Merge branch 'main' into overtrue/docs-1923-free-version-disposition

* test(kms): bound Vault failover progress wait

* ci(nightly): honor manual dispatch ref

* test(kms): preserve Vault worker failures

* test(kms): validate Vault circuit recovery

.github/workflows/nightly-gnu.yml

@@ -39,11 +39,10 @@ jobs:
     env:
       FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: "true"
     steps:
-      - name: Checkout main branch
+      - name: Checkout repository
         uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
         with:
           persist-credentials: false
-          ref: main
 
       - name: Setup Rust environment
         uses: ./.github/actions/setup
@@ -89,11 +88,10 @@ jobs:
       # either casing.
       NO_PROXY: 127.0.0.1,localhost
     steps:
-      - name: Checkout main branch
+      - name: Checkout repository
         uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
         with:
           persist-credentials: false
-          ref: main
 
       - name: Setup Rust environment
         uses: ./.github/actions/setup
@@ -178,11 +176,10 @@ jobs:
       FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: "true"
       NO_PROXY: 127.0.0.1,localhost
     steps:
-      - name: Checkout main branch
+      - name: Checkout repository
         uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
         with:
           persist-credentials: false
-          ref: main
 
       - name: Setup Rust environment
         uses: ./.github/actions/setup

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

@@ -1089,9 +1089,7 @@ impl PeerRestClient {
                     .await?
                     .max_decoding_message_size(BACKGROUND_HEAL_STATUS_MAX_MESSAGE_SIZE);
                 let response = match client
-                    .background_heal_status(Request::new(BackgroundHealStatusRequest {
-                        protocol_version: rustfs_protos::BACKGROUND_HEAL_STATUS_PROTOCOL_VERSION,
-                    }))
+                    .background_heal_status(Request::new(BackgroundHealStatusRequest::default()))
                     .await
                 {
                     Ok(response) => response.into_inner(),

crates/ecstore/src/core/pools.rs

@@ -13,6 +13,8 @@
 // limitations under the License.
 
 use crate::bucket::replication::replication_state_from_filemeta;
+#[cfg(test)]
+use crate::bucket::utils::is_meta_bucketname;
 use crate::bucket::versioning_sys::BucketVersioningSys;
 use crate::bucket::{
     lifecycle::{
@@ -1347,6 +1349,53 @@ fn should_cleanup_decommission_source_entry(decommissioned: usize, total_version
     decommissioned.saturating_add(expired) == total_versions
 }
 
+const DECOMMISSION_FREE_VERSION_MIGRATED_REASON: &str = "tier_free_version_migrated";
+const DECOMMISSION_FREE_VERSION_CONSUMED_REASON: &str = "tier_free_version_already_consumed";
+const DECOMMISSION_FREE_VERSION_RETAINED_REASON: &str = "tier_free_version_migration_failed";
+const DECOMMISSION_FREE_VERSION_SWEEP_REASON: &str = "tier_free_version_unresolved_after_decommission";
+const DECOMMISSION_FREE_VERSION_DISPOSITION_REASON: &str = "tier_free_version_disposition_recorded";
+
+#[derive(Debug, Default, Clone, Copy, PartialEq, Eq)]
+struct DecommissionFreeVersionDisposition {
+    migrated: usize,
+    consumed: usize,
+    retained: usize,
+}
+
+impl DecommissionFreeVersionDisposition {
+    fn record_migrated(&mut self) {
+        self.migrated += 1;
+    }
+
+    fn record_consumed(&mut self) {
+        self.consumed += 1;
+    }
+
+    fn record_retained(&mut self) {
+        self.retained += 1;
+    }
+
+    fn total(self) -> usize {
+        self.migrated.saturating_add(self.consumed).saturating_add(self.retained)
+    }
+}
+
+enum DecommissionFreeVersionAttempt {
+    Migrated,
+    Consumed,
+    CapacityFailure(Error),
+    Retry(Error),
+}
+
+fn classify_decommission_free_version_attempt(result: Result<()>) -> DecommissionFreeVersionAttempt {
+    match result {
+        Ok(()) => DecommissionFreeVersionAttempt::Migrated,
+        Err(err) if is_decommission_copy_cleanup_safe_error(&err) => DecommissionFreeVersionAttempt::Consumed,
+        Err(err) if is_decommission_target_capacity_error(&err) => DecommissionFreeVersionAttempt::CapacityFailure(err),
+        Err(err) => DecommissionFreeVersionAttempt::Retry(err),
+    }
+}
+
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
 #[allow(
     dead_code,
@@ -2615,8 +2664,12 @@ fn determine_decommission_final_state(items_failed: usize, was_cancelled: bool)
     }
 }
 
-fn decommission_remaining_version_count(total_versions: usize, expired: usize) -> usize {
-    total_versions.saturating_sub(expired)
+fn decommission_remaining_version_count(versions: &[rustfs_filemeta::FileInfo], expired: usize) -> usize {
+    versions
+        .iter()
+        .filter(|version| !version.tier_free_version())
+        .count()
+        .saturating_sub(expired)
 }
 
 fn should_skip_decommission_delete_marker(
@@ -2679,6 +2732,7 @@ fn decommission_remote_tiered_opts(
         user_defined: version.metadata.clone(),
         src_pool_idx,
         data_movement: true,
+        incl_free_versions: version.tier_free_version(),
         include_part_checksums: true,
         http_preconditions: Some(crate::data_movement::data_movement_target_precondition()),
         expected_bucket_incarnation_id,
@@ -3822,6 +3876,7 @@ impl ECStore {
 
         let mut decommissioned: usize = 0;
         let mut expired: usize = 0;
+        let mut free_version_disposition = DecommissionFreeVersionDisposition::default();
         let mut cleanup_preflight_allowed_missing = Vec::new();
 
         for version in fivs.versions.iter() {
@@ -3830,6 +3885,115 @@ impl ECStore {
             }
             decommission_cancel_signal_result(rx.is_cancelled())?;
 
+            if version.tier_free_version() {
+                let version_id = version.version_id.map(|v| v.to_string());
+                let mut migration_error = None;
+                let mut migrated = false;
+                let mut consumed = false;
+                let mut capacity_failure = false;
+                for _ in 0..3 {
+                    match classify_decommission_free_version_attempt(
+                        run_decommission_side_effect(&rx, &operation_gate, || async {
+                            self.decommission_tiered_object(
+                                bucket.as_str(),
+                                &version.name,
+                                version,
+                                &decommission_remote_tiered_opts(
+                                    version,
+                                    version_id.clone(),
+                                    idx,
+                                    expected_bucket_incarnation_id,
+                                ),
+                            )
+                            .await
+                        })
+                        .await,
+                    ) {
+                        DecommissionFreeVersionAttempt::Migrated => {
+                            migrated = true;
+                            migration_error = None;
+                            break;
+                        }
+                        DecommissionFreeVersionAttempt::Consumed => {
+                            consumed = true;
+                            migration_error = None;
+                            break;
+                        }
+                        DecommissionFreeVersionAttempt::CapacityFailure(err) => {
+                            capacity_failure = true;
+                            migration_error = Some(err);
+                            break;
+                        }
+                        DecommissionFreeVersionAttempt::Retry(err) => migration_error = Some(err),
+                    }
+                }
+
+                {
+                    let mut pool_meta = self.pool_meta.write().await;
+                    ensure_decommission_generation(&pool_meta, idx, generation)?;
+                    if let Err(err) = count_decommission_item(&mut pool_meta, idx, 0, !migrated && !consumed) {
+                        return Err(with_decommission_entry_context(
+                            "count_decommission_item",
+                            bucket.as_str(),
+                            entry.name.as_str(),
+                            err,
+                        ));
+                    }
+                }
+
+                if migrated || consumed {
+                    decommissioned += 1;
+                    cleanup_preflight_allowed_missing.push(data_movement::source_cleanup_version_identity(version));
+                }
+                if migrated {
+                    free_version_disposition.record_migrated();
+                } else if consumed {
+                    free_version_disposition.record_consumed();
+                } else {
+                    free_version_disposition.record_retained();
+                }
+
+                debug!(
+                    event = EVENT_DECOMMISSION_ENTRY,
+                    component = LOG_COMPONENT_ECSTORE,
+                    subsystem = LOG_SUBSYSTEM_POOLS,
+                    pool_index = idx,
+                    bucket = %bucket,
+                    object = %version.name,
+                    version_id = ?version_id,
+                    result = ?migration_error,
+                    reason = if migrated {
+                        DECOMMISSION_FREE_VERSION_MIGRATED_REASON
+                    } else if consumed {
+                        DECOMMISSION_FREE_VERSION_CONSUMED_REASON
+                    } else {
+                        DECOMMISSION_FREE_VERSION_RETAINED_REASON
+                    },
+                    state = if migrated {
+                        "free_version_migrated"
+                    } else if consumed {
+                        "free_version_consumed"
+                    } else {
+                        "free_version_retained"
+                    },
+                    "Decommission free-version disposition recorded"
+                );
+
+                if capacity_failure {
+                    return Err(with_decommission_entry_context(
+                        "decommission_tier_free_version",
+                        bucket.as_str(),
+                        version.name.as_str(),
+                        migration_error.expect("capacity failure must retain its error"),
+                    ));
+                }
+
+                if !migrated && !consumed {
+                    break;
+                }
+                continue;
+            }
+
             if run_decommission_side_effect(&rx, &operation_gate, || async {
                 should_skip_lifecycle_for_data_movement(
                     self.clone(),
@@ -3850,7 +4014,7 @@ impl ECStore {
                 continue;
             }
 
-            let remaining_versions = decommission_remaining_version_count(fivs.versions.len(), expired);
+            let remaining_versions = decommission_remaining_version_count(&fivs.versions, expired);
             if should_skip_decommission_delete_marker(version, remaining_versions, replication_config.is_some()) {
                 //
                 decommissioned += 1;
@@ -4129,6 +4293,24 @@ impl ECStore {
             }
         }
 
+        if free_version_disposition.total() > 0 {
+            debug!(
+                event = EVENT_DECOMMISSION_ENTRY,
+                component = LOG_COMPONENT_ECSTORE,
+                subsystem = LOG_SUBSYSTEM_POOLS,
+                pool_index = idx,
+                bucket = %bucket,
+                object = %entry.name,
+                free_versions_migrated = free_version_disposition.migrated,
+                free_versions_consumed = free_version_disposition.consumed,
+                free_versions_retained = free_version_disposition.retained,
+                free_versions_total = free_version_disposition.total(),
+                reason = DECOMMISSION_FREE_VERSION_DISPOSITION_REASON,
+                state = "free_version_disposition",
+                "Decommission free-version disposition summary"
+            );
+        }
+
         if should_cleanup_decommission_source_entry(decommissioned, fivs.versions.len(), expired) {
             if bucket_incarnation_fence.as_ref().is_some_and(|guard| guard.is_lock_lost()) {
                 return Err(Error::other("decommission bucket incarnation fence was lost before source cleanup"));
@@ -4298,6 +4480,39 @@ impl ECStore {
         .await
     }
 
+    #[cfg(test)]
+    pub(crate) async fn decommission_entry_for_test_with_bucket_incarnation(
+        self: &Arc<Self>,
+        idx: usize,
+        entry: MetaCacheEntry,
+        bucket: String,
+        set: Arc<SetDisks>,
+    ) -> Result<()> {
+        let expected_bucket_incarnation_id = if is_meta_bucketname(&bucket) {
+            None
+        } else {
+            Some(self.bucket_incarnation_id_from_disk(&bucket).await?)
+        };
+        self.decommission_entry(
+            CancellationToken::new(),
+            idx,
+            OffsetDateTime::now_utc(),
+            entry,
+            bucket,
+            set,
+            None,
+            None,
+            None,
+            expected_bucket_incarnation_id,
+        )
+        .await
+    }
+
+    #[cfg(test)]
+    pub(crate) async fn check_after_decommission_for_test(self: &Arc<Self>, idx: usize) -> Result<()> {
+        self.check_after_decommission(idx).await
+    }
+
     #[tracing::instrument(skip(self, rx))]
     async fn decommission_pool(
         self: &Arc<Self>,
@@ -5162,6 +5377,7 @@ impl ECStore {
                 let lifecycle_config_cb = lifecycle_config.clone();
                 let object_lock_config_cb = object_lock_config.clone();
                 let store = Arc::clone(self);
+                let set_cb = Arc::clone(set);
                 let callback_rx_cb = callback_rx.clone();
 
                 let callback: ListCallback = Arc::new(move |entry: MetaCacheEntry| {
@@ -5171,6 +5387,7 @@ impl ECStore {
                     let lifecycle_config = lifecycle_config_cb.clone();
                     let object_lock_config = object_lock_config_cb.clone();
                     let store = Arc::clone(&store);
+                    let set = Arc::clone(&set_cb);
                     let callback_rx = callback_rx_cb.clone();
                     Box::pin(async move {
                         if callback_rx.is_cancelled() {
@@ -5185,11 +5402,14 @@ impl ECStore {
                             return;
                         }
 
-                        let fivs = match load_decommission_entry_versions(
+                        let fivs = match load_decommission_entry_exact_versions(
+                            &set,
                             &entry,
                             &bucket_name,
                             "check_after_decommission.file_info_versions",
-                        ) {
+                        )
+                        .await
+                        {
                             Ok(fivs) => fivs,
                             Err(err) => {
                                 let mut first_err = entry_error.lock().await;
@@ -5202,7 +5422,23 @@ impl ECStore {
                         };
 
                         let mut remaining = 0;
-                        for version in &fivs.versions {
+                        for version in fivs.versions.iter().chain(fivs.free_versions.iter()) {
+                            if version.tier_free_version() {
+                                remaining += 1;
+                                debug!(
+                                    event = EVENT_DECOMMISSION_ENTRY,
+                                    component = LOG_COMPONENT_ECSTORE,
+                                    subsystem = LOG_SUBSYSTEM_POOLS,
+                                    pool_index = idx,
+                                    bucket = %bucket_name,
+                                    object = %entry.name,
+                                    version_id = ?version.version_id,
+                                    reason = DECOMMISSION_FREE_VERSION_SWEEP_REASON,
+                                    state = "free_version_retained",
+                                    "Decommission final sweep retained a free version"
+                                );
+                                continue;
+                            }
                             if version.deleted {
                                 continue;
                             }
@@ -5317,13 +5553,6 @@ mod tests {
         assert_eq!(determine_decommission_final_state(0, true), DecommissionFinalState::Failed);
     }
 
-    #[test]
-    fn decommission_remaining_version_count_excludes_only_expired_versions() {
-        assert_eq!(decommission_remaining_version_count(1, 0), 1);
-        assert_eq!(decommission_remaining_version_count(2, 1), 1);
-        assert_eq!(decommission_remaining_version_count(1, 1), 0);
-    }
-
     #[test]
     fn lifecycle_action_removes_data_movement_version_rejects_delete_marker_action() {
         assert!(!lifecycle_action_removes_data_movement_version(IlmAction::DeleteAction));
@@ -5380,6 +5609,21 @@ mod tests {
         )));
     }
 
+    #[test]
+    fn decommission_free_version_attempt_treats_missing_source_as_consumed() {
+        let attempt =
+            classify_decommission_free_version_attempt(Err(Error::ObjectNotFound("bucket".to_string(), "object".to_string())));
+
+        assert!(matches!(attempt, DecommissionFreeVersionAttempt::Consumed));
+    }
+
+    #[test]
+    fn decommission_free_version_attempt_preserves_capacity_failure() {
+        let attempt = classify_decommission_free_version_attempt(Err(Error::DiskFull));
+
+        assert!(matches!(attempt, DecommissionFreeVersionAttempt::CapacityFailure(Error::DiskFull)));
+    }
+
     #[test]
     fn decommission_delete_marker_copy_error_rejects_data_movement_overwrite() {
         let err = Error::DataMovementOverwriteErr("bucket".to_string(), "object".to_string(), "version".to_string());
@@ -5561,6 +5805,12 @@ mod tests {
         assert!(opts.include_part_checksums);
         assert!(opts.http_preconditions.is_some());
         assert_eq!(opts.expected_bucket_incarnation_id, Some(incarnation));
+        assert!(!opts.incl_free_versions);
+
+        let mut free_version = version;
+        free_version.set_tier_free_version();
+        let free_opts = decommission_remote_tiered_opts(&free_version, Some("free-version-id".to_string()), 9, Some(incarnation));
+        assert!(free_opts.incl_free_versions);
     }
 
