes
through the same S3-compatible TransitionClient as every other
provider and has no Azure Blob-native client or Azure AD dependency
(confirmed: no azure_* crate anywhere in the workspace), so neither
field can actually be honored today. MinIO's reference implementation
(cmd/warm-backend-azure.go) treats both as first-class: storage_class
sets the blob access tier on every PUT, and sp_auth is a full
alternative to access/secret-key auth via azidentity, mutually
exclusive with it.
Rather than the larger, riskier options (add a native Azure SDK
dependency and a parallel non-S3 client path, or break the persisted
config format by removing the fields), this closes the silent-failure
gap with the minimal safe fix: TierConfigMgr::add now rejects an Azure
tier config with either field set, before backend construction,
returning ERR_TIER_INVALID_CONFIG with an explicit message instead of
accepting and ignoring. The fields stay in the config type (no format
break); already-persisted tiers with these fields set are grandfathered
in un-rejected (edit does not touch sp_auth or storage_class either).
Full support remains a larger follow-up if ever prioritized.
Also removes TierAzure::is_sp_enabled(), which had zero callers
repo-wide (backlog#2055 flagged this) and would have been misleading
dead weight once this decision was made — reusing it for the new gate
would also have been wrong, since it requires *all three* sp_auth
fields non-empty (&&), while the gate must reject on *any* one being set.
Refs rustfs/backlog#2055
(cherry picked from commit 8d148c4e9b2507a1c5075e3d9513adb8b5851ef5)
Validate lifecycle tier references through the tier reference proof path, preserve S3 list CommonPrefix XML compatibility, and make GetObject audit completion use real S3 error status codes.
Co-authored-by: heihutu <heihutu@gmail.com>
refactor(ecstore): huaweicloud/tencent reuse the shared S3 constructor
Migrates the Huaweicloud and Tencent tier warm backends onto the shared
S3-compatible constructor (backlog#2040). Also makes the shared
constructor's outbound-URL validation injectable per provider
(S3CompatibleWarmBackendParams::validate_endpoint) so it can centralize
rustfs/rustfs#6764's SSRF check for the providers that don't need an
exception, while accommodating rustfs/rustfs#6773's RustFS-specific
debug-only loopback opt-in without weakening the other six providers.
Updates scripts/error-other-format-baseline.txt: the one ::other(format!)
call site moves from the two per-provider files into the new shared
call site in warm_backend.rs (net call-site count unchanged).
Refs rustfs/backlog#2042
WarmBackendS3::new already rejects loopback, private, link-local, and
cloud metadata-service endpoints via validate_outbound_url, but the
Aliyun, Azure, Huaweicloud, Tencent, MinIO, R2, RustFS, and GCS warm
backend constructors built their transition clients directly from
conf.endpoint without the same check.
The endpoint comes from the AddTier admin API, gated only by
SetTierAction, which can be a narrower IAM grant than root. Any
principal holding it could point one of these eight tier types at an
internal address (loopback, RFC1918, link-local, or a cloud metadata
IP) and have the server issue authenticated outbound requests to it, a
server-side SSRF vector that the S3 and Wasabi tier types were already
closed against.
Apply the same validate_outbound_url check at construction time for
all eight providers, before any credentials or network client are
built, mirroring the existing WarmBackendS3 pattern. GCS keeps its
default-endpoint behavior when conf.endpoint is empty and only
validates an explicitly configured endpoint.
Add a regression test per provider asserting that a loopback endpoint
is rejected before any backend/network setup, matching the existing
WarmBackendS3 coverage.
Update the error(format!) ratchet baseline: these are one-shot admin
tier-configuration validation errors returned once per AddTier call,
not per-disk I/O errors that flow through reduce_errs quorum
aggregation (backlog#1845), so the new ::other(format!) call sites do
not introduce a quorum-bucketing hazard. They mirror the pre-existing,
already-baselined warm_backend_s3.rs call site.
