ci(s3tests): omit proxy query strings

.config/nextest.toml

@@ -78,12 +78,6 @@ test-group = 'embedded-test-ports'
 filter = 'package(rustfs-ecstore) & test(manual_transition_page_checkpoint_persists_durable_job_progress)'
 test-group = 'ecstore-serial-flaky'
 
-# The durable ILM decommission regressions build isolated multi-pool stores and
-# deliberately take source or target disks offline while checking fencing.
-[[profile.default.overrides]]
-filter = 'package(rustfs-ecstore) & (test(decommission_migrates_and_verifies_registered_durable_ilm_records) | test(decommission_durable_ilm_target_read_error_is_not_masked_by_peer_success) | test(decommission_durable_ilm_terminal_receipt_recovers_failed_source_cleanup) | test(decommission_durable_ilm_receipt_pagination_fails_closed_on_second_page) | test(decommission_durable_ilm_recovery_keeps_multiple_active_sources))'
-test-group = 'ecstore-serial-flaky'
-
 # Serialize the bucket-incarnation / lifecycle-fence tests. They drive
 # init_bucket_metadata_sys and bucket_metadata_sys_of, i.e. process-global
 # OnceLock state that serial_test's #[serial] cannot protect across nextest's
@@ -196,10 +190,6 @@ test-group = 'embedded-test-ports'
 filter = 'package(rustfs-ecstore) & test(manual_transition_page_checkpoint_persists_durable_job_progress)'
 test-group = 'ecstore-serial-flaky'
 
-[[profile.ci.overrides]]
-filter = 'package(rustfs-ecstore) & (test(decommission_migrates_and_verifies_registered_durable_ilm_records) | test(decommission_durable_ilm_target_read_error_is_not_masked_by_peer_success) | test(decommission_durable_ilm_terminal_receipt_recovers_failed_source_cleanup) | test(decommission_durable_ilm_receipt_pagination_fails_closed_on_second_page) | test(decommission_durable_ilm_recovery_keeps_multiple_active_sources))'
-test-group = 'ecstore-serial-flaky'
-
 # Serialize the bucket-incarnation / lifecycle-fence tests under the ci profile
 # too (see the matching default-profile override near the top). No retries.
 [[profile.ci.overrides]]

.github/workflows/e2e-s3tests.yml

@@ -90,6 +90,10 @@ on:
         description: "Optional pytest -m expression"
         required: false
         default: ""
+      testexpr:
+        description: "Optional pytest -k expression"
+        required: false
+        default: ""
   schedule:
     # Weekly full sweep (Sunday 02:00 UTC): full suite, run against BOTH the
     # single-node and the 4-node distributed topologies (matrix below).
@@ -116,6 +120,7 @@ env:
   XDIST: ${{ github.event.inputs.xdist || '4' }}
   MAXFAIL: ${{ github.event.inputs.maxfail || '0' }}
   MARKEXPR: ${{ github.event.inputs.markexpr || '' }}
+  TESTEXPR: ${{ github.event.inputs.testexpr || '' }}
   S3_SHARD_COUNT: ${{ github.event_name == 'schedule' && '4' || github.event.inputs.shard-count || '1' }}
   TEST_TIMEOUT: "300"
 
@@ -269,8 +274,13 @@ jobs:
           EOF
 
           cat > haproxy.cfg <<'EOF'
+          global
+            log stdout format raw local0 info
+
           defaults
             mode http
+            log global
+            log-format '%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %tsc %HM %HP'
             timeout connect 5s
             timeout client  30s
             timeout server  30s
@@ -314,6 +324,7 @@ jobs:
           XDIST="${XDIST}" \
           MAXFAIL="${MAXFAIL}" \
           MARKEXPR="${MARKEXPR}" \
+          TESTEXPR="${TESTEXPR}" \
           ./scripts/s3-tests/run.sh
 
       - name: Publish compatibility report

crates/e2e_test/src/cluster_concurrency_test.rs

@@ -15,12 +15,14 @@
 use crate::common::RustFSTestClusterEnvironment;
 use aws_sdk_s3::Client;
 use aws_sdk_s3::error::SdkError;
+use aws_sdk_s3::types::{CorsConfiguration, CorsRule};
 use bytes::Bytes;
 use std::sync::Arc;
 use tokio::sync::Barrier;
 use tracing::{info, warn};
 
 const BUCKET: &str = "conditional-put-race-bucket";
+const BUCKET_METADATA_RELOAD_BUCKET: &str = "bucket-metadata-reload-barrier";
 
 async fn cleanup_object(client: &Client, key: &str) {
     if let Err(e) = client.delete_object().bucket(BUCKET).key(key).send().await {
@@ -28,6 +30,16 @@ async fn cleanup_object(client: &Client, key: &str) {
     }
 }
 