     #[test]

crates/ecstore/src/data_movement/mod.rs

@@ -1950,6 +1950,19 @@ mod tests {
         assert!(source_cleanup_versions_match_with_allowed_missing(&expected, &current, &allowed_missing));
     }
 
+    #[test]
+    fn test_decommission_cleanup_preflight_accepts_migrated_free_version_consumed_from_source() {
+        let migrated = cleanup_test_file_info("object.txt", Uuid::from_u128(1), "migrated");
+        let mut free_version = cleanup_test_file_info("object.txt", Uuid::from_u128(2), "tier-cleanup");
+        free_version.deleted = true;
+        free_version.set_tier_free_version();
+        let expected = cleanup_test_versions(vec![migrated.clone(), free_version.clone()]);
+        let current = cleanup_test_versions(vec![migrated]);
+        let allowed_missing = vec![source_cleanup_version_identity(&free_version)];
+
+        assert!(source_cleanup_versions_match_with_allowed_missing(&expected, &current, &allowed_missing));
+    }
+
     #[test]
     fn test_decommission_cleanup_preflight_rejects_unexpected_missing_version() {
         let migrated = cleanup_test_file_info("object.txt", Uuid::from_u128(1), "migrated");

crates/ecstore/src/set_disk/mod.rs

@@ -227,6 +227,70 @@ pub(super) fn restore_commit_operation_id_from_metadata(metadata: &HashMap<Strin
     restore_operation_id_from_metadata(metadata)
 }
 
+async fn inspect_decommission_tier_free_version_target(
+    disk: &DiskStore,
+    bucket: &str,
+    object: &str,
+    source: &FileInfo,
+) -> Result<bool> {
+    let raw = match disk.read_xl(bucket, object, false).await {
+        Ok(raw) => raw,
+        Err(DiskError::FileNotFound | DiskError::FileVersionNotFound | DiskError::VolumeNotFound) => return Ok(false),
+        Err(err) => return Err(err.into()),
+    };
+    let meta = FileMeta::load(&raw.buf)?;
+    let source_version_id = source.version_id.filter(|version_id| !version_id.is_nil());
+    let mut matching_count = 0;
+    let mut all_matching_versions_equivalent = true;
+    for existing in meta
+        .versions
+        .iter()
+        .filter(|version| version.header.version_id.filter(|version_id| !version_id.is_nil()) == source_version_id)
+    {
+        matching_count += 1;
+        let existing = existing.into_fileinfo(bucket, object, true)?;
+        existing.validate_for_metadata_read()?;
+        if !existing.tier_free_version() || !crate::store::tiered_data_movement_source_matches(source, &existing)? {
+            all_matching_versions_equivalent = false;
+        }
+    }
+    if matching_count == 0 {
+        return Ok(false);
+    }
+    if matching_count == 1 && all_matching_versions_equivalent {
+        return Ok(true);
+    }
+
+    Err(StorageError::DataMovementOverwriteErr(
+        bucket.to_owned(),
+        object.to_owned(),
+        source_version_id.map(|version_id| version_id.to_string()).unwrap_or_default(),
+    )
+    .into())
+}
+
+fn ensure_decommission_tier_free_version_commit_fence(bucket: &str, object: &str, opts: &ObjectOptions) -> Result<()> {
+    if opts
+        .namespace_lock_fence
+        .as_ref()
+        .is_some_and(NamespaceLockFence::is_lock_lost)
+        || opts
+            .bucket_lifecycle_lock_fence
+            .as_ref()
+            .is_some_and(NamespaceLockFence::is_lock_lost)
+    {
+        return Err(StorageError::NamespaceLockQuorumUnavailable {
+            mode: "decommission_tier_free_version_commit",
+            bucket: bucket.to_string(),
+            object: object.to_string(),
+            required: 1,
+            achieved: 0,
+        });
+    }
+
+    Ok(())
+}
+
 impl SetDisks {
     pub(super) async fn require_current_restore_operation_id(
         &self,
@@ -4665,6 +4729,98 @@ fn resolve_delete_version_state(opts: &ObjectOptions, goi: &ObjectInfo, version_
 }
 
 impl SetDisks {
+    /// Publish an internal tier free-version record without changing its
+    /// delete-marker shape or remote-tier identity. The caller holds the
+    /// source and target object locks; a write quorum is required before the
+    /// source cleanup may remove the original record.
+    #[tracing::instrument(skip(self, fi, opts))]
+    pub(crate) async fn decommission_tier_free_version(
+        &self,
+        bucket: &str,
+        object: &str,
+        fi: &FileInfo,
+        opts: &ObjectOptions,
+    ) -> Result<()> {
+        if !fi.deleted || !fi.tier_free_version() {
+            return Err(Error::other("decommission tier free-version write requires a free version record"));
+        }
+        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;
+
+        self.validate_decommission_tier_free_version_target(bucket, object, fi)
+            .await?;
+        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;
+
+        let disks = self.disks.read().await.clone();
+        let write_quorum = self.default_write_quorum();
+        let futures = disks.into_iter().map(|disk| {
+            let file_info = fi.clone();
+            async move {
+                if let Some(disk) = disk {
+                    disk.write_metadata("", bucket, object, file_info).await
+                } else {
+                    Err(DiskError::DiskNotFound)
+                }
+            }
+        });
+
+        let mut errs = Vec::new();
+        for result in join_all(futures).await {
+            match result {
+                Ok(_) => errs.push(None),
+                Err(err) => errs.push(Some(err)),
+            }
+        }
+
+        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;
+
+        resolve_tiered_decommission_write_quorum_result(&errs, write_quorum, bucket, object)
+    }
+
+    pub(crate) async fn validate_decommission_tier_free_version_target(
+        &self,
+        bucket: &str,
+        object: &str,
+        fi: &FileInfo,
+    ) -> Result<()> {
+        // The caller holds the source and target object locks. Inspect every
+        // target disk before an idempotent return or metadata fan-out so a
+        // sub-quorum conflict cannot be hidden by a successful quorum.
+        let disks = self.disks.read().await.clone();
+        let preflight = disks
+            .iter()
+            .flatten()
+            .map(|disk| inspect_decommission_tier_free_version_target(disk, bucket, object, fi));
+        for result in join_all(preflight).await {
+            result?;
+        }
+        Ok(())
+    }
+
+    pub(crate) async fn has_decommission_tier_free_version_write_quorum(
+        &self,
+        bucket: &str,
+        object: &str,
+        fi: &FileInfo,
+        opts: &ObjectOptions,
+    ) -> Result<bool> {
+        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;
+        let disks = self.disks.read().await.clone();
+        let preflight = disks.iter().map(|disk| async {
+            match disk {
+                Some(disk) => inspect_decommission_tier_free_version_target(disk, bucket, object, fi).await,
+                None => Ok(false),
+            }
+        });
+        let mut equivalent = 0;
+        for result in join_all(preflight).await {
+            if result? {
+                equivalent += 1;
+            }
+        }
+        ensure_decommission_tier_free_version_commit_fence(bucket, object, opts)?;
+        Ok(equivalent >= self.default_write_quorum())
+    }
+
     #[tracing::instrument(skip(self, fi, opts))]
     pub(crate) async fn decommission_tiered_object(
         &self,
@@ -9887,6 +10043,147 @@ mod tests {
         assert_ne!(updated.erasure.distribution, original.erasure.distribution);
     }
 
+    #[tokio::test]
+    async fn decommission_tier_free_version_preserves_remote_identity() {
+        let set_disks = make_local_bucket_test_set_disks().await;
+        let bucket = "free-version-decommission";
+        let object = "object.txt";
+        set_disks
+            .make_bucket(bucket, &MakeBucketOptions::default())
+            .await
+            .expect("target bucket should exist before free-version migration");
+        let version_id = Uuid::new_v4();
+        let mut free_version = FileInfo {
+            name: object.to_string(),
+            volume: bucket.to_string(),
+            version_id: Some(version_id),
+            mod_time: Some(time::OffsetDateTime::now_utc()),
+            deleted: true,
+            transition_tier: "WARM-TIER".to_string(),
+            transitioned_objname: "remote/object".to_string(),
+            ..Default::default()
+        };
+        free_version.set_tier_free_version();
+        // Decoded free versions always carry the on-disk free-version
+        // suffix alongside the in-memory tier marker; mirror that here so
+        // the record satisfies delete-marker metadata validation.
+        rustfs_utils::http::metadata_compat::insert_str(
+            &mut free_version.metadata,
+            rustfs_utils::http::metadata_compat::SUFFIX_FREE_VERSION,
+            String::new(),
+        );
+
+        set_disks
+            .decommission_tier_free_version(bucket, object, &free_version, &ObjectOptions::default())
+            .await
+            .expect("free-version metadata should reach the target quorum");
+        set_disks
+            .decommission_tier_free_version(bucket, object, &free_version, &ObjectOptions::default())
+            .await
+            .expect("replaying the same free-version metadata should be idempotent");
+
+        let versions = set_disks
+            .load_file_info_versions_exact(bucket, object)
+            .await
+            .expect("migrated free-version metadata should decode")
+            .expect("migrated free-version metadata should exist");
+        let migrated = versions
+            .versions
+            .iter()
+            .find(|version| version.version_id == Some(version_id))
+            .expect("free version should be present on the target");
+
+        assert_eq!(
+            versions
+                .versions
+                .iter()
+                .filter(|version| version.version_id == Some(version_id))
+                .count(),
+            1
+        );
+        assert!(migrated.tier_free_version());
+        assert_eq!(migrated.transition_tier, "WARM-TIER");
+        assert_eq!(migrated.transitioned_objname, "remote/object");
+    }
+
+    #[tokio::test]
+    async fn decommission_tier_free_version_resume_requires_write_quorum() {
+        let set_disks = make_local_bucket_test_set_disks_with_drive_count(4).await;
+        let bucket = "free-version-decommission-resume";
+        let object = "object.txt";
+        set_disks
+            .make_bucket(bucket, &MakeBucketOptions::default())
+            .await
+            .expect("target bucket should exist before free-version migration");
+        let mut free_version = FileInfo {
+            name: object.to_string(),
+            volume: bucket.to_string(),
+            version_id: Some(Uuid::new_v4()),
+            mod_time: Some(time::OffsetDateTime::now_utc()),
+            deleted: true,
+            transition_tier: "WARM-TIER".to_string(),
+            transitioned_objname: "remote/object".to_string(),
+            ..Default::default()
+        };
+        free_version.set_tier_free_version();
+        // Decoded free versions always carry the on-disk free-version
+        // suffix alongside the in-memory tier marker; mirror that here so
+        // the record satisfies delete-marker metadata validation.
+        rustfs_utils::http::metadata_compat::insert_str(
+            &mut free_version.metadata,
+            rustfs_utils::http::metadata_compat::SUFFIX_FREE_VERSION,
+            String::new(),
+        );
+        let opts = ObjectOptions::default();
+
+        let disks = set_disks.get_disks_internal().await;
+        for disk in disks.iter().take(2).flatten() {
+            disk.write_metadata("", bucket, object, free_version.clone())
+                .await
+                .expect("partial first attempt should leave equivalent metadata");
+        }
+        assert!(
+            !set_disks
+                .has_decommission_tier_free_version_write_quorum(bucket, object, &free_version, &opts)
+                .await
+                .expect("partial target metadata should remain valid"),
+            "write-quorum-minus-one must not be accepted as an idempotent migration"
+        );
+
+        disks[2]
+            .as_ref()
+            .expect("third target disk should be online")
+            .write_metadata("", bucket, object, free_version.clone())
+            .await
+            .expect("third equivalent target write should complete quorum");
+        assert!(
+            set_disks
+                .has_decommission_tier_free_version_write_quorum(bucket, object, &free_version, &opts)
+                .await
+                .expect("write-quorum target metadata should remain valid")
+        );
+    }
+
+    #[test]
+    fn decommission_tier_free_version_commit_rejects_lost_fence() {
+        let opts = ObjectOptions {
+            namespace_lock_fence: Some(NamespaceLockFence::lost_for_test()),
+            ..Default::default()
+        };
+
+        let err = ensure_decommission_tier_free_version_commit_fence("bucket", "object", &opts)
+            .expect_err("lost target lock must fail the free-version commit");
+        assert!(matches!(
+            err,
+            Error::NamespaceLockQuorumUnavailable {
+                mode: "decommission_tier_free_version_commit",
+                required: 1,
+                achieved: 0,
+                ..
+            }
+        ));
+    }
+
     #[test]
     fn test_resolve_tiered_decommission_write_quorum_result_allows_successful_quorum() {
         let errs = vec![None, None, Some(DiskError::DiskNotFound), None];

crates/ecstore/src/store/init.rs

@@ -553,6 +553,8 @@ mod tests {
         should_retry_format_load, should_retry_local_decommission_resume, wait_for_local_decommission_resume_delay,
     };
     #[cfg(feature = "test-util")]
+    use crate::disk::DiskAPI;
+    #[cfg(feature = "test-util")]
     use crate::{
         bucket::lifecycle::{
             lifecycle::{TRANSITION_PENDING, TransitionOptions},
@@ -1307,6 +1309,77 @@ mod tests {
         });
     }
 
+    #[cfg(feature = "test-util")]
+    async fn seed_transitioned_free_version(
+        ctx: &Arc<crate::runtime::instance::InstanceContext>,
+        store: &Arc<crate::store::ECStore>,
+        bucket: &str,
+        object: &str,
+    ) -> (uuid::Uuid, uuid::Uuid) {
+        let tier_name = format!("DECOMFREE{}", uuid::Uuid::new_v4().simple());
+        register_mock_tier(&ctx.tier_config_mgr(), &tier_name).await;
+
+        let mut reader = PutObjReader::from_vec(b"transitioned source bytes".to_vec());
+        let source = store.pools[0]
+            .put_object(
+                bucket,
+                object,
+                &mut reader,
+                &ObjectOptions {
+                    versioned: true,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("write transitioned decommission source");
+        let source_version = source.version_id.expect("transitioned source must be versioned");
+        store.pools[0]
+            .transition_object(
+                bucket,
+                object,
+                &ObjectOptions {
+                    versioned: true,
+                    version_id: Some(source_version.to_string()),
+                    transition: TransitionOptions {
+                        status: TRANSITION_PENDING.to_string(),
+                        tier: tier_name,
+                        etag: source.etag.clone().expect("transitioned source must have an ETag"),
+                        ..Default::default()
+                    },
+                    mod_time: source.mod_time,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("transition source before decommission");
+        store.pools[0]
+            .delete_object(
+                bucket,
+                object,
+                ObjectOptions {
+                    versioned: true,
+                    version_id: Some(source_version.to_string()),
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("delete transitioned source version");
+
+        let versions = store.pools[0]
+            .get_disks_by_key(object)
+            .load_file_info_versions_exact(bucket, object)
+            .await
+            .expect("source versions should decode after transition delete")
+            .expect("source free version should remain after transition delete");
+        let free_version = versions
+            .versions
+            .iter()
+            .find(|version| version.tier_free_version())
+            .and_then(|version| version.version_id)
+            .expect("transition delete should create a free version");
+        (source_version, free_version)
+    }
+
     async fn write_decommission_test_multipart_source(
         store: &Arc<crate::store::ECStore>,
         pool_idx: usize,
@@ -3846,6 +3919,383 @@ mod tests {
         shutdown.cancel();
     }
 