TransitionClient::new() in crates/s3-client/src/transition_api.rs computes
trailing_header_support = opts.trailing_headers && override_signer_type == SignatureV4,
but override_signer_type is hardcoded to SignatureDefault at construction
and never mutated afterwards, so the expression is always false regardless
of opts.trailing_headers. The resulting field also has no live reader: its
only reference is inside PutObjectOptions::validate() in
crates/s3-client/src/api_put_object.rs, which is itself
#[allow(dead_code, reason = "MinIO-parity ... no caller in this port")],
and even there the reference to trailing_header_support is commented out.
So trailing_headers: true in the seven warm_backend_*.rs constructors has
never had any effect on request signing or chunked/trailing-header
behavior (stream_sha256 signing is gated separately by
metadata.stream_sha256 && !self.secure). Remove the misleading dead
configuration from the seven provider constructors so it doesn't look
like intentional, load-bearing behavior to future readers.
Found during adversarial self-check while implementing rustfs/backlog#2040 (out of that issue's scope).
The seven S3-compatible warm backend providers (Aliyun, Azure, Huaweicloud, Tencent, MinIO, R2, RustFS) each carry a byte-identical copy of the same statically-credentialed TransitionClient construction and of the same optimal_part_size helper. Add both to the module that already owns the WarmBackend trait and WarmBackendS3, so the per-provider migrate step can drop its duplicate without redesigning anything.
bucket_lookup is a parameter rather than a constant because the providers split into two families: Aliyun, Azure, Huaweicloud, and Tencent pin BucketLookupDNS, while MinIO, R2, and RustFS leave it at the BucketLookupAuto default. Hardcoding either value would silently change bucket addressing for the other family during the migrate step.
Error texts, validation order, prefix and host/port normalization are reproduced exactly from the Aliyun/MinIO family. No provider file is touched and no production caller exists yet, so the new unit tests are the first callers.
Refs rustfs/backlog#2040
* Revert "perf(ecstore): use AHashMap for FileInfo metadata fields (#6738)"
This reverts commit 13a2ae212e.
* fix(filemeta): restore standard HashMap metadata (#6742)
Remove the direct ahash dependency added for FileInfo metadata and revert the affected filemeta/ecstore call sites back to std::collections::HashMap.
Co-authored-by: heihutu <heihutu@gmail.com>
---------
Co-authored-by: heihutu <heihutu@gmail.com>
* refactor(ecstore): drop the client shim, import rustfs-s3-client directly
Completes the migration window opened by the rustfs-s3-client extraction (rustfs/backlog#1842 PR3): every consumer now imports the client crate directly and the crate::client shim is deleted.
- All in-crate crate::client:: paths (tier warm backends, tier core, lifecycle tier_sweeper, replication storage boundary, set_disk) now import rustfs_s3_client::* directly; crates/ecstore/src/client/mod.rs and the lib.rs mod client declaration are gone.
- The two server-side modules historically misfiled under client/ move to their real homes: object_api_utils.rs to crates/ecstore/src/object_api/ (it builds engine-side object readers/writers), and object_handlers_common.rs to crates/ecstore/src/bucket/lifecycle/ (it is the lifecycle noncurrent-version cleanup helper). The latter now routes its replication calls through the lifecycle replication_sink boundary (schedule_delete wrapper and the sink's ReplicationObjectBridge re-export), as the lifecycle guard requires.
- The ecstore public facade drops api::client: object_api_utils is exposed as api::object_api_utils, and the rustfs crate takes admin_handler_utils (AdminError) from rustfs-s3-client directly (new dependency).
- Guard updates: the migration guard no longer pins mod client in ecstore's lib.rs or the admin_handler_utils facade module (it pins the new api::object_api_utils facade instead), and the module-lint register follows object_api_utils.rs to its new path.
Verification: cargo check -p rustfs-ecstore --all-targets and -p rustfs; cargo fmt --all; tier/transition/lifecycle-focused nextest (626 passed) and the decommission/rebalance/heal families in a filtered run (603 passed; the full-suite parallel run only fails on this machine's known decommission/rebalance baseline flakes, which pass in filtered reruns and fail identically on pristine origin/main); layer/migration/s3s/logging/error-format/doc-path guard scripts all pass.