+async fn assert_bucket_cors_missing(client: &Client) {
+    let result = client.get_bucket_cors().bucket(BUCKET_METADATA_RELOAD_BUCKET).send().await;
+    match result {
+        Err(SdkError::ServiceError(error)) => {
+            assert_eq!(error.err().meta().code(), Some("NoSuchCORSConfiguration"));
+        }
+        result => panic!("expected the peer to report a missing CORS configuration: {result:?}"),
+    }
+}
+
 async fn conditional_put(
     client: &Client,
     key: &str,
@@ -236,3 +248,48 @@ async fn test_conditional_put_basic_cluster() -> Result<(), Box<dyn std::error::
     cleanup_object(&client, test_key).await;
     Ok(())
 }
+
+#[tokio::test]
+async fn test_bucket_cors_write_is_visible_on_peer_before_response() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    crate::common::init_logging();
+
+    let mut cluster = RustFSTestClusterEnvironment::new(2).await?;
+    cluster.start().await?;
+    cluster.create_test_bucket(BUCKET_METADATA_RELOAD_BUCKET).await?;
+
+    let writer = cluster.create_s3_client(0)?;
+    let reader = cluster.create_s3_client(1)?;
+    assert_bucket_cors_missing(&reader).await;
+
+    let rule = CorsRule::builder()
+        .allowed_methods("GET")
+        .allowed_origins("https://example.com")
+        .build()?;
+    let configuration = CorsConfiguration::builder().cors_rules(rule).build()?;
+
+    writer
+        .put_bucket_cors()
+        .bucket(BUCKET_METADATA_RELOAD_BUCKET)
+        .cors_configuration(configuration)
+        .send()
+        .await?;
+
+    let response = reader.get_bucket_cors().bucket(BUCKET_METADATA_RELOAD_BUCKET).send().await?;
+    let rules = response.cors_rules();
+    assert_eq!(
+        rules.len(),
+        1,
+        "peer should observe the committed CORS rule before the write response returns"
+    );
+    assert_eq!(rules[0].allowed_methods(), ["GET"]);
+    assert_eq!(rules[0].allowed_origins(), ["https://example.com"]);
+
+    writer
+        .delete_bucket_cors()
+        .bucket(BUCKET_METADATA_RELOAD_BUCKET)
+        .send()
+        .await?;
+    assert_bucket_cors_missing(&reader).await;
+    writer.delete_bucket().bucket(BUCKET_METADATA_RELOAD_BUCKET).send().await?;
+    Ok(())
+}

crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs

@@ -9136,7 +9136,6 @@ mod tests {
         assert_eq!(loaded.report.scanned, 37);
         assert_eq!(loaded.report.eligible, 11);
         assert_eq!(loaded.report.enqueued, 5);
-        assert_eq!(loaded.cursor_revision, Some(1));
         assert!(loaded.lease_expires_at_unix_nanos > 0);
         let token = loaded
             .report
@@ -9155,30 +9154,6 @@ mod tests {
         assert_eq!(admission.lease_id, loaded.lease_id);
         assert_eq!(admission.lease_expires_at_unix_nanos, loaded.lease_expires_at_unix_nanos);
 
-        let mut same_marker_report = report.clone();
-        same_marker_report.scanned += 1;
-        persist_manual_transition_page_checkpoint(
-            &checkpoint_options,
-            &same_marker_report,
-            Some("logs/page-end".to_string()),
-            Some("opaque-next-version".to_string()),
-        )
-        .await
-        .expect("same-marker version checkpoint should persist through the durable progress sink");
-        let same_marker_checkpointed = load_manual_transition_job_record(ecstore.clone(), job_id)
-            .await
-            .expect("same-marker version checkpoint should reload");
-        assert_eq!(same_marker_checkpointed.cursor_revision, Some(2));
-        let (_, version_marker) = decode_manual_transition_continuation_token(
-            same_marker_checkpointed
-                .report
-                .continuation_token
-                .as_deref()
-                .expect("same-marker version checkpoint should persist a cursor"),
-        )
-        .expect("same-marker version cursor should decode");
-        assert_eq!(version_marker.as_deref(), Some("opaque-next-version"));
-
         create_test_bucket(&ecstore, &bucket).await;
         let lifecycle_xml = format!(
             r#"<?xml version="1.0" encoding="UTF-8"?>
@@ -9235,7 +9210,6 @@ mod tests {
         assert_eq!(checkpointed.report.scanned, 1000);
         assert_eq!(checkpointed.report.eligible, 1000);
         assert_eq!(checkpointed.report.dry_run_eligible, 1000);
-        assert_eq!(checkpointed.cursor_revision, Some(3));
         let token = checkpointed
             .report
             .continuation_token

crates/ecstore/src/bucket/lifecycle/manual_transition_job.rs

@@ -24,10 +24,6 @@ use crate::bucket::lifecycle::bucket_lifecycle_ops::{
     ManualTransitionQueueSnapshot, ManualTransitionRunOptions, ManualTransitionRunReport,
 };
 use crate::bucket::lifecycle::config_boundary;
-use crate::bucket::lifecycle::durable_namespace::{
-    MANUAL_TRANSITION_JOB_NAMESPACE, MANUAL_TRANSITION_SCOPE_NAMESPACE, MANUAL_TRANSITION_TASK_NAMESPACE,
-    MANUAL_TRANSITION_WORKER_RESULT_NAMESPACE,
-};
 use crate::disk::RUSTFS_META_BUCKET;
 use crate::error::{Error, Result as EcstoreResult};
 use crate::object_api::ObjectOptions;
@@ -38,10 +34,10 @@ use crate::store::ECStore;
 pub const MANUAL_TRANSITION_JOB_SCHEMA: &str = "rustfs-manual-transition-job-v1";
 pub const MANUAL_TRANSITION_TASK_SCHEMA: &str = "rustfs-manual-transition-task-v1";
 pub const MANUAL_TRANSITION_WORKER_RESULT_SCHEMA: &str = "rustfs-manual-transition-worker-result-v1";
-pub const MANUAL_TRANSITION_JOB_RECORD_PREFIX: &str = MANUAL_TRANSITION_JOB_NAMESPACE.prefix;
-pub const MANUAL_TRANSITION_SCOPE_RECORD_PREFIX: &str = MANUAL_TRANSITION_SCOPE_NAMESPACE.prefix;
-pub const MANUAL_TRANSITION_TASK_PREFIX: &str = MANUAL_TRANSITION_TASK_NAMESPACE.prefix;
-pub const MANUAL_TRANSITION_WORKER_RESULT_PREFIX: &str = MANUAL_TRANSITION_WORKER_RESULT_NAMESPACE.prefix;
+pub const MANUAL_TRANSITION_JOB_RECORD_PREFIX: &str = "ilm/manual-transition/jobs";
+pub const MANUAL_TRANSITION_SCOPE_RECORD_PREFIX: &str = "ilm/manual-transition/scopes";
+pub const MANUAL_TRANSITION_TASK_PREFIX: &str = "ilm/manual-transition/tasks";
+pub const MANUAL_TRANSITION_WORKER_RESULT_PREFIX: &str = "ilm/manual-transition/results";
 pub const MAX_MANUAL_TRANSITION_JOB_RECORD_SIZE: usize = 64 * 1024;
 pub const MAX_MANUAL_TRANSITION_TASK_RECORD_SIZE: usize = 16 * 1024;
 pub const MAX_MANUAL_TRANSITION_WORKER_RESULT_RECORD_SIZE: usize = 8 * 1024;
@@ -199,8 +195,6 @@ pub struct ManualTransitionJobRecord {
     pub updated_at_unix_nanos: i128,
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub completed_at_unix_nanos: Option<i128>,
-    #[serde(default, skip_serializing_if = "Option::is_none")]
-    pub cursor_revision: Option<u64>,
     pub report: ManualTransitionRunReport,
     pub queue_snapshot: ManualTransitionQueueSnapshot,
     #[serde(default, skip_serializing_if = "Option::is_none")]
@@ -230,7 +224,6 @@ impl ManualTransitionJobRecord {
             created_at_unix_nanos: now,
             updated_at_unix_nanos: now,
             completed_at_unix_nanos: None,
-            cursor_revision: Some(0),
             report: ManualTransitionRunReport {
                 bucket: bucket.to_string(),
                 prefix: options.prefix.clone(),
@@ -245,7 +238,7 @@ impl ManualTransitionJobRecord {
 