+    #[cfg(feature = "test-util")]
+    #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+    #[serial_test::serial(storage_class_env)]
+    async fn decommission_entry_skips_cleanup_only_marker_when_free_version_is_present() {
+        let temp_dir = tempfile::tempdir().expect("create free-version decommission store dir");
+        let (ctx, store, shutdown) =
+            without_storage_class_env(build_isolated_test_store(temp_dir.path(), "decommission-free-marker", &[4, 4])).await;
+        crate::bucket::metadata_sys::init_bucket_metadata_sys(store.clone(), Vec::new()).await;
+        let bucket = format!("decom-free-marker-{}", uuid::Uuid::new_v4());
+        let object = "free-marker-object";
+        store
+            .make_bucket(&bucket, &MakeBucketOptions::default())
+            .await
+            .expect("create free-version decommission bucket");
+        let (_, free_version) = seed_transitioned_free_version(&ctx, &store, &bucket, object).await;
+        let source_free = store.pools[0]
+            .get_disks_by_key(object)
+            .load_file_info_versions_exact(&bucket, object)
+            .await
+            .expect("source free-version metadata should decode")
+            .and_then(|versions| {
+                versions
+                    .versions
+                    .into_iter()
+                    .find(|version| version.version_id == Some(free_version) && version.tier_free_version())
+            })
+            .expect("source free-version identity should be present before decommission");
+        let mut target_reader = PutObjReader::from_vec(b"target ordinary bytes".to_vec());
+        let target_w = store.pools[1]
+            .put_object(
+                &bucket,
+                object,
+                &mut target_reader,
+                &ObjectOptions {
+                    versioned: true,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("write unrelated target version");
+        let target_w_version = target_w.version_id.expect("target version should have an id");
+        let marker = store.pools[0]
+            .delete_object(
+                &bucket,
+                object,
+                ObjectOptions {
+                    versioned: true,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("write cleanup-only delete marker");
+        assert!(marker.delete_marker);
+
+        mark_test_pool_decommissioning(&store, 0).await;
+        let source_set = store.pools[0].get_disks_by_key(object);
+        store
+            .decommission_entry_for_test_with_bucket_incarnation(
+                0,
+                MetaCacheEntry {
+                    name: object.to_string(),
+                    ..Default::default()
+                },
+                bucket.clone(),
+                source_set.clone(),
+            )
+            .await
+            .expect("real decommission entry should migrate the free version");
+
+        let target_versions = store.pools[1]
+            .get_disks_by_key(object)
+            .load_file_info_versions_exact(&bucket, object)
+            .await
+            .expect("target versions should decode")
+            .expect("target free version should be present");
+        assert!(
+            target_versions
+                .versions
+                .iter()
+                .any(|version| { version.version_id == Some(free_version) && version.tier_free_version() })
+        );
+        let migrated_free = target_versions
+            .versions
+            .iter()
+            .find(|version| version.version_id == Some(free_version) && version.tier_free_version())
+            .expect("migrated free-version identity should remain readable from target disks");
+        assert!(
+            crate::store::tiered_data_movement_source_matches(&source_free, migrated_free)
+                .expect("migrated free-version identity should decode")
+        );
+        let retained_w = target_versions
+            .versions
+            .iter()
+            .find(|version| version.version_id == Some(target_w_version))
+            .expect("unrelated target version should remain");
+        assert!(!retained_w.deleted && !retained_w.tier_free_version());
+        assert_eq!(retained_w.size, target_w.size);
+        assert_eq!(retained_w.get_etag(), target_w.etag);
+        assert!(
+            target_versions
+                .versions
+                .iter()
+                .all(|version| { version.tier_free_version() || !version.deleted })
+        );
+        assert!(
+            source_set
+                .load_file_info_versions_exact(&bucket, object)
+                .await
+                .expect("source versions should be readable after cleanup")
+                .is_none(),
+            "successful free-version migration should permit source cleanup"
+        );
+        let (heal_versions, _, _) = store
+            .heal_walk_versions_page(1, 0, &bucket, "", None, 2, 16, true)
+            .await
+            .expect("heal walk should decode the migrated free version");
+        let free_version_string = free_version.to_string();
+        let healed_free = heal_versions
+            .iter()
+            .find(|version| version.version_id.as_deref() == Some(free_version_string.as_str()))
+            .expect("heal walk should surface the migrated free version");
+        let healed_info = healed_free
+            .lifecycle_object_info
+            .as_ref()
+            .expect("heal walk should retain lifecycle identity for the migrated free version");
+        assert!(healed_info.transitioned_object.free_version);
+        assert_eq!(healed_info.transitioned_object.tier, source_free.transition_tier);
+        assert_eq!(healed_info.transitioned_object.name, source_free.transitioned_objname);
+        shutdown.cancel();
+    }
+
+    #[cfg(feature = "test-util")]
+    #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+    #[serial_test::serial(storage_class_env)]
+    async fn decommission_entry_allows_free_version_consumed_before_source_lock() {
+        let temp_dir = tempfile::tempdir().expect("create consumed free-version store dir");
+        let (ctx, store, shutdown) =
+            without_storage_class_env(build_isolated_test_store(temp_dir.path(), "decommission-free-consumed", &[4, 4])).await;
+        crate::bucket::metadata_sys::init_bucket_metadata_sys(store.clone(), Vec::new()).await;
+        let bucket = format!("decom-free-consumed-{}", uuid::Uuid::new_v4());
+        let object = "free-consumed-object";
+        store
+            .make_bucket(&bucket, &MakeBucketOptions::default())
+            .await
+            .expect("create consumed free-version bucket");
+        let (_, free_version) = seed_transitioned_free_version(&ctx, &store, &bucket, object).await;
+
+        mark_test_pool_decommissioning(&store, 0).await;
+        let source_set = store.pools[0].get_disks_by_key(object);
+        let barrier = crate::store::object::DecommissionFreeVersionSourceRaceBarrier::install(&bucket, object);
+        let decommission = tokio::spawn({
+            let store = store.clone();
+            let bucket = bucket.clone();
+            let source_set = source_set.clone();
+            async move {
+                store
+                    .decommission_entry_for_test_with_bucket_incarnation(
+                        0,
+                        MetaCacheEntry {
+                            name: object.to_string(),
+                            ..Default::default()
+                        },
+                        bucket,
+                        source_set,
+                    )
+                    .await
+            }
+        });
+
+        barrier.wait_until_paused().await;
+        store.pools[0]
+            .delete_object(
+                &bucket,
+                object,
+                ObjectOptions {
+                    versioned: true,
+                    version_id: Some(free_version.to_string()),
+                    incl_free_versions: true,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("lifecycle should consume the source free version before decommission locks it");
+        assert!(
+            source_set
+                .load_file_info_versions_exact(&bucket, object)
+                .await
+                .expect("consumed source metadata should remain readable")
+                .is_none(),
+            "the lifecycle delete should remove the source free version"
+        );
+
+        barrier.release();
+        decommission
+            .await
+            .expect("decommission task should join")
+            .expect("a concurrently consumed free version should not fail source cleanup");
+        assert!(
+            store.pools[1]
+                .get_disks_by_key(object)
+                .load_file_info_versions_exact(&bucket, object)
+                .await
+                .expect("target metadata should remain readable")
+                .is_none(),
+            "an already consumed free version should not be recreated on the target"
+        );
+        shutdown.cancel();
+    }
+
+    #[cfg(feature = "test-util")]
+    #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+    #[serial_test::serial(storage_class_env)]
+    async fn decommission_entry_rejects_subquorum_free_version_conflict_and_retains_source() {
+        let temp_dir = tempfile::tempdir().expect("create sub-quorum free-version store dir");
+        let (ctx, store, shutdown) =
+            without_storage_class_env(build_isolated_test_store(temp_dir.path(), "decommission-free-conflict", &[4, 4])).await;
+        crate::bucket::metadata_sys::init_bucket_metadata_sys(store.clone(), Vec::new()).await;
+        let bucket = format!("decom-free-conflict-{}", uuid::Uuid::new_v4());
+        let object = "free-conflict-object";
+        store
+            .make_bucket(&bucket, &MakeBucketOptions::default())
+            .await
+            .expect("create sub-quorum conflict bucket");
+        let (_, free_version) = seed_transitioned_free_version(&ctx, &store, &bucket, object).await;
+        let source_free = store.pools[0]
+            .get_disks_by_key(object)
+            .load_file_info_versions_exact(&bucket, object)
+            .await
+            .expect("source free version should decode before crash replay setup")
+            .and_then(|versions| {
+                versions
+                    .versions
+                    .into_iter()
+                    .find(|version| version.version_id == Some(free_version))
+            })
+            .expect("source free version should be available for crash replay setup");
+
+        let mut target_reader = PutObjReader::from_vec(b"target ordinary bytes".to_vec());
+        let target = store.pools[1]
+            .put_object(
+                &bucket,
+                object,
+                &mut target_reader,
+                &ObjectOptions {
+                    versioned: true,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("seed ordinary target version");
+        let target_version = target.version_id.expect("target version must have an ID");
+        let target_disks = store.pools[1].get_disks_by_key(object).disks.read().await.clone();
+        for disk in target_disks.iter().skip(1) {
+            disk.as_ref()
+                .expect("target crash replay quorum disk should be online")
+                .write_metadata("", &bucket, object, source_free.clone())
+                .await
+                .expect("seed an equivalent free version on the target quorum");
+        }
+        let conflict_path = temp_dir
+            .path()
+            .join(format!("pool1/set0/disk0/{bucket}/{object}/{STORAGE_FORMAT_FILE}"));
+        let encoded = tokio::fs::read(&conflict_path)
+            .await
+            .expect("target metadata should be readable");
+        let mut metadata = FileMeta::load(&encoded).expect("target metadata should decode");
+        let target_index = metadata
+            .versions
+            .iter()
+            .position(|version| version.header.version_id == Some(target_version))
+            .expect("target version should be present on the conflict disk");
+        let mut target_meta = metadata.versions[target_index]
+            .parse_version_meta()
+            .expect("target version metadata should decode");
+        target_meta
+            .object
+            .as_mut()
+            .expect("target conflict must remain an ordinary object")
+            .version_id = Some(free_version);
+        metadata.versions[target_index] = target_meta.try_into().expect("conflict metadata should encode");
+        let expected_conflict_meta = metadata.versions[target_index].meta.clone();
+        let expected_conflict = metadata.versions[target_index]
+            .into_fileinfo(&bucket, object, true)
+            .expect("conflict metadata should decode as an ordinary object");
+        let duplicate_free: rustfs_filemeta::FileMetaShallowVersion = rustfs_filemeta::FileMetaVersion::from(source_free.clone())
+            .try_into()
+            .expect("duplicate free metadata should encode");
+        metadata.versions.insert(target_index, duplicate_free);
+        tokio::fs::write(&conflict_path, metadata.marshal_msg().expect("conflict metadata should encode"))
+            .await
+            .expect("write sub-quorum conflict metadata");
+
+        mark_test_pool_decommissioning(&store, 0).await;
+        let source_set = store.pools[0].get_disks_by_key(object);
+        store
+            .decommission_entry_for_test_with_bucket_incarnation(
+                0,
+                MetaCacheEntry {
+                    name: object.to_string(),
+                    ..Default::default()
+                },
+                bucket.clone(),
+                source_set.clone(),
+            )
+            .await
+            .expect("conflicted decommission entry should retain the source and retry later");
+
+        let source_versions = source_set
+            .load_file_info_versions_exact(&bucket, object)
+            .await
+            .expect("retained source versions should decode")
+            .expect("source free version should be retained after conflict");
+        assert!(
+            source_versions
+                .versions
+                .iter()
+                .any(|version| { version.version_id == Some(free_version) && version.tier_free_version() })
+        );
+        let post_encoded = tokio::fs::read(&conflict_path)
+            .await
+            .expect("conflict metadata should remain readable");
+        let post_metadata = FileMeta::load(&post_encoded).expect("post-conflict metadata should decode");
+        let same_id = post_metadata
+            .versions
+            .iter()
+            .filter(|version| version.header.version_id == Some(free_version))
+            .collect::<Vec<_>>();
+        assert_eq!(same_id.len(), 2, "conflict metadata should retain both same-ID records");
+        assert_eq!(same_id.iter().filter(|version| version.header.free_version()).count(), 1);
+        assert_eq!(same_id.iter().filter(|version| !version.header.free_version()).count(), 1);
+        let post_conflict = same_id
+            .into_iter()
+            .find(|version| !version.header.free_version())
+            .expect("ordinary conflict version must remain addressable by the source ID");
+        let post_conflict_info = post_conflict
+            .into_fileinfo(&bucket, object, true)
+            .expect("post-conflict ordinary metadata should decode");
+        assert!(!post_conflict_info.deleted && !post_conflict_info.tier_free_version());
+        assert_eq!(post_conflict.meta, expected_conflict_meta);
+        assert_eq!(post_conflict_info.size, expected_conflict.size);
+        assert_eq!(post_conflict_info.data_dir, expected_conflict.data_dir);
+        assert_eq!(post_conflict_info.metadata, expected_conflict.metadata);
+        assert_eq!(post_conflict_info.get_etag(), expected_conflict.get_etag());
+
+        for disk_index in 0..4 {
+            let target_path = temp_dir
+                .path()
+                .join(format!("pool1/set0/disk{disk_index}/{bucket}/{object}/{STORAGE_FORMAT_FILE}"));
+            let target_encoded = tokio::fs::read(&target_path)
+                .await
+                .expect("target metadata should remain readable");
+            let target_meta = FileMeta::load(&target_encoded).expect("target metadata should decode");
+            let same_id = target_meta
+                .versions
+                .iter()
+                .filter(|version| version.header.version_id == Some(free_version))
+                .collect::<Vec<_>>();
+            if disk_index == 0 {
+                assert_eq!(same_id.len(), 2);
+                assert!(same_id[0].header.free_version());
+                assert!(!same_id[1].header.free_version());
+            } else {
+                assert_eq!(same_id.len(), 1);
+                assert!(same_id[0].header.free_version());
+            }
+        }
+        let sweep_err = store
+            .check_after_decommission_for_test(0)
+            .await
+            .expect_err("final sweep must report the retained free version");
+        assert!(
+            sweep_err.to_string().contains("version(s) were found"),
+            "unexpected final sweep error: {sweep_err}"
+        );
+        shutdown.cancel();
+    }
+
     #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
     #[serial_test::serial(storage_class_env)]
     async fn versioned_batch_delete_marker_skips_decommission_source() {

crates/ecstore/src/store/mod.rs

@@ -151,7 +151,7 @@ pub(crate) mod init_format;
 pub(crate) mod list_objects;
 mod multipart;
 mod object;
-pub(crate) use object::{ObjectLockDiagGuard, SourceCleanupMutationFence};
+pub(crate) use object::{ObjectLockDiagGuard, SourceCleanupMutationFence, tiered_data_movement_source_matches};
 pub use object::{
     PrepareSelectObjectSnapshotError, PreparedGetObjectReader, SelectObjectSnapshot, SelectObjectSnapshotReadError,
     SnapshotConsistencyError,

crates/ecstore/src/store/object.rs

@@ -490,6 +490,82 @@ fn decommission_mutation_fence_for_test(
         .map(|hook| hook.fence.clone())
 }
 