* docs(architecture): record the S3 client extraction and reword invariant 4 (#6669)
Closes the documentation step of rustfs/backlog#1842. ARCHITECTURE.md invariant 4 now states the serving-vs-consuming distinction the adversarial ruling asked for: ecstore must not serve HTTP/S3 wire types, while consuming remote S3 endpoints is a legitimate engine capability that lives in the extracted rustfs-s3-client crate. The violation note is updated from the pre-extraction snapshot (58 files, embedded client) to the current ratcheted state (shrink-only S3S_ECSTORE_FILES_BASELINE in scripts/check_s3s_footprint.sh, object_lock converted first), and the crate map gains s3-client. ecstore-module-split-plan.md gets the client-directory entry the plan was missing: a Current Shape row and a completed-extraction section describing the pure-move + shim + direct-import sequence and the re-homing of the two misfiled server-side modules.
The storage engine embedded a ~8.4K-line hand-written S3 HTTP client under crates/ecstore/src/client (rustfs/backlog#1842). That client is a legitimate engine capability — it consumes remote S3-compatible endpoints for ILM tier warm backends and transition targets — but it was misfiled inside the engine, dragging s3s/hyper wire types into ecstore and blocking ARCHITECTURE.md invariant 4.
This PR is the pure-move step: 21 modules move verbatim to the new crates/s3-client crate (rustfs-s3-client), and crates/ecstore/src/client/mod.rs becomes a re-export shim so every in-crate crate::client:: path keeps working. The two server-side modules that were historically misfiled under client/ — object_api_utils.rs and object_handlers_common.rs — stay in ecstore.
Three reverse dependencies from the client into engine internals are severed so the move can be pure:
- transition_api::ReaderImpl::ObjectBody held ecstore's GetObjectReader; the client only ever reads the body, so the variant now holds an ObjectReader newtype over Box<dyn AsyncRead + Send + Sync + Unpin> with the same read_all() surface. The single production construction site (set_disk transition upload) and the two engine-side consumers were adjusted.
- api_list/api_remove used ecstore's storage_api_contracts / object_api types; api_list now imports BucketInfo from rustfs-storage-api directly, and api_remove uses the client's own transition_api::ObjectInfo (only .name/.version_id were read; the error-path bucket name is now threaded as a parameter instead of read from the deleted objects).
- the api_put_object_streaming regression tests built a GetObjectReader by hand; they now wrap the duplex stream in ObjectReader::new.
Guard updates: the s3s footprint ratchet gains an ecstore-scoped counter (42 files, shrink-only, per rustfs/backlog#1842), the ecstore module-lint-blanket register follows the moved files into crates/s3-client so the blanket ratchet keeps covering them, the logging guardrail path pin follows transition_api.rs, and the ::other(format!) baseline is regenerated (moved call sites left ecstore).
Verification: cargo check -p rustfs-s3-client -p rustfs-ecstore; cargo nextest run -p rustfs-s3-client (43 passed) and -p rustfs-ecstore (4515/4523; the 8 failures reproduce identically on pristine origin/main on the same machine); cargo clippy --all-targets; scripts/check_layer_dependencies.sh, check_architecture_migration_rules.sh, check_s3s_footprint.sh, check_logging_guardrails.sh, check_error_other_format_ratchet.sh, check_doc_paths.sh, check_ci_paths_sync.sh all pass.
Replace every bare `#[allow(dead_code)]` in ecstore with either a deletion or a per-item allow carrying a `reason`. Blanket allows at module, struct, and impl level silence the lint for future members too, so each is narrowed to the members that are actually dead.
Delete the dead cluster in `config/heal.rs` (`Config`, its three methods, `RUSTFS_BITROT_CYCLE_IN_MONTHS`, `parse_bitrot_config`) rather than annotate it: it has no callers and is unreachable outside the crate, and `parse_bitrot_config` would panic on its disabled path via `Duration::from_secs_f64(-1.0)`. `DEFAULT_KVS` stays, since the config registry uses it.
Correct two `reason` strings on `Checksum::new` and `PutObjReader::md5_current_hex_string`, which are methods but carried a field-only rationale.