     pub fn complete(&mut self, report: ManualTransitionRunReport, queue_snapshot: ManualTransitionQueueSnapshot) {
         self.scan_completed = true;
-        self.merge_scan_report(&report);
+        self.report.merge_scan_report_preserving_worker(&report);
         self.queue_snapshot = queue_snapshot;
         self.error = None;
         self.mark_terminal_if_worker_drained();
@@ -323,7 +316,7 @@ impl ManualTransitionJobRecord {
             }
         }
         self.queue_snapshot = queue_snapshot;
-        self.advance_updated_at();
+        self.updated_at_unix_nanos = OffsetDateTime::now_utc().unix_timestamp_nanos();
         self.mark_terminal_if_worker_drained();
     }
 
@@ -366,14 +359,14 @@ impl ManualTransitionJobRecord {
         self.report.tier_failure = scan_tier_failure.saturating_add(transition_failed);
         self.report.tier_failure_by_reason = scan_tier_failure_by_reason;
         self.queue_snapshot = queue_snapshot;
-        self.advance_updated_at();
+        self.updated_at_unix_nanos = OffsetDateTime::now_utc().unix_timestamp_nanos();
         self.mark_terminal_if_worker_drained();
         true
     }
 
     pub fn mark_cancel_requested(&mut self) {
         self.cancel_requested = true;
-        self.advance_updated_at();
+        self.updated_at_unix_nanos = OffsetDateTime::now_utc().unix_timestamp_nanos();
     }
 
     pub fn claim_recovery_lease(&mut self, owner_id: impl Into<String>, queue_snapshot: ManualTransitionQueueSnapshot) {
@@ -387,7 +380,7 @@ impl ManualTransitionJobRecord {
     pub fn abandon_recovery_lease(&mut self, lease_id: Uuid) {
         if self.state == ManualTransitionJobState::Running && self.lease_id == lease_id {
             self.lease_expires_at_unix_nanos = 0;
-            self.advance_updated_at();
+            self.updated_at_unix_nanos = OffsetDateTime::now_utc().unix_timestamp_nanos();
         }
     }
 
@@ -405,7 +398,7 @@ impl ManualTransitionJobRecord {
 
     pub fn renew_lease(&mut self, queue_snapshot: ManualTransitionQueueSnapshot) {
         let now = OffsetDateTime::now_utc().unix_timestamp_nanos();
-        self.updated_at_unix_nanos = self.updated_at_unix_nanos.saturating_add(1).max(now);
+        self.updated_at_unix_nanos = now;
         self.lease_expires_at_unix_nanos = manual_transition_job_lease_expires_at(now);
         self.queue_snapshot = queue_snapshot;
     }
@@ -445,18 +438,11 @@ impl ManualTransitionJobRecord {
 
     pub fn update_running_progress(&mut self, report: ManualTransitionRunReport, queue_snapshot: ManualTransitionQueueSnapshot) {
         if self.state == ManualTransitionJobState::Running {
-            self.merge_scan_report(&report);
+            self.report.merge_scan_report_preserving_worker(&report);
             self.renew_lease(queue_snapshot);
         }
     }
 