+#[cfg(test)]
+struct DecommissionFreeVersionSourceRaceState {
+    bucket: String,
+    object: String,
+    arrived: tokio::sync::Notify,
+    release: tokio::sync::Notify,
+}
+
+#[cfg(test)]
+pub(crate) struct DecommissionFreeVersionSourceRaceBarrier {
+    state: Arc<DecommissionFreeVersionSourceRaceState>,
+}
+
+#[cfg(test)]
+static DECOMMISSION_FREE_VERSION_SOURCE_RACE_BARRIER: std::sync::OnceLock<
+    std::sync::Mutex<Option<Arc<DecommissionFreeVersionSourceRaceState>>>,
+> = std::sync::OnceLock::new();
+
+#[cfg(test)]
+impl DecommissionFreeVersionSourceRaceBarrier {
+    pub(crate) fn install(bucket: &str, object: &str) -> Self {
+        let state = Arc::new(DecommissionFreeVersionSourceRaceState {
+            bucket: bucket.to_string(),
+            object: object.to_string(),
+            arrived: tokio::sync::Notify::new(),
+            release: tokio::sync::Notify::new(),
+        });
+        let mut slot = DECOMMISSION_FREE_VERSION_SOURCE_RACE_BARRIER
+            .get_or_init(|| std::sync::Mutex::new(None))
+            .lock()
+            .expect("decommission free-version source race barrier should not poison");
+        assert!(slot.is_none(), "decommission free-version source race barrier must be unique");
+        *slot = Some(Arc::clone(&state));
+        Self { state }
+    }
+
+    pub(crate) async fn wait_until_paused(&self) {
+        tokio::time::timeout(Duration::from_secs(30), self.state.arrived.notified())
+            .await
+            .expect("decommission should pause before acquiring the free-version source lock");
+    }
+
+    pub(crate) fn release(&self) {
+        self.state.release.notify_one();
+    }
+}
+
+#[cfg(test)]
+impl Drop for DecommissionFreeVersionSourceRaceBarrier {
+    fn drop(&mut self) {
+        self.state.release.notify_one();
+        let mut slot = DECOMMISSION_FREE_VERSION_SOURCE_RACE_BARRIER
+            .get_or_init(|| std::sync::Mutex::new(None))
+            .lock()
+            .expect("decommission free-version source race barrier should not poison");
+        if slot.as_ref().is_some_and(|state| Arc::ptr_eq(state, &self.state)) {
+            *slot = None;
+        }
+    }
+}
+
+#[cfg(test)]
+async fn pause_decommission_free_version_before_source_lock(bucket: &str, object: &str) {
+    let state = DECOMMISSION_FREE_VERSION_SOURCE_RACE_BARRIER
+        .get_or_init(|| std::sync::Mutex::new(None))
+        .lock()
+        .expect("decommission free-version source race barrier should not poison")
+        .as_ref()
+        .filter(|state| state.bucket == bucket && state.object == object)
+        .cloned();
+    if let Some(state) = state {
+        state.arrived.notify_one();
+        state.release.notified().await;
+    }
+}
+
 pub(crate) struct SourceCleanupMutationFence {
     guard: ObjectLockDiagGuard,
     source_lock_covered: bool,
@@ -1494,13 +1570,15 @@ fn is_equivalent_data_movement_tiered_object(source: &rustfs_filemeta::FileInfo,
         && source_actual_size == target_actual_size
 }
 
-fn tiered_data_movement_source_matches(
+pub(crate) fn tiered_data_movement_source_matches(
     expected: &rustfs_filemeta::FileInfo,
     current: &rustfs_filemeta::FileInfo,
 ) -> Result<bool> {
     let expected_backend = crate::services::tier::tier::tier_destination_id_from_metadata(&expected.metadata)?;
     let current_backend = crate::services::tier::tier::tier_destination_id_from_metadata(&current.metadata)?;
     Ok(expected.version_id == current.version_id
+        && expected.deleted == current.deleted
+        && expected.tier_free_version() == current.tier_free_version()
         && expected.data_dir == current.data_dir
         && expected.mod_time == current.mod_time
         && expected.size == current.size
@@ -1514,6 +1592,15 @@ fn tiered_data_movement_source_matches(
         && expected_backend == current_backend)
 }
 
+fn decommission_free_version_overwrite_error(bucket: &str, object: &str, version_id: Option<Uuid>) -> Error {
+    StorageError::DataMovementOverwriteErr(
+        bucket.to_owned(),
+        object.to_owned(),
+        version_id.map(|id| id.to_string()).unwrap_or_default(),
+    )
+    .into()
+}
+
 fn should_check_data_movement_resume_target(src_pool_idx: usize, target_pool_idx: usize) -> bool {
     target_pool_idx != src_pool_idx
 }
@@ -2221,6 +2308,23 @@ impl ECStore {
         )
     }
 
+    async fn has_equivalent_data_movement_tier_free_version(
+        &self,
+        bucket: &str,
+        object: &str,
+        source: &rustfs_filemeta::FileInfo,
+        opts: &ObjectOptions,
+        target_pool_idx: usize,
+    ) -> Result<bool> {
+        let pool = self
+            .pools
+            .get(target_pool_idx)
+            .ok_or_else(|| Error::other(format!("invalid tiered data movement target pool {target_pool_idx}")))?;
+        pool.get_disks_by_key(object)
+            .has_decommission_tier_free_version_write_quorum(bucket, object, source, opts)
+            .await
+    }
+
     fn resolve_decommission_target_pool_idx_result(result: Result<usize>, bucket: &str, object: &str) -> Result<usize> {
         result.map_err(|err| Error::other(format!("failed to select decommission target pool for {bucket}/{object}: {err}")))
     }
@@ -2240,6 +2344,10 @@ impl ECStore {
         check_put_object_args(bucket, object)?;
 
         let mut opts = opts.clone();
+        let is_free_version = fi.tier_free_version();
+        if is_free_version {
+            opts.incl_free_versions = true;
+        }
         let bucket_incarnation_fence = if is_meta_bucketname(bucket) {
             None
         } else {
@@ -2277,6 +2385,10 @@ impl ECStore {
                 &object,
             )?
         };
+        #[cfg(test)]
+        if is_free_version {
+            pause_decommission_free_version_before_source_lock(bucket, logical_object).await;
+        }
         let _object_guards = self
             .acquire_data_movement_object_write_locks(bucket, &object, opts.src_pool_idx, idx, &mut opts)
             .await?;
@@ -2294,7 +2406,7 @@ impl ECStore {
                 versions
                     .versions
                     .iter()
-                    .find(|current| current.version_id == fi.version_id && !current.tier_free_version())
+                    .find(|current| current.version_id == fi.version_id && current.tier_free_version() == is_free_version)
             })
             .ok_or_else(|| to_object_err(StorageError::FileNotFound, vec![bucket, object.as_str()]))?;
         if !tiered_data_movement_source_matches(fi, current_source)? {
@@ -2309,24 +2421,40 @@ impl ECStore {
                 .get_available_pool_idx_excluding(bucket, &object, fi.size, opts.src_pool_idx)
                 .await;
             let target_pool_idx = resolve_data_movement_resume_target_pool(idx, resume_target_pool_idx, opts.src_pool_idx);
+            if is_free_version && target_pool_idx == opts.src_pool_idx {
+                return Err(Error::DiskFull);
+            }
+            let equivalent = if is_free_version {
+                self.has_equivalent_data_movement_tier_free_version(bucket, &object, &fi, &opts, target_pool_idx)
+                    .await?
+            } else {
+                self.has_equivalent_data_movement_tiered_object(bucket, &object, &fi, &opts, target_pool_idx)
+                    .await?
+            };
+            if equivalent {
+                return Ok(());
+            }
+
+            return Err(decommission_free_version_overwrite_error(bucket, &object, fi.version_id));
+        }
+
+        let result = if is_free_version {
             if self
-                .has_equivalent_data_movement_tiered_object(bucket, &object, &fi, &opts, target_pool_idx)
+                .has_equivalent_data_movement_tier_free_version(bucket, &object, &fi, &opts, idx)
                 .await?
             {
                 return Ok(());
             }
-
-            return Err(StorageError::DataMovementOverwriteErr(
-                bucket.to_owned(),
-                object.to_owned(),
-                opts.version_id.clone().unwrap_or_default(),
-            ));
-        }
-
-        let result = self.pools[idx]
-            .get_disks_by_key(&object)
-            .decommission_tiered_object(bucket, &object, &fi, &opts)
-            .await;
+            self.pools[idx]
+                .get_disks_by_key(&object)
+                .decommission_tier_free_version(bucket, &object, &fi, &opts)
+                .await
+        } else {
+            self.pools[idx]
+                .get_disks_by_key(&object)
+                .decommission_tiered_object(bucket, &object, &fi, &opts)
+                .await
+        };
         if matches!(result, Err(Error::PreconditionFailed)) {
             if self
                 .has_equivalent_data_movement_tiered_object(bucket, &object, &fi, &opts, idx)

crates/filemeta/src/filemeta.rs

@@ -90,6 +90,21 @@ fn legacy_data_key_for_version(version_id: Option<Uuid>) -> Option<String> {
 pub const TRANSITION_COMPLETE: &str = "complete";
 pub const TRANSITION_PENDING: &str = "pending";
 
+/// xl.meta key marking a tier free-version record.
+///
+/// A free version is a delete-marker-shaped cleanup hint appended by
+/// [`MetaObject::delete_version`] when a version whose remote transition
+/// completed is removed from xl.meta; it carries the remote tier identity for
+/// an idempotent remote delete and is never a user-visible version
+/// (`num_versions` excludes it). While the record exists it is consumed by the
+/// lifecycle free-version recovery scan and the usage scanner, which re-enqueue
+/// the pending remote delete, and by heal metadata walks. On S3 and lifecycle
+/// delete paths the same obligation is also carried by a committed tier-journal
+/// entry; deletes without such an entry (for example a removed version whose
+/// transition state decodes as unknown) rely on this record alone until the
+/// worker removes it after a successful remote delete. Decommission preserves
+/// the record and its remote identity on the target pool before source cleanup
+/// — see docs/architecture/decommission-compatibility.md.
 pub const FREE_VERSION: &str = "free-version";
 
 pub const TRANSITION_STATUS: &str = "transition-status";
@@ -447,6 +462,10 @@ impl FileMeta {
         };
 
         if let Some(fidx) = existing_idx {
+            let existing = self.versions[fidx].parse_version_meta()?;
+            if existing.free_version() != version.free_version() {
+                return Err(Error::other("cannot replace a free version with a non-free version"));
+            }
             return self.set_idx(fidx, version);
         }
 

crates/filemeta/src/filemeta/version.rs

@@ -2725,6 +2725,15 @@ impl MetaObject {
         self.meta_sys.retain(|k, _| !k.starts_with("X-Amz-Restore"));
     }
 
+    /// Builds the free-version cleanup record appended when a transitioned
+    /// version is removed from xl.meta. The record keeps the remote tier
+    /// identity so the lifecycle worker can issue the idempotent remote delete
+    /// and only then remove the record; until then the recovery scan and the
+    /// usage scanner keep re-enqueueing it. S3 and lifecycle deletes also
+    /// persist a committed tier-journal entry for the same remote delete. The
+    /// decommission path copies this record unchanged before source cleanup,
+    /// including when the transition state is unknown — see
+    /// docs/architecture/decommission-compatibility.md.
     pub fn init_free_version(&self, fi: &FileInfo) -> Result<(FileMetaVersion, bool)> {
         if fi.skip_tier_free_version() {
             return Ok((FileMetaVersion::default(), false));

crates/heal/src/heal/erasure_healer.rs

@@ -13,10 +13,10 @@
 // limitations under the License.
 
 use crate::heal::{
-    progress::{HealProgress, add_bytes, increment_counter},
+    progress::HealProgress,
     resume::{
-        CheckpointManager, CheckpointObjectOutcome, CheckpointObjectOutcomeRecord, ReplacementTargetIdentity, ResumeManager,
-        ResumeUtils, compose_key, replacement_target_identities_match,
+        CheckpointManager, ReplacementTargetIdentity, ResumeManager, ResumeUtils, compose_key,
+        replacement_target_identities_match,
     },
     storage::{HealStorageAPI, next_heal_listing_token},
     task::{demote_to_debug_when, is_missing_object_dir_heal_result, take_failure_log_sample},
@@ -410,9 +410,6 @@ impl ErasureSetHealer {
                 && state.successful_objects == 0
                 && state.failed_objects == 0
                 && state.skipped_objects == 0
-                && state.skipped_new_versions == 0
-                && state.skipped_ilm_expired == 0
-                && state.processed_bytes == 0
             {
                 // schedule_retry persists the authoritative resume reset before
                 // resetting the checkpoint. Reapply the checkpoint reset after
@@ -477,23 +474,6 @@ impl ErasureSetHealer {
 
         // 2. initialize progress
         self.initialize_progress(buckets, &state).await;
-        let (baseline_known, baseline_count, baseline_size, baseline_generation) = {
-            let baseline = self.progress.read().await;
-            (
-                baseline.baseline_known,
-                baseline.objects_total_count,
-                baseline.objects_total_size,
-                baseline.baseline_generation,
-            )
-        };
-        if baseline_known {
-            resume_manager
-                .set_progress_baseline(baseline_count, baseline_size, baseline_generation)
-                .await?;
-            checkpoint_manager
-                .set_progress_baseline(baseline_count, baseline_size, baseline_generation)
-                .await?;
-        }
 
         // 3. continue from checkpoint
         let current_bucket_index = checkpoint.current_bucket_index;
@@ -503,66 +483,12 @@ impl ErasureSetHealer {
         let mut successful_objects = state.successful_objects;
         let mut failed_objects = state.failed_objects;
         let mut skipped_objects = state.skipped_objects;
-        let checkpoint_has_progress = checkpoint.baseline_known
-            || checkpoint.successful_objects > 0
-            || checkpoint.failed_object_count > 0
-            || checkpoint.skipped_object_count > 0
-            || checkpoint.skipped_new_versions > 0
-            || checkpoint.skipped_ilm_expired > 0
-            || checkpoint.processed_bytes > 0
-            || checkpoint.total_objects > 0
-            || checkpoint.total_bytes > 0
-            || checkpoint.baseline_generation.is_some()
-            || checkpoint.counter_unknown;
-        let checkpoint_generation_mismatch = checkpoint.baseline_known && checkpoint.baseline_generation != baseline_generation;
-        let mut restored_counter_unknown = state.counter_unknown || checkpoint.counter_unknown;
-        if checkpoint_has_progress {
-            successful_objects = checkpoint.successful_objects;
-            failed_objects = checkpoint.failed_object_count;
-            skipped_objects = checkpoint.skipped_object_count;
-            let restored_processed_objects = successful_objects
-                .checked_add(failed_objects)
-                .and_then(|value| value.checked_add(skipped_objects))
-                .and_then(|value| value.checked_add(checkpoint.skipped_new_versions))
-                .and_then(|value| value.checked_add(checkpoint.skipped_ilm_expired));
-            let checkpoint_counter_overflow = restored_processed_objects.is_none();
-            restored_counter_unknown |= checkpoint_counter_overflow;
-            processed_objects = restored_processed_objects.unwrap_or(u64::MAX);
-            let mut progress = self.progress.write().await;
-            progress.objects_scanned = processed_objects;
-            progress.objects_healed = successful_objects;
-            progress.objects_failed = failed_objects;
-            progress.skipped_objects = skipped_objects;
-            progress.skipped_new_versions = checkpoint.skipped_new_versions;
-            progress.skipped_ilm_expired = checkpoint.skipped_ilm_expired;
-            if checkpoint.baseline_known && !checkpoint_generation_mismatch {
-                progress.objects_total_count = checkpoint.total_objects;
-                progress.objects_total_size = checkpoint.total_bytes;
-                progress.baseline_generation = checkpoint.baseline_generation;
-                progress.baseline_known = true;
-            }
-            progress.bytes_processed = checkpoint.processed_bytes;
-            progress.counter_unknown = state.counter_unknown || checkpoint.counter_unknown;
-            progress.refresh_progress_percentage();
-            if checkpoint_generation_mismatch || checkpoint_counter_overflow || progress.counter_unknown {
-                progress.mark_unknown();
-            }
-        }
-        if checkpoint_generation_mismatch {
-            restored_counter_unknown = true;
-        }
-        if restored_counter_unknown {
-            checkpoint_manager.mark_counter_unknown().await?;
-            resume_manager.mark_counter_unknown().await?;
-        }
         let mut failed_buckets = 0u64;
 
         // 4. process remaining buckets
         for (bucket_idx, bucket) in buckets.iter().enumerate().skip(current_bucket_index) {
             // check if completed
             if state.completed_buckets.contains(bucket) {
-                checkpoint_manager.complete_bucket(bucket_idx.saturating_add(1)).await?;
-                current_object_index = 0;
                 continue;
             }
 
@@ -590,42 +516,13 @@ impl ErasureSetHealer {
                 return bucket_result;
             }
 