Refs backlog#1823
Co-authored-by: houseme <housemecn@gmail.com>
Removing the blanket exposes twenty-five items across tier, notification and rebalance. Only eight are deleted — the lowest ratio of this burn-down so far, and the reason is that these subsystems carry heavy test coverage, so the blanket was mostly hiding test-only seams rather than dead weight.
Deleted:
- crates/ecstore/src/services/tier/warm_backend_s3sdk.rs entirely (200 lines). Its WarmBackendS3 is never constructed; the type of the same name in warm_backend_s3.rs is the live one, wrapped by the Azure backend. Two implementations of one S3 warm backend, one of them never wired.
- TierConfigMgr::begin_publish_transition and publish_candidate_inner, thin wrappers whose _with_allowed_mutation_blocks siblings carry every real caller, plus retire_driver.
- The GCS backend's MAX_MULTIPART_PUT_OBJECT_SIZE, MAX_PARTS_COUNT and MIN_PART_SIZE, and its write-only storage_class field.
- mark_started_rebalance_pools_stopped and the RStats alias.
Two deletions were withdrawn after a per-name grep, both because of an inference rather than a check:
AsyncBatchProcessor::new was deleted on the strength of grepping only BATCH_PROCESSOR_OPERATION_CUSTOM, whose two hits are its definition and its use inside new. That looked like a self-contained dead pair; new in fact has seven test callers. The warning listed both items, and only one of them was actually checked.
Deleting the two dead publish wrappers then revealed a second layer — publish_candidate_owned, remove_and_save_with, clear_and_save_with, save_tiering_config_if_current. These are not dead: publish_candidate, their caller, is #[cfg(test)], so a callee that lives in the main body has no caller in the lib build and a live one in the test build. rustc reports the roots of a dead subgraph, and the next layer down can have a different character, so each layer needs its own grep.
Kept with allows: the tier mutation-intent record helpers (asserted by store::init tests), affected_targets, tier_object_blocks_target_rebind, the rebalance snapshot and retry-wait helpers, notification_sys's tier_config_reload_worker_active and call_peer_with_timeout, and active_operation_lease_count, whose only caller sits behind #[cfg(feature = "test-util")].
Also kept, with a module note rather than removal: the ecstore-side EventNotifier. All four of its methods are unreachable and init_bucket_targets logs that it is a no-op in this build; the working stack is rustfs-notify, whose own EventNotifier drives bucket configuration. Removing it means also retiring the InstanceContext slot that holds it (backlog#939 Phase 5), which belongs in its own PR.
Worth a separate issue: MAX_MULTIPART_PUT_OBJECT_SIZE, MAX_PARTS_COUNT and MIN_PART_SIZE are declared independently in eight warm-backend files plus client/constants.rs. Only the GCS copies were dead; the other seven backends each use their own.
Verification, four lanes warning-free: default, --tests, --features rio-v2 --tests, --features test-util --tests. cargo nextest run -p rustfs-ecstore 4041 passed; clippy --lib --tests -D warnings clean; make pre-commit exit 0.
Ref rustfs/backlog#1823 (step 2).
Add a deterministic four-hot AddTier regression for the reporter path where one peer has already durably committed prepare replay and the remaining peers still need commit fanout before local publish.
Co-authored-by: heihutu <heihutu@gmail.com>
* fix(tier): gate exact remote version consumption
Reject non-empty remote tier versions before transitioned GET and remote delete backend I/O when the tier backend does not support exact version operations.
Co-Authored-By: heihutu <heihutu@gmail.com>
* test(tier): split mock remote version validation fault
Separate one-shot mock remote version validation failures from persistent unsupported-backend behavior so cleanup durability tests can still verify exact-version recovery while #1358 fail-closed gate tests keep asserting no backend I/O.
Co-Authored-By: heihutu <heihutu@gmail.com>
---------
Co-authored-by: heihutu <heihutu@gmail.com>
Advance transition transactions to UploadOutcomeUnknown before remote tier PUT so response-loss windows can be recovered through provider-authoritative probing.
Co-authored-by: heihutu <heihutu@gmail.com>
Keep coordinator Prepared intents durable when peer abort recovery fails, then make reload retry abort or commit based on the persisted tier config digest.
Co-authored-by: heihutu <heihutu@gmail.com>
Zhengchao An