-    fn merge_scan_report(&mut self, report: &ManualTransitionRunReport) {
-        if self.report.continuation_token != report.continuation_token {
-            self.cursor_revision = Some(self.cursor_revision.unwrap_or(0).saturating_add(1));
-        }
-        self.report.merge_scan_report_preserving_worker(report);
-    }
-
     pub fn mark_unknown_if_unowned(&mut self) {
         if self.state == ManualTransitionJobState::Running {
             self.state = ManualTransitionJobState::Unknown;
@@ -477,13 +463,9 @@ impl ManualTransitionJobRecord {
     }
 
     fn mark_updated_terminal(&mut self) {
-        self.advance_updated_at();
-        self.completed_at_unix_nanos = Some(self.updated_at_unix_nanos);
-    }
-
-    fn advance_updated_at(&mut self) {
         let now = OffsetDateTime::now_utc().unix_timestamp_nanos();
-        self.updated_at_unix_nanos = self.updated_at_unix_nanos.saturating_add(1).max(now);
+        self.updated_at_unix_nanos = now;
+        self.completed_at_unix_nanos = Some(now);
     }
 
     fn mark_terminal_if_worker_drained(&mut self) {
@@ -1127,8 +1109,7 @@ pub fn manual_transition_scope_record_object_name(scope_key: &str) -> Result<Str
 pub async fn save_manual_transition_job_record(api: Arc<ECStore>, job: &ManualTransitionJobRecord) -> EcstoreResult<()> {
     let object = manual_transition_job_record_object_name(job.job_id).map_err(manual_transition_job_store_error)?;
     let data = job.encode().map_err(manual_transition_job_store_error)?;
-    config_boundary::save_config(api.clone(), &object, data.clone()).await?;
-    api.record_durable_ilm_decommission_progress(&object, &data).await
+    config_boundary::save_config(api, &object, data).await
 }
 
 pub async fn load_manual_transition_job_record(api: Arc<ECStore>, job_id: Uuid) -> EcstoreResult<ManualTransitionJobRecord> {
@@ -1161,9 +1142,9 @@ pub async fn save_manual_transition_job_record_if_current(
     let object = manual_transition_job_record_object_name(job.job_id).map_err(manual_transition_job_store_error)?;
     let data = job.encode().map_err(manual_transition_job_store_error)?;
     config_boundary::save_config_with_opts_quiet(
-        api.clone(),
+        api,
         &object,
-        data.clone(),
+        data,
         &ObjectOptions {
             max_parity: true,
             http_preconditions: Some(HTTPPreconditions {
@@ -1173,8 +1154,7 @@ pub async fn save_manual_transition_job_record_if_current(
             ..Default::default()
         },
     )
-    .await?;
-    api.record_durable_ilm_decommission_progress(&object, &data).await
+    .await
 }
 
 /// Applies a job-record mutation with optimistic concurrency control.
@@ -1612,9 +1592,9 @@ pub async fn save_manual_transition_scope_admission_if_absent(
     let object = manual_transition_scope_record_object_name(&admission.scope_key).map_err(manual_transition_job_store_error)?;
     let data = serde_json::to_vec(admission).map_err(Error::other)?;
     config_boundary::save_config_with_opts(
-        api.clone(),
+        api,
         &object,
-        data.clone(),
+        data,
         &ObjectOptions {
             max_parity: true,
             http_preconditions: Some(HTTPPreconditions {
@@ -1624,8 +1604,7 @@ pub async fn save_manual_transition_scope_admission_if_absent(
             ..Default::default()
         },
     )
-    .await?;
-    api.record_durable_ilm_decommission_progress(&object, &data).await
+    .await
 }
 
 pub async fn load_manual_transition_scope_admission(
@@ -1663,9 +1642,9 @@ pub async fn save_manual_transition_scope_admission_if_current(
     let object = manual_transition_scope_record_object_name(&admission.scope_key).map_err(manual_transition_job_store_error)?;
     let data = serde_json::to_vec(admission).map_err(Error::other)?;
     match config_boundary::save_config_with_opts(
-        api.clone(),
+        api,
         &object,
-        data.clone(),
+        data,
         &ObjectOptions {
             max_parity: true,
             http_preconditions: Some(HTTPPreconditions {
@@ -1681,8 +1660,7 @@ pub async fn save_manual_transition_scope_admission_if_current(
             Err(Error::PreconditionFailed)
         }
         result => result,
-    }?;
-    api.record_durable_ilm_decommission_progress(&object, &data).await
+    }
 }
 
 pub async fn claim_manual_transition_scope_admission(
@@ -1975,15 +1953,13 @@ pub async fn delete_manual_transition_scope_admission_if_current(
     job_id: Uuid,
     lease_id: Uuid,
 ) -> EcstoreResult<bool> {
-    let (admission, etag) = match load_manual_transition_scope_admission_with_etag(api.clone(), scope_key).await {
-        Ok((admission, etag)) if admission.job_id == job_id && admission.lease_id == lease_id => (admission, etag),
+    let etag = match load_manual_transition_scope_admission_with_etag(api.clone(), scope_key).await {
+        Ok((admission, etag)) if admission.job_id == job_id && admission.lease_id == lease_id => etag,
         Ok(_) => return Ok(false),
         Err(Error::ConfigNotFound) => return Ok(true),
         Err(err) => return Err(err),
     };
     let object = manual_transition_scope_record_object_name(scope_key).map_err(manual_transition_job_store_error)?;
-    let data = serde_json::to_vec(&admission).map_err(Error::other)?;
-    api.record_durable_ilm_decommission_terminal(&object, &data).await?;
     match config_boundary::delete_config_if_match(api, &object, &etag).await {
         Ok(()) | Err(Error::ConfigNotFound) => Ok(true),
         Err(Error::PreconditionFailed) => Ok(false),
@@ -2601,25 +2577,6 @@ mod tests {
         assert!(decoded.report.tier_failure_by_reason.is_empty());
     }
 