-            // update progress
-            let progress_snapshot = self.progress.read().await;
-            let bytes_processed = progress_snapshot.bytes_processed;
-            let skipped_new_versions = progress_snapshot.skipped_new_versions;
-            let skipped_ilm_expired = progress_snapshot.skipped_ilm_expired;
-            let counter_unknown = progress_snapshot.counter_unknown;
-            drop(progress_snapshot);
-            // The checkpoint is the recovery authority for object progress.
-            // Publish its counters and fence before the resume summary so a
-            // crash between the two stores cannot make recovery select newer
-            // summary bytes with an older checkpoint ledger.
-            if counter_unknown {
-                checkpoint_manager.mark_counter_unknown().await?;
-            }
-            checkpoint_manager
-                .update_progress(successful_objects, failed_objects, skipped_objects, bytes_processed)
-                .await?;
-            checkpoint_manager
-                .set_skipped_version_counts(skipped_new_versions, skipped_ilm_expired)
-                .await?;
+            // update checkpoint position
             checkpoint_manager.update_position(bucket_idx, current_object_index).await?;
+
+            // update progress
             resume_manager
-                .update_progress_with_bytes(
-                    processed_objects,
-                    successful_objects,
-                    failed_objects,
-                    skipped_objects,
-                    bytes_processed,
-                )
+                .update_progress(processed_objects, successful_objects, failed_objects, skipped_objects)
                 .await?;
-            resume_manager
-                .set_skipped_version_counts(skipped_new_versions, skipped_ilm_expired)
-                .await?;
-            if counter_unknown {
-                resume_manager.mark_counter_unknown().await?;
-            }
 
             // check cancel status
             if self.cancel_token.is_cancelled() {
@@ -645,7 +542,6 @@ impl ErasureSetHealer {
             match bucket_result {
                 Ok(_) => {
                     resume_manager.complete_bucket(bucket).await?;
-                    checkpoint_manager.complete_bucket(bucket_idx.saturating_add(1)).await?;
                     debug!(
                         target: "rustfs::heal::erasure_healer",
                         event = EVENT_HEAL_ERASURE_BUCKET_STATE,
@@ -671,9 +567,7 @@ impl ErasureSetHealer {
                         error = %e,
                         "Erasure set bucket heal failed"
                     );
-                    // A single durable cursor and ledger cannot safely preserve
-                    // this bucket while processing a later one.
-                    break;
+                    // continue to next bucket, do not interrupt the whole process
                 }
             }
 
@@ -881,49 +775,20 @@ impl ErasureSetHealer {
 
                 // Per-version dedup identity — the single canonical key.
                 let key = compose_key(&item.name, item.version_id.as_deref());
-                if checkpoint.processed_objects.contains(&key)
-                    || checkpoint.failed_objects.contains(&key)
-                    || checkpoint.skipped_objects.contains(&key)
-                {
+                if checkpoint.processed_objects.contains(&key) || checkpoint.skipped_objects.contains(&key) {
                     continue;
                 }
 
                 if should_skip_new_version(item.mod_time_unix_nanos, started_at_secs) {
-                    let counter_ok = increment_counter(processed_objects);
+                    checkpoint_manager.add_processed_object(key).await?;
+                    *processed_objects = processed_objects.saturating_add(1);
                     completed_in_page = completed_in_page.saturating_add(1);
                     counter!("rustfs_heal_skipped_new_versions_total").increment(1);
-                    let (outcome_record, counter_unknown) = {
+                    {
                         let mut progress = self.progress.write().await;
                         progress.record_skipped_new_version();
                         progress.set_current_object(Some(format!("skipped_new: {bucket}/{}", item.name)));
-                        progress.update_object_progress(
-                            *processed_objects,
-                            *successful_objects,
-                            *failed_objects,
-                            *skipped_objects,
-                            bytes_processed,
-                        );
-                        if !counter_ok {
-                            progress.mark_unknown();
-                        }
-                        (
-                            CheckpointObjectOutcomeRecord {
-                                object: key,
-                                outcome: CheckpointObjectOutcome::Processed,
-                                successful: progress.objects_healed,
-                                failed: progress.objects_failed,
-                                skipped: progress.skipped_objects,
-                                bytes: progress.bytes_processed,
-                                skipped_new_versions: progress.skipped_new_versions,
-                                skipped_ilm_expired: progress.skipped_ilm_expired,
-                                counter_unknown: progress.counter_unknown,
-                            },
-                            progress.counter_unknown,
-                        )
-                    };
-                    checkpoint_manager.record_object_outcome(outcome_record).await?;
-                    if counter_unknown {
-                        resume_manager.mark_counter_unknown().await?;
+                        progress.update_progress(*processed_objects, *successful_objects, *failed_objects, bytes_processed);
                     }
                     debug!(
                         target: "rustfs::heal::erasure_healer",
@@ -955,41 +820,15 @@ impl ErasureSetHealer {
                         )
                         .await?
                 {
-                    let counter_ok = increment_counter(processed_objects);
+                    checkpoint_manager.add_processed_object(key).await?;
+                    *processed_objects = processed_objects.saturating_add(1);
                     completed_in_page = completed_in_page.saturating_add(1);
                     counter!("rustfs_heal_skipped_ilm_expired_total").increment(1);
-                    let (outcome_record, counter_unknown) = {
+                    {
                         let mut progress = self.progress.write().await;
                         progress.record_skipped_ilm_expired();
                         progress.set_current_object(Some(format!("skipped_ilm: {bucket}/{}", item.name)));
-                        progress.update_object_progress(
-                            *processed_objects,
-                            *successful_objects,
-                            *failed_objects,
-                            *skipped_objects,
-                            bytes_processed,
-                        );
-                        if !counter_ok {
-                            progress.mark_unknown();
-                        }
-                        (
-                            CheckpointObjectOutcomeRecord {
-                                object: key,
-                                outcome: CheckpointObjectOutcome::Processed,
-                                successful: progress.objects_healed,
-                                failed: progress.objects_failed,
-                                skipped: progress.skipped_objects,
-                                bytes: progress.bytes_processed,
-                                skipped_new_versions: progress.skipped_new_versions,
-                                skipped_ilm_expired: progress.skipped_ilm_expired,
-                                counter_unknown: progress.counter_unknown,
-                            },
-                            progress.counter_unknown,
-                        )
-                    };
-                    checkpoint_manager.record_object_outcome(outcome_record).await?;
-                    if counter_unknown {
-                        resume_manager.mark_counter_unknown().await?;
+                        progress.update_progress(*processed_objects, *successful_objects, *failed_objects, bytes_processed);
                     }
                     debug!(
                         target: "rustfs::heal::erasure_healer",
@@ -1115,11 +954,11 @@ impl ErasureSetHealer {
 
             while let Some((key, object, version_id, result)) = page_tasks.next().await {
                 let (object_size, result) = result;
-                let mut telemetry_unknown = false;
-                let checkpoint_outcome = match result {
+                match result {
                     Ok(true) => {
-                        telemetry_unknown |= !increment_counter(successful_objects);
-                        telemetry_unknown |= !add_bytes(&mut bytes_processed, object_size);
+                        *successful_objects += 1;
+                        bytes_processed = bytes_processed.saturating_add(object_size);
+                        checkpoint_manager.add_processed_object(key).await?;
                         debug!(
                             target: "rustfs::heal::erasure_healer",
                             event = EVENT_HEAL_ERASURE_OBJECT_STATE,
@@ -1132,11 +971,11 @@ impl ErasureSetHealer {
                             state = "healed",
                             "Erasure set object healed"
                         );
-                        CheckpointObjectOutcome::Processed
                     }
                     Ok(false) => {
-                        telemetry_unknown |= !increment_counter(successful_objects);
-                        telemetry_unknown |= !add_bytes(&mut bytes_processed, object_size);
+                        checkpoint_manager.add_processed_object(key).await?;
+                        *successful_objects += 1;
+                        bytes_processed = bytes_processed.saturating_add(object_size);
                         debug!(
                             target: "rustfs::heal::erasure_healer",
                             event = EVENT_HEAL_ERASURE_OBJECT_STATE,
@@ -1149,12 +988,12 @@ impl ErasureSetHealer {
                             state = "missing_treated_as_ok",
                             "Erasure set missing object treated as ok"
                         );
-                        CheckpointObjectOutcome::Processed
                     }
                     Err(err @ Error::TaskCancelled) | Err(err @ Error::TaskTimeout) => return Err(err),
                     Err(Error::TransientSkip { message }) => {
-                        telemetry_unknown |= !increment_counter(skipped_objects);
-                        telemetry_unknown |= !add_bytes(&mut bytes_processed, object_size);
+                        *skipped_objects += 1;
+                        bytes_processed = bytes_processed.saturating_add(object_size);
+                        checkpoint_manager.add_skipped_object(key).await?;
                         demote_to_debug_when!(!take_failure_log_sample(&mut transient_skip_samples_logged), warn, target: "rustfs::heal::erasure_healer", {
                             event = EVENT_HEAL_ERASURE_OBJECT_STATE,
                             component = LOG_COMPONENT_HEAL,
@@ -1167,11 +1006,11 @@ impl ErasureSetHealer {
                             error = %message,
                             "Erasure set object heal skipped due to transient error"
                         });
-                        CheckpointObjectOutcome::Skipped
                     }
                     Err(err) => {
-                        telemetry_unknown |= !increment_counter(failed_objects);
-                        telemetry_unknown |= !add_bytes(&mut bytes_processed, object_size);
+                        *failed_objects += 1;
+                        bytes_processed = bytes_processed.saturating_add(object_size);
+                        checkpoint_manager.add_failed_object(key).await?;
                         demote_to_debug_when!(!take_failure_log_sample(&mut failure_samples_logged), warn, target: "rustfs::heal::erasure_healer", {
                             event = EVENT_HEAL_ERASURE_OBJECT_STATE,
                             component = LOG_COMPONENT_HEAL,
@@ -1184,43 +1023,15 @@ impl ErasureSetHealer {
                             error = %err,
                             "Erasure set object heal failed"
                         });
-                        CheckpointObjectOutcome::Failed
                     }
-                };
+                }
 
-                telemetry_unknown |= !increment_counter(processed_objects);
+                *processed_objects += 1;
                 completed_in_page += 1;
-                let (outcome_record, counter_unknown) = {
+                {
                     let mut progress = self.progress.write().await;
                     progress.set_current_object(Some(format!("{bucket}/{object}")));
-                    progress.update_object_progress(
-                        *processed_objects,
-                        *successful_objects,
-                        *failed_objects,
-                        *skipped_objects,
-                        bytes_processed,
-                    );
-                    if telemetry_unknown {
-                        progress.mark_unknown();
-                    }
-                    (
-                        CheckpointObjectOutcomeRecord {
-                            object: key,
-                            outcome: checkpoint_outcome,
-                            successful: progress.objects_healed,
-                            failed: progress.objects_failed,
-                            skipped: progress.skipped_objects,
-                            bytes: progress.bytes_processed,
-                            skipped_new_versions: progress.skipped_new_versions,
-                            skipped_ilm_expired: progress.skipped_ilm_expired,
-                            counter_unknown: progress.counter_unknown,
-                        },
-                        progress.counter_unknown,
-                    )
-                };
-                checkpoint_manager.record_object_outcome(outcome_record).await?;
-                if counter_unknown {
-                    resume_manager.mark_counter_unknown().await?;
+                    progress.update_progress(*processed_objects, *successful_objects, *failed_objects, bytes_processed);
                 }
 
                 if completed_in_page.is_multiple_of(100) {
@@ -1230,22 +1041,16 @@ impl ErasureSetHealer {
 
             *current_object_index = global_obj_idx;
 
-            // Persist the checkpoint ledger and page position before exposing
-            // the next resume cursor. A crash before cursor publication keeps
-            // the page identities available for exact-once replay.
-            checkpoint_manager.advance_page(bucket_index, *current_object_index).await?;
+            // Persist the authoritative cursor FIRST (points at the next page
+            // boundary), then prune the per-version dedup sets. Both are
+            // idempotent under crash: heal_object re-heals safely.
+            let next_cursor = if is_truncated { next_token.clone() } else { None };
+            resume_manager.set_resume_cursor(next_cursor.clone()).await?;
+            checkpoint_manager.complete_page(bucket_index, *current_object_index).await?;
             // Check if there are more pages
             if !is_truncated {
                 break;
             }
-            continuation_token = next_heal_listing_token(bucket, "", next_token, is_truncated)?;
-            if continuation_token.is_none() {
-                // A truncated page without a continuation token is terminal.
-                // Retain its ledger until bucket completion is durable.
-                break;
-            }
-            resume_manager.set_resume_cursor(continuation_token.clone()).await?;
-            checkpoint_manager.prune_completed_page().await?;
 
             // Anti-loop guard: an empty page reported as truncated cannot advance
             // the cursor (there is no last identity to move past), so treat it as a
@@ -1264,6 +1069,12 @@ impl ErasureSetHealer {
                 )));
             }
             previous_page_last = page_last;
+
+            continuation_token = next_heal_listing_token(bucket, "", next_token, is_truncated)?;
+            if continuation_token.is_none() {
+                // Truncated but no continuation token: treat as end of listing.
+                break;
+            }
         }
 
         Ok(())
@@ -1272,66 +1083,10 @@ impl ErasureSetHealer {
     /// initialize progress tracking
     async fn initialize_progress(&self, _buckets: &[String], state: &crate::heal::resume::ResumeState) {
         let mut progress = self.progress.write().await;
-        let existing_baseline = (
-            progress.objects_total_count,
-            progress.objects_total_size,
-            progress.baseline_generation,
-            progress.progress_state,
-            progress.baseline_known,
-        );
-        let baseline_generation_mismatch =
-            state.baseline_known && existing_baseline.4 && state.baseline_generation != existing_baseline.2;
-        let use_persisted_baseline = state.baseline_known && !baseline_generation_mismatch;
-        progress.objects_scanned = state.processed_objects;
+        progress.objects_scanned = state.total_objects;
         progress.objects_healed = state.successful_objects;
         progress.objects_failed = state.failed_objects;
-        progress.skipped_objects = state.skipped_objects;
-        progress.skipped_new_versions = state.skipped_new_versions;
-        progress.skipped_ilm_expired = state.skipped_ilm_expired;
-        progress.bytes_processed = state.processed_bytes;
-        progress.counter_unknown = state.counter_unknown;
-        if use_persisted_baseline
-            || existing_baseline.0 > 0
-            || existing_baseline.1 > 0
-            || existing_baseline.2.is_some()
-            || existing_baseline.4
-        {
-            progress.objects_total_count = if use_persisted_baseline {
-                state.total_objects
-            } else {
-                existing_baseline.0
-            };
-            progress.objects_total_size = if use_persisted_baseline {
-                state.total_bytes
-            } else {
-                existing_baseline.1
-            };
-            progress.baseline_generation = if use_persisted_baseline {
-                state.baseline_generation
-            } else {
-                existing_baseline.2
-            };
-            progress.baseline_known = use_persisted_baseline
-                || existing_baseline.0 > 0
-                || existing_baseline.1 > 0
-                || existing_baseline.2.is_some()
-                || existing_baseline.4;
-        }
-        progress.progress_state = if use_persisted_baseline
-            || existing_baseline.0 > 0
-            || existing_baseline.1 > 0
-            || existing_baseline.2.is_some()
-            || existing_baseline.4
-        {
-            crate::heal::progress::HealProgressState::Running
-        } else {
-            crate::heal::progress::HealProgressState::Indeterminate
-        };
-        if baseline_generation_mismatch || state.counter_unknown {
-            progress.mark_unknown();
-        }
-        progress.ledger_complete = false;
-        progress.refresh_progress_percentage();
+        progress.bytes_processed = 0; // Resume state tracks object counts, not byte counters.
         progress.start_time = UNIX_EPOCH.checked_add(Duration::from_secs(state.start_time));
         progress.last_update_time = UNIX_EPOCH.checked_add(Duration::from_secs(state.last_update));
         progress.set_current_object(state.current_object.clone());
@@ -1509,8 +1264,8 @@ mod resume_loop_tests {
     };
     use crate::heal::progress::HealProgress;
     use crate::heal::resume::{
-        CheckpointManager, CheckpointObjectOutcome, CheckpointObjectOutcomeRecord, RESUME_CHECKPOINT_FILE,
-        ReplacementTargetIdentity, ResumeDeleteFailure, ResumeManager, ResumeUtils, compose_key,
+        CheckpointManager, RESUME_CHECKPOINT_FILE, ReplacementTargetIdentity, ResumeDeleteFailure, ResumeManager, ResumeUtils,
+        compose_key,
     };
     use crate::heal::storage::{HealLifecycleExpiryContext, HealListItem, HealObjectInfo, HealStorageAPI};
     use crate::heal::storage_api::status::BucketInfo;
@@ -1650,7 +1405,6 @@ mod resume_loop_tests {
         list_include_lifecycle_object_info: Mutex<Vec<bool>>,
         replacement_target_identity_sequences: Mutex<VecDeque<Vec<ReplacementTargetIdentity>>>,
         fail_listing: AtomicBool,
-        fail_listing_buckets: Mutex<HashSet<String>>,
     }
 
     impl FakeStorage {
@@ -1687,9 +1441,6 @@ mod resume_loop_tests {
         fn fail_listing(&self) {
             self.fail_listing.store(true, Ordering::SeqCst);
         }
-        fn fail_bucket_listing(&self, bucket: &str) {
-            self.fail_listing_buckets.lock().unwrap().insert(bucket.to_string());
-        }
     }
 