-    #[test]
-    fn manual_transition_job_record_decodes_legacy_cursor_without_revision() {
-        let options = ManualTransitionRunOptions::default();
-        let record = ManualTransitionJobRecord::new(Uuid::new_v4(), "bucket", &options, TEST_OWNER);
-        let encoded = record.encode().expect("job record should encode");
-        let mut value: serde_json::Value = serde_json::from_slice(&encoded).expect("encoded job should be json");
-        value["job"]
-            .as_object_mut()
-            .expect("job should be object")
-            .remove("cursor_revision");
-        let record_bytes = serde_json::to_vec(&value["job"]).expect("legacy job should encode");
-        value["content_sha256"] = serde_json::Value::String(hex_sha256(&record_bytes, ToOwned::to_owned));
-        let legacy = serde_json::to_vec(&value).expect("legacy envelope should encode");
-
-        let decoded = ManualTransitionJobRecord::decode(record.job_id, &legacy).expect("legacy job should decode");
-
-        assert_eq!(decoded.cursor_revision, None);
-    }
-
     #[test]
     fn manual_transition_job_record_rejects_unknown_report_fields() {
         let options = ManualTransitionRunOptions::default();

crates/ecstore/src/bucket/lifecycle/mod.rs

@@ -16,7 +16,6 @@ pub mod bucket_lifecycle_audit;
 pub mod bucket_lifecycle_ops;
 mod config_boundary;
 pub mod core;
-mod durable_namespace;
 pub mod evaluator;
 pub mod manual_transition_job;
 mod metadata_boundary;
@@ -32,8 +31,3 @@ pub mod tier_free_version_recovery;
 pub mod tier_last_day_stats;
 pub mod tier_sweeper;
 pub mod transition_transaction;
-
-pub(crate) use durable_namespace::{
-    DurableIlmRecordCheckpoint, ILM_META_PREFIX, ValidatedDurableIlmRecord, classify_durable_ilm_record,
-    validate_durable_ilm_record,
-};

crates/ecstore/src/bucket/lifecycle/tier_delete_journal.rs

@@ -20,7 +20,6 @@ use tokio_util::sync::CancellationToken;
 use tracing::{debug, warn};
 
 use crate::bucket::lifecycle::config_boundary;
-use crate::bucket::lifecycle::durable_namespace::TIER_DELETE_JOURNAL_NAMESPACE;
 use crate::bucket::lifecycle::runtime_boundary;
 use crate::bucket::lifecycle::tier_sweeper::{
     Jentry, TierDeleteJournalState, TierDeleteSourceIdentity,
@@ -50,7 +49,7 @@ const TIER_DELETE_JOURNAL_VERSION: u8 = 2;
 const TIER_DELETE_JOURNAL_EXACT_VERSION: u8 = 3;
 const TIER_DELETE_JOURNAL_STATE_VERSION: u8 = 4;
 const TIER_DELETE_JOURNAL_TRANSACTION_VERSION: u8 = 5;
-pub(crate) const TIER_DELETE_JOURNAL_PREFIX: &str = TIER_DELETE_JOURNAL_NAMESPACE.prefix;
+pub(crate) const TIER_DELETE_JOURNAL_PREFIX: &str = "ilm/tier-delete-journal/";
 
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 #[serde(deny_unknown_fields)]
@@ -433,21 +432,6 @@ async fn process_committed_tier_delete_journal_entry(api: Arc<ECStore>, je: &Jen
         )
         .await?;
     }
-    let path = tier_delete_journal_object_name(je);
-    let data = encode_tier_delete_journal_entry(je).map_err(std::io::Error::other)?;
-    let target_pool_indices = api
-        .record_durable_ilm_decommission_terminal_target_pools(&path, &data)
-        .await
-        .map_err(std::io::Error::other)?;
-    if let Some(target_pool_indices) = target_pool_indices {
-        for target_pool_idx in target_pool_indices {
-            match config_boundary::delete_config(api.pools[target_pool_idx].clone(), &path).await {
-                Ok(()) | Err(Error::ConfigNotFound) => {}
-                Err(err) => return Err(std::io::Error::other(err)),
-            }
-        }
-        return Ok(());
-    }
     remove_tier_delete_journal_entry(api, je).await
 }
 

crates/ecstore/src/bucket/lifecycle/transition_transaction.rs

@@ -21,7 +21,6 @@ use tracing::{debug, warn};
 use uuid::Uuid;
 
 use crate::bucket::lifecycle::config_boundary;
-use crate::bucket::lifecycle::durable_namespace::TRANSITION_TRANSACTION_NAMESPACE;
 use crate::bucket::lifecycle::lifecycle::TRANSITION_COMPLETE;
 use crate::bucket::lifecycle::tier_sweeper::{
     delete_confirmed_transition_candidate_exact_with_lease_idempotent,
@@ -43,7 +42,7 @@ const TRANSITION_TRANSACTION_RECOVERY_INTERVAL: Duration = Duration::from_secs(6
 const TRANSITION_TRANSACTION_RECOVERY_TIMEOUT: Duration = Duration::from_secs(300);
 pub const TRANSITION_TRANSACTION_SCHEMA: &str = "rustfs-transition-transaction-v1";
 pub const TRANSITION_TRANSACTION_PREFIX: &str = "ilm/transition-transactions";
-pub const TRANSITION_TRANSACTION_RECORD_PREFIX: &str = TRANSITION_TRANSACTION_NAMESPACE.prefix;
+pub const TRANSITION_TRANSACTION_RECORD_PREFIX: &str = "ilm/transition-transactions/records";
 pub const MAX_TRANSITION_TRANSACTION_SIZE: usize = 64 * 1024;
 
 pub type Result<T> = std::result::Result<T, TransitionTransactionError>;
@@ -585,8 +584,7 @@ pub(crate) async fn save_transition_transaction_record(
     let object =
         transition_transaction_record_object_name(transaction.transaction_id).map_err(transition_transaction_store_error)?;
     let data = transaction.encode().map_err(transition_transaction_store_error)?;
-    config_boundary::save_config(api.clone(), &object, data.clone()).await?;
-    api.record_durable_ilm_decommission_progress(&object, &data).await
+    config_boundary::save_config(api, &object, data).await
 }
 
 pub(crate) async fn load_transition_transaction_record(
@@ -598,14 +596,8 @@ pub(crate) async fn load_transition_transaction_record(
     TransitionTransaction::decode(transaction_id, &data).map_err(transition_transaction_store_error)
 }
 