     #[async_trait::async_trait]
@@ -1780,7 +1531,7 @@ mod resume_loop_tests {
         }
         async fn list_objects_for_heal_page(
             &self,
-            bucket: &str,
+            _bucket: &str,
             _prefix: &str,
             continuation_token: Option<&str>,
             include_lifecycle_object_info: bool,
@@ -1789,7 +1540,7 @@ mod resume_loop_tests {
                 .lock()
                 .unwrap()
                 .push(include_lifecycle_object_info);
-            if self.fail_listing.load(Ordering::SeqCst) || self.fail_listing_buckets.lock().unwrap().contains(bucket) {
+            if self.fail_listing.load(Ordering::SeqCst) {
                 return Err(Error::other("injected listing failure"));
             }
             let key = continuation_token.map(str::to_string);
@@ -2167,49 +1918,6 @@ mod resume_loop_tests {
         assert!(state.completed_buckets.is_empty(), "the failed bucket must remain resumable");
     }
 
-    #[tokio::test]
-    async fn bucket_failure_stops_before_a_later_bucket_checkpoint() {
-        let env = make_env().await;
-        let task_id = ResumeUtils::generate_task_id();
-        let buckets = vec!["a".to_string(), "b".to_string()];
-        let resume = ResumeManager::new(
-            env.healer.disk.clone(),
-            task_id.clone(),
-            "erasure_set".to_string(),
-            "pool_0_set_0".to_string(),
-            buckets.clone(),
-        )
-        .await
-        .unwrap();
-        let checkpoint = CheckpointManager::new(env.healer.disk.clone(), task_id.clone())
-            .await
-            .unwrap();
-        env.storage.fail_bucket_listing("a");
-        for _ in 0..3 {
-            assert!(resume.schedule_retry().await.unwrap());
-        }
-
-        env.healer
-            .execute_heal_with_resume(&buckets, "pool_0_set_0", &resume, &checkpoint)
-            .await
-            .expect_err("the first bucket failure must keep the pass incomplete");
-        let persisted = checkpoint.get_checkpoint().await;
-        assert_eq!(persisted.current_bucket_index, 0);
-        assert!(resume.get_state().await.completed_buckets.is_empty());
-
-        let resumed = ResumeManager::load_from_disk(env.healer.disk.clone(), &task_id)
-            .await
-            .unwrap();
-        let checkpoint = CheckpointManager::load_from_disk(env.healer.disk.clone(), &task_id)
-            .await
-            .unwrap();
-        env.healer
-            .execute_heal_with_resume(&buckets, "pool_0_set_0", &resumed, &checkpoint)
-            .await
-            .expect_err("recovery must retry the earlier failed bucket");
-        assert!(!resumed.get_state().await.completed);
-    }
-
     #[tokio::test]
     async fn completed_resume_state_is_not_selected_for_a_new_heal() {
         let env = make_env().await;
@@ -2399,175 +2107,8 @@ mod resume_loop_tests {
         let mut names: Vec<String> = env.storage.calls().into_iter().map(|(n, _)| n).collect();
         names.sort();
         assert_eq!(names, vec!["a", "b", "c", "d"], "every object exactly once, none dropped/doubled");
-        // Keep the final page cursor until the outer loop durably completes the
-        // bucket, so a crash can replay only this page against its identities.
-        assert_eq!(env.resume.resume_cursor().await, Some("t1".to_string()));
-    }
-
-    #[tokio::test]
-    async fn persisted_failure_waits_for_the_bounded_retry_after_page_replay() {
-        let env = make_env().await;
-        env.storage.set_page(
-            None,
-            Page {
-                items: vec![item("object", Some("v1"), false)],
-                next: None,
-                truncated: false,
-            },
-        );
-        env.checkpoint
-            .record_object_outcome(CheckpointObjectOutcomeRecord {
-                object: compose_key("object", Some("v1")),
-                outcome: CheckpointObjectOutcome::Failed,
-                successful: 0,
-                failed: 1,
-                skipped: 0,
-                bytes: 0,
-                skipped_new_versions: 0,
-                skipped_ilm_expired: 0,
-                counter_unknown: false,
-            })
-            .await
-            .unwrap();
-        env.checkpoint.advance_page(0, 1).await.unwrap();
-
-        let resumed = ResumeManager::load_from_disk(env.healer.disk.clone(), &env.task_id)
-            .await
-            .unwrap();
-        let checkpoint = CheckpointManager::load_from_disk(env.healer.disk.clone(), &env.task_id)
-            .await
-            .unwrap();
-
-        env.healer
-            .execute_heal_with_resume(&["b".to_string()], "pool_0_set_0", &resumed, &checkpoint)
-            .await
-            .expect_err("the persisted failure must schedule a bounded retry");
-        assert!(
-            env.storage.calls().is_empty(),
-            "the failed identity must not be repeated in the same pass"
-        );
-
-        env.healer
-            .execute_heal_with_resume(&["b".to_string()], "pool_0_set_0", &resumed, &checkpoint)
-            .await
-            .expect("the bounded retry must heal the object");
-        assert_eq!(env.storage.calls(), vec![("object".to_string(), Some("v1".to_string()))]);
-        let state = resumed.get_state().await;
-        assert_eq!(state.successful_objects, 1);
-        assert_eq!(state.failed_objects, 0);
-    }
-
-    #[tokio::test]
-    async fn final_page_crash_replays_only_the_retained_page_identities() {
-        let env = make_env().await;
-        env.storage.set_page(
-            None,
-            Page {
-                items: vec![item("first", Some("v1"), false)],
-                next: Some("final-page".to_string()),
-                truncated: true,
-            },
-        );
-        env.storage.set_page(
-            Some("final-page"),
-            Page {
-                items: vec![item("last", Some("v1"), false)],
-                next: None,
-                truncated: false,
-            },
-        );
-
-        let (processed, successful, failed, skipped, result) = run(&env).await;
-        result.expect("the bucket pass must finish before the simulated crash");
-        assert_eq!((processed, successful, failed, skipped), (2, 2, 0, 0));
-
-        let resumed = ResumeManager::load_from_disk(env.healer.disk.clone(), &env.task_id)
-            .await
-            .unwrap();
-        let checkpoint = CheckpointManager::load_from_disk(env.healer.disk.clone(), &env.task_id)
-            .await
-            .unwrap();
-        env.healer
-            .execute_heal_with_resume(&["b".to_string()], "pool_0_set_0", &resumed, &checkpoint)
-            .await
-            .expect("the retained final-page ledger must make recovery exact");
-
-        assert_eq!(
-            env.storage.calls(),
-            vec![
-                ("first".to_string(), Some("v1".to_string())),
-                ("last".to_string(), Some("v1".to_string()))
-            ]
-        );
-        let state = resumed.get_state().await;
-        assert_eq!(state.successful_objects, 2);
-        assert_eq!(state.processed_objects, 2);
-    }
-
-    #[tokio::test]
-    async fn truncated_page_without_token_retains_its_replay_ledger() {
-        let env = make_env().await;
-        env.storage.set_page(
-            None,
-            Page {
-                items: vec![item("object", Some("v1"), false)],
-                next: None,
-                truncated: true,
-            },
-        );
-
-        let (processed, successful, failed, skipped, result) = run(&env).await;
-        result.expect("the tokenless truncated page is a terminal page");
-        assert_eq!((processed, successful, failed, skipped), (1, 1, 0, 0));
-
-        let resumed = ResumeManager::load_from_disk(env.healer.disk.clone(), &env.task_id)
-            .await
-            .unwrap();
-        let checkpoint = CheckpointManager::load_from_disk(env.healer.disk.clone(), &env.task_id)
-            .await
-            .unwrap();
-        env.healer
-            .execute_heal_with_resume(&["b".to_string()], "pool_0_set_0", &resumed, &checkpoint)
-            .await
-            .expect("terminal-page recovery must not replay a durable identity");
-
-        assert_eq!(env.storage.calls(), vec![("object".to_string(), Some("v1".to_string()))]);
-        let state = resumed.get_state().await;
-        assert_eq!(state.successful_objects, 1);
-        assert_eq!(state.processed_objects, 1);
-    }
-
-    #[tokio::test]
-    async fn completed_bucket_reconciles_its_final_page_checkpoint_after_crash() {
-        let env = make_env().await;
-        env.storage.set_page(
-            None,
-            Page {
-                items: vec![item("object", Some("v1"), false)],
-                next: None,
-                truncated: false,
-            },
-        );
-
-        let (_, _, _, _, result) = run(&env).await;
-        result.expect("the bucket pass must finish before the simulated crash");
-        env.resume.complete_bucket("b").await.unwrap();
-
-        let resumed = ResumeManager::load_from_disk(env.healer.disk.clone(), &env.task_id)
-            .await
-            .unwrap();
-        let checkpoint = CheckpointManager::load_from_disk(env.healer.disk.clone(), &env.task_id)
-            .await
-            .unwrap();
-        env.healer
-            .execute_heal_with_resume(&["b".to_string()], "pool_0_set_0", &resumed, &checkpoint)
-            .await
-            .expect("recovery must finish the checkpoint transition without replaying the bucket");
-
-        assert_eq!(env.storage.calls(), vec![("object".to_string(), Some("v1".to_string()))]);
-        let checkpoint = checkpoint.get_checkpoint().await;
-        assert_eq!(checkpoint.current_bucket_index, 1);
-        assert!(checkpoint.processed_objects.is_empty());
+        // Final page not truncated => cursor cleared.
+        assert_eq!(env.resume.resume_cursor().await, None);
     }
 
     #[tokio::test]

crates/heal/src/heal/manager.rs

@@ -2011,11 +2011,34 @@ impl HealManager {
             return None;
         }
 
-        let mut progresses = Vec::with_capacity(active_tasks.len());
+        let mut snapshot = HealProgress::default();
         for task in active_tasks {
-            progresses.push(task.get_progress().await);
+            let progress = task.get_progress().await;
+            snapshot.objects_scanned = snapshot.objects_scanned.saturating_add(progress.objects_scanned);
+            snapshot.objects_healed = snapshot.objects_healed.saturating_add(progress.objects_healed);
+            snapshot.objects_failed = snapshot.objects_failed.saturating_add(progress.objects_failed);
+            snapshot.skipped_new_versions = snapshot.skipped_new_versions.saturating_add(progress.skipped_new_versions);
+            snapshot.skipped_ilm_expired = snapshot.skipped_ilm_expired.saturating_add(progress.skipped_ilm_expired);
+            snapshot.objects_total_count = snapshot.objects_total_count.saturating_add(progress.objects_total_count);
+            snapshot.objects_total_size = snapshot.objects_total_size.saturating_add(progress.objects_total_size);
+            snapshot.bytes_processed = snapshot.bytes_processed.saturating_add(progress.bytes_processed);
+            snapshot.start_time = match (snapshot.start_time, progress.start_time) {
+                (Some(current), Some(next)) => Some(current.min(next)),
+                (None, next) => next,
+                (current, None) => current,
+            };
+            snapshot.last_update_time = match (snapshot.last_update_time, progress.last_update_time) {
+                (Some(current), Some(next)) => Some(current.max(next)),
+                (None, next) => next,
+                (current, None) => current,
+            };
+            if progress.current_object.is_some() {
+                snapshot.current_object = progress.current_object;
+            }
         }
-        crate::heal::progress::aggregate_heal_progress(progresses)
+        snapshot.refresh_progress_percentage();
+        snapshot.refresh_estimated_completion_time();
+        Some(snapshot)
     }
 }
 

crates/heal/src/heal/progress.rs

@@ -15,91 +15,15 @@
 use serde::{Deserialize, Serialize};
 use std::time::{Duration, SystemTime};
 
-pub(crate) fn stable_generation(parts: &[&[u8]]) -> u64 {
-    let mut hash = 0xcbf29ce484222325u64;
-    for part in parts {
-        for byte in (part.len() as u64).to_be_bytes().into_iter().chain(part.iter().copied()) {
-            hash ^= u64::from(byte);
-            hash = hash.wrapping_mul(0x100000001b3);
-        }
-    }
-    hash
-}
-
-#[cfg(test)]
-mod stable_generation_tests {
-    use super::stable_generation;
-
-    #[test]
-    fn stable_generation_has_a_fixed_vector() {
-        assert_eq!(stable_generation(&[b"rustfs", b"heal", b"42"]), 11_007_672_338_488_385_056);
-    }
-}
-
-pub(crate) fn increment_counter(counter: &mut u64) -> bool {
-    match counter.checked_add(1) {
-        Some(next) => {
-            *counter = next;
-            true
-        }
-        None => {
-            *counter = u64::MAX;
-            false
-        }
-    }
-}
-
-pub(crate) fn add_bytes(total: &mut u64, amount: u64) -> bool {
-    match total.checked_add(amount) {
-        Some(next) => {
-            *total = next;
-            true
-        }
-        None => {
-            *total = u64::MAX;
-            false
-        }
-    }
-}
-
-#[derive(Debug, Default, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[derive(Debug, Default, Clone, Serialize, Deserialize)]
 #[serde(rename_all = "camelCase")]
-pub enum HealProgressKind {
-    #[default]
-    Unknown,
-    Stage,
-    ObjectSweep,
-}
-
-/// Whether the object ledger can produce a meaningful percentage.
-///
-/// A zero-valued baseline is not a completed scan: it means that no complete
-/// usage snapshot was available.  Keep this state explicit so callers do not
-/// mistake the legacy `0.0` wire value for a measured zero-percent result.
-#[derive(Debug, Default, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
-#[serde(rename_all = "camelCase")]
-pub enum HealProgressState {
-    #[default]
-    Unknown,
-    Indeterminate,
-    Running,
-    Completed,
-}
-
-#[derive(Debug, Default, Clone, PartialEq, Serialize, Deserialize)]
-#[serde(default, rename_all = "camelCase")]
 pub struct HealProgress {
-    #[serde(default)]
-    pub kind: HealProgressKind,
     /// Objects scanned
     pub objects_scanned: u64,
     /// Objects healed
     pub objects_healed: u64,
     /// Objects failed
     pub objects_failed: u64,
-    /// Versions deferred for a later retry pass.
-    #[serde(default)]
-    pub skipped_objects: u64,
     /// Versions skipped because they were written after this heal started
     pub skipped_new_versions: u64,
     /// Versions skipped because lifecycle already selected them for expiry
@@ -120,38 +44,11 @@ pub struct HealProgress {
     pub last_update_time: Option<SystemTime>,
     /// Estimated completion time
     pub estimated_completion_time: Option<SystemTime>,
-    /// Current stage number. Stage updates are intentionally independent from
-    /// the object ledger below.
-    #[serde(default)]
-    pub stage_current: u64,
-    /// Number of stages in the current task.
-    #[serde(default)]
-    pub stage_total: u64,
-    /// Explicitly distinguishes a missing usage baseline from measured 0%.
-    #[serde(default)]
-    pub progress_state: HealProgressState,
-    /// True only after the task's durable completion ledger was committed.
-    #[serde(default)]
-    pub ledger_complete: bool,
-    /// Generation of the usage snapshot used for the baseline, if available.
-    #[serde(default)]
-    pub baseline_generation: Option<u64>,
-    /// Whether the baseline was explicitly observed.  This is separate from
-    /// the counters so a known empty scope (0 objects, 0 bytes) is not
-    /// confused with a legacy snapshot that omitted the baseline fields.
-    #[serde(default)]
-    pub baseline_known: bool,
-    /// Internal telemetry fence set when an aggregate counter overflows or
-    /// becomes inconsistent.  It prevents a later refresh from fabricating a
-    /// percentage from the poisoned values.
-    #[serde(default)]
-    pub counter_unknown: bool,
 }
 
 impl HealProgress {
     pub fn new() -> Self {
         Self {
-            kind: HealProgressKind::Unknown,
             start_time: Some(SystemTime::now()),
             last_update_time: Some(SystemTime::now()),
             ..Default::default()
@@ -159,87 +56,12 @@ impl HealProgress {
     }
 
     pub fn update_progress(&mut self, scanned: u64, healed: u64, failed: u64, bytes: u64) {
-        self.update_object_sweep_progress(scanned, healed, failed, bytes);
-    }
-
-    pub fn update_object_sweep_progress(&mut self, scanned: u64, healed: u64, failed: u64, bytes: u64) {
-        self.kind = HealProgressKind::ObjectSweep;
         self.objects_scanned = scanned;
         self.objects_healed = healed;
         self.objects_failed = failed;
         self.bytes_processed = bytes;
         self.last_update_time = Some(SystemTime::now());
 