-pub(crate) async fn delete_transition_transaction_record(
-    api: Arc<ECStore>,
-    transaction: &TransitionTransaction,
-) -> EcstoreResult<()> {
-    let object =
-        transition_transaction_record_object_name(transaction.transaction_id).map_err(transition_transaction_store_error)?;
-    let data = transaction.encode().map_err(transition_transaction_store_error)?;
-    api.record_durable_ilm_decommission_terminal(&object, &data).await?;
+pub(crate) async fn delete_transition_transaction_record(api: Arc<ECStore>, transaction_id: Uuid) -> EcstoreResult<()> {
+    let object = transition_transaction_record_object_name(transaction_id).map_err(transition_transaction_store_error)?;
     match config_boundary::delete_config(api, &object).await {
         Ok(()) | Err(Error::ConfigNotFound) => Ok(()),
         Err(err) => Err(err),
@@ -821,7 +813,7 @@ pub async fn finalize_missing_transition_transaction_for_operator(
     if probe != TransitionOperatorProbe::Missing {
         return Err(TransitionOperatorError::CandidateNotMissing(probe));
     }
-    delete_transition_transaction_record(api, &transaction)
+    delete_transition_transaction_record(api, transaction_id)
         .await
         .map_err(TransitionOperatorError::Store)
 }
@@ -857,22 +849,22 @@ pub async fn process_transition_transaction_record(
     match transaction.state {
         TransitionTransactionState::Uploaded => {
             delete_transition_remote_candidate(api.clone(), transaction).await?;
-            delete_transition_transaction_record(api, transaction).await?;
+            delete_transition_transaction_record(api, transaction.transaction_id).await?;
             Ok(TransitionTransactionRecoveryOutcome::RemoteCandidateDeleted)
         }
         TransitionTransactionState::CleanupPending => match local_commit_matches_transaction(api.clone(), transaction).await {
             Ok(true) => {
-                delete_transition_transaction_record(api, transaction).await?;
+                delete_transition_transaction_record(api, transaction.transaction_id).await?;
                 Ok(TransitionTransactionRecoveryOutcome::RecordDeleted)
             }
             Ok(false) => {
                 delete_transition_remote_candidate(api.clone(), transaction).await?;
-                delete_transition_transaction_record(api, transaction).await?;
+                delete_transition_transaction_record(api, transaction.transaction_id).await?;
                 Ok(TransitionTransactionRecoveryOutcome::RemoteCandidateDeleted)
             }
             Err(err) if transition_source_is_missing(&err) => {
                 delete_transition_remote_candidate(api.clone(), transaction).await?;
-                delete_transition_transaction_record(api, transaction).await?;
+                delete_transition_transaction_record(api, transaction.transaction_id).await?;
                 Ok(TransitionTransactionRecoveryOutcome::RemoteCandidateDeleted)
             }
             Err(err) => Err(err),
@@ -880,7 +872,7 @@ pub async fn process_transition_transaction_record(
         TransitionTransactionState::LocalCommitStarted => {
             match local_commit_matches_transaction(api.clone(), transaction).await {
                 Ok(true) => {
-                    delete_transition_transaction_record(api, transaction).await?;
+                    delete_transition_transaction_record(api, transaction.transaction_id).await?;
                     Ok(TransitionTransactionRecoveryOutcome::RecordDeleted)
                 }
                 Ok(false) => Ok(TransitionTransactionRecoveryOutcome::Retained),
@@ -889,7 +881,7 @@ pub async fn process_transition_transaction_record(
             }
         }
         TransitionTransactionState::AbortedNoRemote | TransitionTransactionState::Committed => {
-            delete_transition_transaction_record(api, transaction).await?;
+            delete_transition_transaction_record(api, transaction.transaction_id).await?;
             Ok(TransitionTransactionRecoveryOutcome::RecordDeleted)
         }
         TransitionTransactionState::UploadOutcomeUnknown => recover_unknown_upload_outcome(api, transaction).await,
@@ -915,7 +907,7 @@ async fn recover_unknown_upload_outcome(
         .map_err(Error::other)?
     {
         TransitionCandidateProbe::Missing => {
-            delete_transition_transaction_record(api, transaction).await?;
+            delete_transition_transaction_record(api, transaction.transaction_id).await?;
             Ok(TransitionTransactionRecoveryOutcome::RecordDeleted)
         }
         TransitionCandidateProbe::UnversionedPresent => {
@@ -933,7 +925,7 @@ async fn recover_unknown_upload_outcome(
             )
             .await
             .map_err(Error::other)?;
-            delete_transition_transaction_record(api, transaction).await?;
+            delete_transition_transaction_record(api, transaction.transaction_id).await?;
             Ok(TransitionTransactionRecoveryOutcome::RemoteCandidateDeleted)
         }
         TransitionCandidateProbe::VersionedPresent(version_id) => {
@@ -966,7 +958,7 @@ async fn cleanup_recovered_unknown_upload_candidate(
         .map_err(transition_transaction_store_error)?;
     save_transition_transaction_record(api.clone(), &cleanup).await?;
     delete_transition_remote_candidate(api.clone(), &cleanup).await?;
-    delete_transition_transaction_record(api, &cleanup).await?;
+    delete_transition_transaction_record(api, cleanup.transaction_id).await?;
     Ok(TransitionTransactionRecoveryOutcome::RemoteCandidateDeleted)
 }
 

crates/ecstore/src/cluster/rpc/peer_rest_client.rs

@@ -85,6 +85,7 @@ const PEER_REST_RECOVERY_MAX_ATTEMPTS: u32 = 60;
 const PEER_REST_RECOVERY_MAX_BACKOFF: Duration = Duration::from_secs(30);
 const SCANNER_ACTIVITY_MAX_MESSAGE_SIZE: usize = 1024;
 const REPLICATION_STATS_MAX_MESSAGE_SIZE: usize = 8 * 1024 * 1024;
+const BUCKET_METADATA_RELOAD_TIMEOUT: Duration = Duration::from_secs(5);
 