-        let explicit_skipped = match self.skipped_new_versions.checked_add(self.skipped_ilm_expired) {
-            Some(value) => value,
-            None => {
-                self.mark_unknown();
-                0
-            }
-        };
-        let skipped = healed
-            .checked_add(failed)
-            .and_then(|value| value.checked_add(explicit_skipped))
-            .and_then(|value| scanned.checked_sub(value))
-            .unwrap_or(0);
-        self.update_object_progress(scanned, healed, failed, skipped, bytes);
-    }
-
-    /// Update task stage progress without modifying object counters.
-    pub fn update_stage(&mut self, current: u64, total: u64) {
-        let object_sweep_active = matches!(self.kind, HealProgressKind::ObjectSweep);
-        if !object_sweep_active {
-            self.kind = HealProgressKind::Stage;
-        }
-        self.ledger_complete = false;
-        self.stage_current = current.min(total);
-        self.stage_total = total;
-        if object_sweep_active {
-            self.last_update_time = Some(SystemTime::now());
-            self.refresh_progress_percentage();
-            return;
-        }
-        self.progress_state = if total == 0 {
-            HealProgressState::Indeterminate
-        } else {
-            HealProgressState::Running
-        };
-        self.progress_percentage = if total == 0 {
-            0.0
-        } else {
-            (current as f64 / total as f64 * 100.0).min(100.0)
-        };
-        self.last_update_time = Some(SystemTime::now());
-    }
-
-    /// Update the disjoint object ledger. `scanned` is the number of terminal
-    /// object outcomes and must equal healed + failed + deferred skipped plus
-    /// the two terminal skip classes. Overflow is a corrupt/unknown counter
-    /// state, not a reason to abort a completed heal.
-    pub fn update_object_progress(&mut self, scanned: u64, healed: u64, failed: u64, skipped: u64, bytes: u64) {
-        self.kind = HealProgressKind::ObjectSweep;
-        // `skipped` is the transient/deferred class.  The two explicit skip
-        // counters are terminal classifications too, so include them in the
-        // same ledger without making callers maintain a second aggregate.
-        let outcomes = healed
-            .checked_add(failed)
-            .and_then(|value| value.checked_add(skipped))
-            .and_then(|value| value.checked_add(self.skipped_new_versions))
-            .and_then(|value| value.checked_add(self.skipped_ilm_expired));
-        self.objects_scanned = scanned;
-        self.objects_healed = healed;
-        self.objects_failed = failed;
-        self.skipped_objects = skipped;
-        self.bytes_processed = bytes;
-        self.last_update_time = Some(SystemTime::now());
-        self.ledger_complete = false;
-        if outcomes != Some(scanned) {
-            // Telemetry corruption must not abort a heal.  Preserve the
-            // counters for diagnostics, but do not derive a percentage from a
-            // double-counted or overflowing ledger.
-            self.mark_unknown();
-            return;
-        }
         self.refresh_progress_percentage();
         self.refresh_estimated_completion_time();
     }
@@ -247,88 +69,50 @@ impl HealProgress {
     pub fn set_total_baseline(&mut self, objects_total_count: u64, objects_total_size: u64) {
         self.objects_total_count = objects_total_count;
         self.objects_total_size = objects_total_size;
-        self.baseline_known = true;
         self.last_update_time = Some(SystemTime::now());
         self.refresh_progress_percentage();
         self.refresh_estimated_completion_time();
     }
 
-    pub fn set_total_baseline_with_generation(&mut self, objects_total_count: u64, objects_total_size: u64, generation: u64) {
-        self.baseline_generation = Some(generation);
-        self.set_total_baseline(objects_total_count, objects_total_size);
-    }
-
     pub fn record_skipped_new_version(&mut self) {
-        let Some(next) = self.skipped_new_versions.checked_add(1) else {
-            self.mark_unknown();
-            return;
-        };
-        self.skipped_new_versions = next;
+        self.skipped_new_versions = self.skipped_new_versions.saturating_add(1);
         self.last_update_time = Some(SystemTime::now());
         self.refresh_progress_percentage();
         self.refresh_estimated_completion_time();
     }
 
     pub fn record_skipped_ilm_expired(&mut self) {
-        let Some(next) = self.skipped_ilm_expired.checked_add(1) else {
-            self.mark_unknown();
-            return;
-        };
-        self.skipped_ilm_expired = next;
+        self.skipped_ilm_expired = self.skipped_ilm_expired.saturating_add(1);
         self.last_update_time = Some(SystemTime::now());
         self.refresh_progress_percentage();
         self.refresh_estimated_completion_time();
     }
 
-    fn completed_for_baseline(&self) -> Option<u64> {
+    fn completed_for_baseline(&self) -> u64 {
         self.objects_healed
-            .checked_add(self.objects_failed)?
-            .checked_add(self.skipped_objects)?
-            .checked_add(self.skipped_new_versions)?
-            .checked_add(self.skipped_ilm_expired)
+            .saturating_add(self.objects_failed)
+            .saturating_add(self.skipped_new_versions)
+            .saturating_add(self.skipped_ilm_expired)
     }
 
     pub(crate) fn refresh_progress_percentage(&mut self) {
-        if self.ledger_complete {
-            self.progress_state = HealProgressState::Completed;
-            self.progress_percentage = 100.0;
-            return;
-        }
-        if self.counter_unknown {
-            self.progress_state = HealProgressState::Unknown;
-            self.progress_percentage = 0.0;
-            return;
-        }
-        if !self.baseline_known {
-            self.progress_state = HealProgressState::Indeterminate;
-            self.progress_percentage = 0.0;
-            self.estimated_completion_time = None;
-            return;
-        }
         if self.objects_total_size > 0 {
             self.progress_percentage = ((self.bytes_processed as f64 / self.objects_total_size as f64) * 100.0).min(100.0);
-            self.progress_percentage = self.progress_percentage.min(99.999);
-            self.progress_state = HealProgressState::Running;
             return;
         }
         if self.objects_total_count > 0 {
-            let Some(completed) = self.completed_for_baseline() else {
-                self.progress_state = HealProgressState::Unknown;
-                self.progress_percentage = 0.0;
-                return;
-            };
+            let completed = self.completed_for_baseline();
             self.progress_percentage = ((completed as f64 / self.objects_total_count as f64) * 100.0).min(100.0);
-            self.progress_percentage = self.progress_percentage.min(99.999);
-            self.progress_state = HealProgressState::Running;
             return;
         }
-        if self.baseline_known {
-            self.progress_state = HealProgressState::Running;
-            self.progress_percentage = 0.0;
-            return;
+
+        let total = self
+            .objects_scanned
+            .saturating_add(self.objects_healed)
+            .saturating_add(self.objects_failed);
+        if total > 0 {
+            self.progress_percentage = (self.objects_healed as f64 / total as f64) * 100.0;
         }
-        self.progress_state = HealProgressState::Indeterminate;
-        self.progress_percentage = 0.0;
     }
 
     pub fn set_current_object(&mut self, object: Option<String>) {
@@ -341,11 +125,7 @@ impl HealProgress {
             self.estimated_completion_time = None;
             return;
         };
-        if self.is_completed()
-            || self.progress_percentage <= 0.0
-            || self.progress_percentage >= 100.0
-            || self.bytes_processed == 0
-        {
+        if self.is_completed() || !(0.0..100.0).contains(&self.progress_percentage) || self.bytes_processed == 0 {
             self.estimated_completion_time = None;
             return;
         }
@@ -362,39 +142,18 @@ impl HealProgress {
     }
 
     pub fn is_completed(&self) -> bool {
-        self.ledger_complete
-    }
-
-    /// Mark telemetry unknown while allowing the underlying heal operation to
-    /// continue.  This is used for corrupt/overflowing counters at the
-    /// observability boundary; it must never turn a successful heal into an
-    /// execution error.
-    pub fn mark_unknown(&mut self) {
-        self.counter_unknown = true;
-        self.progress_state = HealProgressState::Unknown;
-        self.ledger_complete = false;
-        self.progress_percentage = 0.0;
-        self.estimated_completion_time = None;
-        self.last_update_time = Some(SystemTime::now());
-    }
-
-    /// Mark the object ledger terminal only after the enclosing task has
-    /// committed all durable resume state and cleanup fences.
-    pub fn mark_completed(&mut self) {
-        let telemetry_unknown = self.counter_unknown || self.progress_state == HealProgressState::Unknown;
-        self.ledger_complete = true;
-        if !telemetry_unknown {
-            self.progress_state = HealProgressState::Completed;
+        if self.progress_percentage >= 100.0 {
+            return true;
         }
-        self.progress_percentage = 100.0;
-        self.last_update_time = Some(SystemTime::now());
-        self.estimated_completion_time = None;
+        if self.objects_total_count > 0 || self.objects_total_size > 0 {
+            return false;
+        }
+
+        self.objects_scanned > 0 && self.objects_healed.saturating_add(self.objects_failed) >= self.objects_scanned
     }
 
     pub fn get_success_rate(&self) -> f64 {
-        let Some(total) = self.objects_healed.checked_add(self.objects_failed) else {
-            return 0.0;
-        };
+        let total = self.objects_healed + self.objects_failed;
         if total > 0 {
             (self.objects_healed as f64 / total as f64) * 100.0
         } else {
@@ -403,101 +162,6 @@ impl HealProgress {
     }
 }
 
-pub fn aggregate_heal_progress(progresses: impl IntoIterator<Item = HealProgress>) -> Option<HealProgress> {
-    let mut snapshot = HealProgress::default();
-    let mut found = false;
-    let mut has_object_sweep = false;
-    let mut all_object_baselines_known = true;
-    let mut baseline_generation = None;
-    let mut baseline_generation_consistent = true;
-    let mut all_ledgers_complete = true;
-    let mut counter_overflow = false;
-
-    for progress in progresses {
-        found = true;
-        let object_sweep = matches!(progress.kind, HealProgressKind::ObjectSweep);
-        has_object_sweep |= object_sweep;
-        all_ledgers_complete &= progress.ledger_complete;
-        if object_sweep {
-            all_object_baselines_known &= progress.baseline_known;
-            match baseline_generation {
-                None => baseline_generation = Some(progress.baseline_generation),
-                Some(generation) => baseline_generation_consistent &= generation == progress.baseline_generation,
-            }
-        }
-        counter_overflow |= progress.counter_unknown || matches!(progress.progress_state, HealProgressState::Unknown);
-        for (target, value) in [
-            (&mut snapshot.objects_scanned, progress.objects_scanned),
-            (&mut snapshot.objects_healed, progress.objects_healed),
-            (&mut snapshot.objects_failed, progress.objects_failed),
-            (&mut snapshot.skipped_objects, progress.skipped_objects),
-            (&mut snapshot.skipped_new_versions, progress.skipped_new_versions),
-            (&mut snapshot.skipped_ilm_expired, progress.skipped_ilm_expired),
-            (&mut snapshot.objects_total_count, progress.objects_total_count),
-            (&mut snapshot.objects_total_size, progress.objects_total_size),
-            (&mut snapshot.bytes_processed, progress.bytes_processed),
-            (&mut snapshot.stage_current, progress.stage_current),
-            (&mut snapshot.stage_total, progress.stage_total),
-        ] {
-            match target.checked_add(value) {
-                Some(sum) => *target = sum,
-                None => {
-                    *target = u64::MAX;
-                    counter_overflow = true;
-                }
-            }
-        }
-        snapshot.start_time = match (snapshot.start_time, progress.start_time) {
-            (Some(current), Some(next)) => Some(current.min(next)),
-            (None, next) => next,
-            (current, None) => current,
-        };
-        snapshot.last_update_time = match (snapshot.last_update_time, progress.last_update_time) {
-            (Some(current), Some(next)) => Some(current.max(next)),
-            (None, next) => next,
-            (current, None) => current,
-        };
-        if progress.current_object.is_some() {
-            snapshot.current_object = progress.current_object;
-        }
-    }
-
-    if !found {
-        return None;
-    }
-
-    snapshot.kind = if has_object_sweep {
-        HealProgressKind::ObjectSweep
-    } else {
-        HealProgressKind::Stage
-    };
-    snapshot.baseline_known = has_object_sweep && all_object_baselines_known && baseline_generation_consistent;
-    snapshot.baseline_generation = if snapshot.baseline_known && baseline_generation_consistent {
-        baseline_generation.flatten()
-    } else {
-        None
-    };
-    snapshot.ledger_complete = all_ledgers_complete;
-    snapshot.counter_unknown = counter_overflow;
-    if counter_overflow {
-        snapshot.progress_state = HealProgressState::Unknown;
-        snapshot.progress_percentage = if snapshot.ledger_complete { 100.0 } else { 0.0 };
-    } else if snapshot.ledger_complete {
-        snapshot.progress_state = HealProgressState::Completed;
-        snapshot.progress_percentage = 100.0;
-    } else if has_object_sweep {
-        snapshot.refresh_progress_percentage();
-    } else if snapshot.stage_total == 0 {
-        snapshot.progress_state = HealProgressState::Indeterminate;
-        snapshot.progress_percentage = 0.0;
-    } else {
-        snapshot.progress_state = HealProgressState::Running;
-        snapshot.progress_percentage = ((snapshot.stage_current as f64 / snapshot.stage_total as f64) * 100.0).min(99.999);
-    }
-    snapshot.refresh_estimated_completion_time();
-    Some(snapshot)
-}
-
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct HealStatistics {
     /// Total heal tasks
@@ -566,7 +230,6 @@ mod tests {
         assert_eq!(progress.objects_scanned, 0);
         assert_eq!(progress.objects_healed, 0);
         assert_eq!(progress.objects_failed, 0);
-        assert_eq!(progress.skipped_objects, 0);
         assert_eq!(progress.skipped_new_versions, 0);
         assert_eq!(progress.skipped_ilm_expired, 0);
         assert_eq!(progress.objects_total_count, 0);
@@ -587,8 +250,10 @@ mod tests {
         assert_eq!(progress.objects_healed, 8);
         assert_eq!(progress.objects_failed, 2);
         assert_eq!(progress.bytes_processed, 1024);
-        assert_eq!(progress.progress_state, HealProgressState::Indeterminate);
-        assert_eq!(progress.progress_percentage, 0.0);
+        // Progress percentage should be calculated based on healed/total
+        // total = scanned + healed + failed = 10 + 8 + 2 = 20
+        // healed/total = 8/20 = 0.4 = 40%
+        assert!((progress.progress_percentage - 40.0).abs() < 0.001);
         assert!(progress.last_update_time.is_some());
     }
 