 /// Error for a peer that reported `success = false` without an `error_info` payload.
 ///
@@ -1328,27 +1329,38 @@ impl PeerRestClient {
     }
 
     pub async fn load_bucket_metadata(&self, bucket: &str, scanner_maintenance_change: bool) -> Result<()> {
-        self.finalize_result(
-            async {
-                let mut client = self.get_client().await?;
-                let mut request = Request::new(LoadBucketMetadataRequest {
-                    bucket: bucket.to_string(),
-                    scanner_maintenance_change,
-                });
-                set_tonic_mutation_body_digest(&mut request)?;
-
-                let response = client.load_bucket_metadata(request).await?.into_inner();
-                if !response.success {
-                    if let Some(msg) = response.error_info {
-                        return Err(Error::other(msg));
-                    }
-                    return Err(peer_failure_without_details("load_bucket_metadata", Some(bucket)));
-                }
-                Ok(())
+        let result = tokio::time::timeout(BUCKET_METADATA_RELOAD_TIMEOUT, async {
+            let result = self.load_bucket_metadata_once(bucket, scanner_maintenance_change).await;
+            if let Err(err) = &result
+                && Self::is_network_like_error(err)
+            {
+                self.prepare_retry().await;
+                return self.load_bucket_metadata_once(bucket, scanner_maintenance_change).await;
             }
-            .await,
-        )
+            result
+        })
         .await
+        .unwrap_or_else(|_| Err(Error::other(format!("load_bucket_metadata({bucket}) timed out"))));
+        self.finalize_result(result).await
+    }
+
+    async fn load_bucket_metadata_once(&self, bucket: &str, scanner_maintenance_change: bool) -> Result<()> {
+        let mut client = self.get_client().await?;
+        let mut request = Request::new(LoadBucketMetadataRequest {
+            bucket: bucket.to_string(),
+            scanner_maintenance_change,
+        });
+        set_tonic_mutation_body_digest(&mut request)?;
+        request.set_timeout(BUCKET_METADATA_RELOAD_TIMEOUT);
+
+        let response = client.load_bucket_metadata(request).await?.into_inner();
+        if !response.success {
+            if let Some(msg) = response.error_info {
+                return Err(Error::other(msg));
+            }
+            return Err(peer_failure_without_details("load_bucket_metadata", Some(bucket)));
+        }
+        Ok(())
     }
 
     pub async fn delete_bucket_metadata(&self, bucket: &str) -> Result<()> {

crates/ecstore/src/config/com.rs

@@ -406,25 +406,6 @@ where
     Ok(data)
 }
 
-pub(crate) async fn read_config_limited_preserve_empty<S>(api: Arc<S>, file: &str, max_bytes: usize) -> Result<Vec<u8>>
-where
-    S: EcstoreObjectIO,
-{
-    let (data, _obj) = read_config_limited_preserve_empty_with_metadata(api, file, max_bytes).await?;
-    Ok(data)
-}
-
-pub(crate) async fn read_config_limited_preserve_empty_with_metadata<S>(
-    api: Arc<S>,
-    file: &str,
-    max_bytes: usize,
-) -> Result<(Vec<u8>, ObjectInfo)>
-where
-    S: EcstoreObjectIO,
-{
-    read_config_with_metadata_inner(api, file, &ObjectOptions::default(), true, Some(max_bytes)).await
-}
-
 /// Read an existing config object without treating an empty payload as absent.
 /// Callers that validate their own payload format need to distinguish corruption
 /// from `ConfigNotFound`.

crates/ecstore/src/set_disk/ops/object.rs

@@ -37,7 +37,7 @@ use crate::bucket::lifecycle::{
     transition_transaction::{
         TransitionRemoteVersion, TransitionSourceIdentity, TransitionSourceVersionMode, TransitionTransaction,
         TransitionTransactionInit, TransitionTransactionState, delete_transition_transaction_record,
-        load_transition_transaction_record, save_transition_transaction_record,
+        save_transition_transaction_record,
     },
 };
 use crate::bucket::quota::reservation;
@@ -4246,12 +4246,7 @@ fn record_transition_uploaded_save_attempt(transaction: &TransitionTransaction,
 
 async fn delete_transition_transaction_if_available(api: Option<&Arc<ECStore>>, transaction_id: Uuid) -> Result<()> {
     if let Some(api) = api {
-        let transaction = match load_transition_transaction_record(api.clone(), transaction_id).await {
-            Ok(transaction) => transaction,
-            Err(Error::ConfigNotFound) => return Ok(()),
-            Err(err) => return Err(err),
-        };
-        return delete_transition_transaction_record(api.clone(), &transaction).await;
+        return delete_transition_transaction_record(api.clone(), transaction_id).await;
     }
     Ok(())
 }

rustfs/src/app/bucket_usecase.rs

@@ -513,13 +513,15 @@ fn sr_bucket_meta_item(bucket: String, item_type: &str) -> SRBucketMeta {
     }
 }
 
-fn notify_bucket_metadata_reload(
+async fn notify_bucket_metadata_reload(
     bucket: String,
     operation: &'static str,
     request_context: Option<request_context::RequestContext>,
     scanner_maintenance_change: bool,
 ) {
     record_local_scanner_maintenance_reload(&bucket, scanner_maintenance_change);
+    // Keep reload detached across request cancellation, but wait before a healthy peer can serve the previous config.
+    let (completed_tx, completed_rx) = tokio::sync::oneshot::channel();
     spawn_background_with_context(request_context, async move {
         if let Some(notification_sys) = current_notification_system() {
             let result = if scanner_maintenance_change {
@@ -531,7 +533,9 @@ fn notify_bucket_metadata_reload(
                 warn!(bucket = %bucket, error = %err, "failed to notify peers after {operation}");
             }
         }
+        let _ = completed_tx.send(());
     });
+    let _ = completed_rx.await;
 }
 
 fn record_local_scanner_maintenance_reload(bucket: &str, scanner_maintenance_change: bool) {
@@ -1476,7 +1480,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "delete bucket encryption", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "delete bucket encryption", request_context, false).await;
 
         let item = sr_bucket_meta_item(bucket.clone(), "sse-config");
         if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1508,7 +1512,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "delete bucket cors", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "delete bucket cors", request_context, false).await;
 
         let item = sr_bucket_meta_item(bucket.clone(), "cors-config");
         if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1540,7 +1544,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "delete bucket lifecycle", request_context, true);
+        notify_bucket_metadata_reload(bucket.clone(), "delete bucket lifecycle", request_context, true).await;
 
         let item = sr_bucket_meta_item(bucket.clone(), "lc-config");
         if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1572,7 +1576,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "delete bucket policy", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "delete bucket policy", request_context, false).await;
 
         let item = sr_bucket_meta_item(bucket.clone(), "policy");
         if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1630,7 +1634,7 @@ impl DefaultBucketUsecase {
         }
         drop(targets_guard);
 
-        notify_bucket_metadata_reload(bucket.clone(), "delete bucket replication", request_context, true);
+        notify_bucket_metadata_reload(bucket.clone(), "delete bucket replication", request_context, true).await;
 
         let item = sr_bucket_meta_item(bucket.clone(), "replication-config");
         if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1655,7 +1659,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "delete bucket tagging", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "delete bucket tagging", request_context, false).await;
 
         let item = sr_bucket_meta_item(bucket.clone(), "tags");
         if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1688,7 +1692,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "delete public access block", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "delete public access block", request_context, false).await;
 
         Ok(S3Response::with_status(DeletePublicAccessBlockOutput::default(), StatusCode::NO_CONTENT))
     }
@@ -2143,7 +2147,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "put bucket encryption", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "put bucket encryption", request_context, false).await;
 
         let mut item = sr_bucket_meta_item(bucket.clone(), "sse-config");
         item.sse_config = Some(
@@ -2222,7 +2226,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "put bucket lifecycle", request_context, true);
+        notify_bucket_metadata_reload(bucket.clone(), "put bucket lifecycle", request_context, true).await;
 
         let mut item = sr_bucket_meta_item(bucket.clone(), "lc-config");
         item.expiry_lc_config =
@@ -2307,7 +2311,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "put bucket notification", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "put bucket notification", request_context, false).await;
 
         let region = resolve_notification_region(self.global_region(), request_region);
         let notify = current_notify_interface_for_context(self.context.as_deref());
@@ -2412,7 +2416,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "put bucket policy", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "put bucket policy", request_context, false).await;
 
         let mut item = sr_bucket_meta_item(bucket.clone(), "policy");
         item.policy = Some(serde_json::from_str(&policy).map_err(|e| s3_error!(InvalidArgument, "parse policy failed {:?}", e))?);
@@ -2447,7 +2451,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "put bucket cors", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "put bucket cors", request_context, false).await;
 
         let mut item = sr_bucket_meta_item(bucket.clone(), "cors-config");
         item.cors =
@@ -2491,7 +2495,7 @@ impl DefaultBucketUsecase {
             .map_err(ApiError::from)?;
         drop(targets_guard);
 
-        notify_bucket_metadata_reload(bucket.clone(), "put bucket replication", request_context, true);
+        notify_bucket_metadata_reload(bucket.clone(), "put bucket replication", request_context, true).await;
 
         let mut item = sr_bucket_meta_item(bucket.clone(), "replication-config");
         item.replication_config = Some(
@@ -2531,7 +2535,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "put public access block", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "put public access block", request_context, false).await;
 
         Ok(S3Response::new(PutPublicAccessBlockOutput::default()))
     }
@@ -2560,7 +2564,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "put bucket tagging", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "put bucket tagging", request_context, false).await;
 
         let mut item = sr_bucket_meta_item(bucket.clone(), "tags");
         item.tags = Some(serialize_config(&tagging).and_then(|bytes| String::from_utf8(bytes).map_err(to_internal_error))?);
@@ -2593,7 +2597,7 @@ impl DefaultBucketUsecase {
             .await
             .map_err(ApiError::from)?;
 
-        notify_bucket_metadata_reload(bucket.clone(), "put bucket versioning", request_context, false);
+        notify_bucket_metadata_reload(bucket.clone(), "put bucket versioning", request_context, false).await;
 
         let mut item = sr_bucket_meta_item(bucket.clone(), "version-config");
         item.versioning = Some(
@@ -3044,7 +3048,7 @@ mod tests {
                 "{method} should identify the bucket metadata operation in reload logs"
             );
             let expected_reload = format!(
-                "notify_bucket_metadata_reload(bucket.clone(), \"{operation}\", request_context, {scanner_maintenance_change});"
+                "notify_bucket_metadata_reload(bucket.clone(), \"{operation}\", request_context, {scanner_maintenance_change}).await;"
             );
             assert!(
                 body.contains(&expected_reload),