@@ -597,8 +262,7 @@ mod tests {
         let mut progress = HealProgress::new();
         progress.start_time = Some(SystemTime::now() - Duration::from_secs(10));
 
-        progress.set_total_baseline(100, 16384);
-        progress.update_progress(25, 25, 0, 4096);
+        progress.update_progress(100, 25, 0, 4096);
 
         let eta = progress
             .estimated_completion_time
@@ -611,7 +275,7 @@ mod tests {
         let mut progress = HealProgress::new();
         progress.set_total_baseline(10, 8192);
 
-        progress.update_progress(25, 25, 0, 4096);
+        progress.update_progress(100, 25, 0, 4096);
 
         assert!((progress.progress_percentage - 50.0).abs() < 0.001);
     }
@@ -621,7 +285,7 @@ mod tests {
         let mut progress = HealProgress::new();
         progress.set_total_baseline(10, 0);
 
-        progress.update_progress(5, 3, 2, 0);
+        progress.update_progress(100, 3, 2, 0);
 
         assert!((progress.progress_percentage - 50.0).abs() < 0.001);
     }
@@ -631,7 +295,7 @@ mod tests {
         let mut progress = HealProgress::new();
         progress.set_total_baseline(10, 0);
 
-        progress.update_progress(5, 3, 2, 0);
+        progress.update_progress(100, 3, 2, 0);
         progress.record_skipped_new_version();
 
         assert_eq!(progress.skipped_new_versions, 1);
@@ -672,8 +336,7 @@ mod tests {
     fn test_heal_progress_update_progress_all_healed() {
         let mut progress = HealProgress::new();
         // When scanned=0, healed=10, failed=0: total=10, progress = 10/10 = 100%
-        progress.update_progress(10, 10, 0, 2048);
-        progress.mark_completed();
+        progress.update_progress(0, 10, 0, 2048);
 
         // All healed, should be 100%
         assert!((progress.progress_percentage - 100.0).abs() < 0.001);
@@ -731,7 +394,6 @@ mod tests {
         assert_eq!(json["objectsScanned"], 10);
         assert_eq!(json["objectsHealed"], 8);
         assert_eq!(json["objectsFailed"], 2);
-        assert_eq!(json["skippedObjects"], 0);
         assert_eq!(json["skippedNewVersions"], 0);
         assert_eq!(json["skippedIlmExpired"], 0);
         assert_eq!(json["bytesProcessed"], 1024);
@@ -743,7 +405,6 @@ mod tests {
     fn test_heal_progress_is_completed_by_percentage() {
         let mut progress = HealProgress::new();
         progress.update_progress(10, 10, 0, 1024);
-        progress.mark_completed();
 
         assert!(progress.is_completed());
     }
@@ -754,7 +415,7 @@ mod tests {
         progress.objects_scanned = 10;
         progress.objects_healed = 8;
         progress.objects_failed = 2;
-        progress.mark_completed();
+        // healed + failed = 8 + 2 = 10 >= scanned = 10
         assert!(progress.is_completed());
     }
 
@@ -794,108 +455,6 @@ mod tests {
         assert!((progress.get_success_rate() - 100.0).abs() < 0.001);
     }
 
-    #[test]
-    fn single_object_progress_reaches_terminal_100() {
-        let mut progress = HealProgress::new();
-        progress.update_object_progress(1, 1, 0, 0, 128);
-        assert!(!progress.is_completed());
-        progress.mark_completed();
-        assert!(progress.is_completed());
-        assert_eq!(progress.progress_percentage, 100.0);
-    }
-
-    #[test]
-    fn progress_without_baseline_is_indeterminate() {
-        let mut progress = HealProgress::new();
-        progress.update_object_progress(1, 1, 0, 0, 128);
-        assert_eq!(progress.progress_state, HealProgressState::Indeterminate);
-        assert_eq!(progress.progress_percentage, 0.0);
-        assert!(progress.estimated_completion_time.is_none());
-    }
-
-    #[test]
-    fn progress_retry_is_exactly_once() {
-        let mut progress = HealProgress::new();
-        progress.set_total_baseline(1, 128);
-        progress.update_object_progress(1, 1, 0, 0, 128);
-        progress.update_object_progress(1, 1, 0, 0, 128);
-        assert_eq!(progress.objects_scanned, 1);
-        assert_eq!(progress.objects_healed, 1);
-        assert_eq!(progress.bytes_processed, 128);
-    }
-
-    #[test]
-    fn progress_never_triggers_cleanup_before_terminal_ledger_empty() {
-        let mut progress = HealProgress::new();
-        progress.progress_percentage = 100.0;
-        assert!(!progress.is_completed());
-        progress.mark_completed();
-        assert!(progress.is_completed());
-    }
-
-    #[test]
-    fn progress_counter_overflow_is_marked_unknown_without_aborting_completed_heal() {
-        let mut progress = HealProgress::new();
-        progress.update_object_progress(u64::MAX, u64::MAX, 1, 0, 0);
-        assert_eq!(progress.progress_state, HealProgressState::Unknown);
-        progress.mark_completed();
-        assert!(progress.is_completed());
-        assert_eq!(progress.progress_state, HealProgressState::Unknown);
-
-        let aggregate = aggregate_heal_progress([progress]).expect("progress should aggregate");
-        assert!(aggregate.ledger_complete);
-        assert!(aggregate.counter_unknown);
-        assert_eq!(aggregate.progress_state, HealProgressState::Unknown);
-        assert_eq!(aggregate.progress_percentage, 100.0);
-    }
-
-    #[test]
-    fn aggregate_rejects_mixed_baseline_generations() {
-        let progress = |generation| HealProgress {
-            kind: HealProgressKind::ObjectSweep,
-            objects_scanned: 5,
-            objects_total_count: 10,
-            progress_state: HealProgressState::Running,
-            baseline_generation: Some(generation),
-            baseline_known: true,
-            ..Default::default()
-        };
-
-        let aggregate = aggregate_heal_progress([progress(1), progress(2)]).expect("progress should aggregate");
-        assert!(!aggregate.baseline_known);
-        assert_eq!(aggregate.baseline_generation, None);
-        assert_eq!(aggregate.progress_state, HealProgressState::Indeterminate);
-        assert_eq!(aggregate.progress_percentage, 0.0);
-    }
-
-    #[test]
-    fn aggregate_accepts_multiple_sets_from_one_snapshot_generation() {
-        let progress = |objects_scanned| HealProgress {
-            kind: HealProgressKind::ObjectSweep,
-            objects_scanned,
-            objects_total_count: 10,
-            progress_state: HealProgressState::Running,
-            baseline_generation: Some(7),
-            baseline_known: true,
-            ..Default::default()
-        };
-
-        let aggregate = aggregate_heal_progress([progress(5), progress(3)]).expect("progress should aggregate");
-        assert!(aggregate.baseline_known);
-        assert_eq!(aggregate.baseline_generation, Some(7));
-    }
-
-    #[test]
-    fn stage_updates_do_not_double_count_object_outcomes() {
-        let mut progress = HealProgress::new();
-        progress.update_object_progress(2, 1, 0, 1, 256);
-        progress.update_stage(3, 4);
-        assert_eq!(progress.kind, HealProgressKind::ObjectSweep);
-        assert_eq!(progress.objects_scanned, 2);
-        assert_eq!(progress.objects_healed, 1);
-        assert_eq!(progress.skipped_objects, 1);
-    }
-
     #[test]
     fn test_heal_statistics_new() {
         let stats = HealStatistics::new();

crates/heal/src/heal/resume.rs

@@ -31,7 +31,7 @@ mod checkpoint;
 mod replacement;
 mod utils;
 
-pub use checkpoint::{CheckpointManager, CheckpointObjectOutcome, CheckpointObjectOutcomeRecord, ResumeCheckpoint};
+pub use checkpoint::{CheckpointManager, ResumeCheckpoint};
 pub(crate) use replacement::replacement_target_identities_match;
 use replacement::replacement_targets_match_identities;
 pub use replacement::{
@@ -340,12 +340,6 @@ pub struct ResumeState {
     pub failed_objects: u64,
     /// skipped objects
     pub skipped_objects: u64,
-    /// Terminal versions skipped because they were newer than the heal start.
-    #[serde(default)]
-    pub skipped_new_versions: u64,
-    /// Terminal versions handed to lifecycle expiry.
-    #[serde(default)]
-    pub skipped_ilm_expired: u64,
     /// current bucket
     pub current_bucket: Option<String>,
     /// current object
@@ -360,24 +354,6 @@ pub struct ResumeState {
     pub retry_count: u32,
     /// max retries
     pub max_retries: u32,
-    /// Bytes accounted by the object ledger; additive for old snapshots.
-    #[serde(default)]
-    pub processed_bytes: u64,
-    /// Total bytes from a complete usage snapshot, when available.
-    #[serde(default)]
-    pub total_bytes: u64,
-    /// Generation of the usage snapshot used for the baseline.
-    #[serde(default)]
-    pub baseline_generation: Option<u64>,
-    /// Whether the usage baseline is known. Missing in old snapshots means
-    /// indeterminate rather than a measured zero baseline.
-    #[serde(default)]
-    pub baseline_known: bool,
-    /// Persistent telemetry fence for counter/byte overflow or corruption.
-    /// It must survive a restart so a saturated snapshot is never presented as
-    /// a measured percentage on the next resume.
-    #[serde(default)]
-    pub counter_unknown: bool,
 }
 
 impl ResumeState {
@@ -401,8 +377,6 @@ impl ResumeState {
             successful_objects: 0,
             failed_objects: 0,
             skipped_objects: 0,
-            skipped_new_versions: 0,
-            skipped_ilm_expired: 0,
             current_bucket: None,
             current_object: None,
             completed_buckets: Vec::new(),
@@ -410,11 +384,6 @@ impl ResumeState {
             error_message: None,
             retry_count: 0,
             max_retries: 3,
-            processed_bytes: 0,
-            total_bytes: 0,
-            baseline_generation: None,
-            baseline_known: false,
-            counter_unknown: false,
         }
     }
 
@@ -443,39 +412,6 @@ impl ResumeState {
         self.last_update = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
     }
 
-    pub fn update_progress_with_bytes(
-        &mut self,
-        processed: u64,
-        successful: u64,
-        failed: u64,
-        skipped: u64,
-        processed_bytes: u64,
-    ) {
-        self.update_progress(processed, successful, failed, skipped);
-        self.processed_bytes = processed_bytes;
-    }
-
-    pub fn set_skipped_version_counts(&mut self, new_versions: u64, ilm_expired: u64) {
-        self.skipped_new_versions = new_versions;
-        self.skipped_ilm_expired = ilm_expired;
-        self.last_update = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
-    }
-
-    pub fn set_progress_baseline(&mut self, total_objects: u64, total_bytes: u64, generation: Option<u64>) {
-        self.total_objects = total_objects;
-        self.total_bytes = total_bytes;
-        self.baseline_generation = generation;
-        // This method is called only after a complete usage snapshot has been
-        // validated.  A complete but empty snapshot is still a known baseline.
-        self.baseline_known = true;
-        self.last_update = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
-    }
-
-    pub fn mark_counter_unknown(&mut self) {
-        self.counter_unknown = true;
-        self.last_update = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
-    }
-
     pub fn set_current_item(&mut self, bucket: Option<String>, object: Option<String>) {
         self.current_bucket = bucket;
         self.current_object = object;
@@ -501,7 +437,6 @@ impl ResumeState {
         if let Some(pos) = self.pending_buckets.iter().position(|b| b == bucket) {
             self.pending_buckets.remove(pos);
         }
-        self.resume_cursor = None;
         self.last_update = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
     }
 
@@ -519,10 +454,6 @@ impl ResumeState {
         self.successful_objects = 0;
         self.failed_objects = 0;
         self.skipped_objects = 0;
-        self.skipped_new_versions = 0;
-        self.skipped_ilm_expired = 0;
-        self.processed_bytes = 0;
-        self.counter_unknown = false;
         self.completed = false;
         // A retry re-scans every bucket from the beginning, so the version
         // cursor must be cleared too — otherwise the retry would resume mid-scan.
@@ -545,28 +476,14 @@ impl ResumeState {
     }
 
     pub fn get_progress_percentage(&self) -> f64 {
-        if self.completed {
-            return 100.0;
-        }
-        if self.counter_unknown {
-            return 0.0;
-        }
-        if !self.baseline_known {
-            return 0.0;
-        }
-        if self.total_bytes > 0 {
-            return ((self.processed_bytes as f64 / self.total_bytes as f64) * 100.0).min(99.999);
-        }
         if self.total_objects == 0 {
             return 0.0;
         }
-        ((self.processed_objects as f64 / self.total_objects as f64) * 100.0).min(99.999)
+        (self.processed_objects as f64 / self.total_objects as f64) * 100.0
     }
 
     pub fn get_success_rate(&self) -> f64 {
-        let Some(total) = self.successful_objects.checked_add(self.failed_objects) else {
-            return 0.0;
-        };
+        let total = self.successful_objects + self.failed_objects;
         if total == 0 {
             return 0.0;
         }
@@ -837,14 +754,6 @@ impl ResumeManager {
             state.successful_objects = 0;
             state.failed_objects = 0;
             state.skipped_objects = 0;
-            state.skipped_new_versions = 0;
-            state.skipped_ilm_expired = 0;
-            state.processed_bytes = 0;
-            state.total_objects = 0;
-            state.total_bytes = 0;
-            state.baseline_generation = None;
-            state.baseline_known = false;
-            state.counter_unknown = false;
             state.completed = false;
             state.completed_buckets.clear();
             state.schema_version = CURRENT_RESUME_SCHEMA;
@@ -929,41 +838,6 @@ impl ResumeManager {
         self.save_state_throttled().await
     }
 
-    pub async fn update_progress_with_bytes(
-        &self,
-        processed: u64,
-        successful: u64,
-        failed: u64,
-        skipped: u64,
-        processed_bytes: u64,
-    ) -> Result<()> {
-        let mut state = self.state.write().await;
-        state.update_progress_with_bytes(processed, successful, failed, skipped, processed_bytes);
-        drop(state);
-        self.save_state_throttled().await
-    }
-
-    pub async fn set_progress_baseline(&self, total_objects: u64, total_bytes: u64, generation: Option<u64>) -> Result<()> {
-        let mut state = self.state.write().await;
-        state.set_progress_baseline(total_objects, total_bytes, generation);
-        drop(state);
-        self.save_state_throttled().await
-    }
-
-    pub async fn mark_counter_unknown(&self) -> Result<()> {
-        let mut state = self.state.write().await;
-        state.mark_counter_unknown();
-        drop(state);
-        self.save_state().await
-    }
-
-    pub async fn set_skipped_version_counts(&self, new_versions: u64, ilm_expired: u64) -> Result<()> {
-        let mut state = self.state.write().await;
-        state.set_skipped_version_counts(new_versions, ilm_expired);
-        drop(state);
-        self.save_state_throttled().await
-    }
-
     /// Set current item. Called once per healed object, so persistence is
     /// throttled: the in-memory state always updates, but the snapshot is only
     /// written every `PERSIST_EVERY_MUTATIONS` calls or `PERSIST_INTERVAL`.
@@ -1008,7 +882,7 @@ impl ResumeManager {
         let mut state = self.state.write().await;
         state.complete_bucket(bucket);
         drop(state);
-        self.save_state().await
+        self.save_state_throttled().await
     }
 
     /// mark task completed

crates/heal/src/heal/resume/checkpoint.rs

@@ -29,30 +29,10 @@ use super::{
 
 const EVENT_HEAL_CHECKPOINT_STATE: &str = "heal_checkpoint_state";
 
-/// Current on-disk schema version for `ResumeCheckpoint`. Schema 5 could
-/// persist dedup identities without the aggregate counters needed to restore
-/// them safely, so stale checkpoints are discarded and replayed.
-pub(super) const CURRENT_CHECKPOINT_SCHEMA: u32 = 6;
-
-#[derive(Debug, Clone, Copy)]
-pub enum CheckpointObjectOutcome {
-    Processed,
-    Failed,
-    Skipped,
-}
-
-#[derive(Debug)]
-pub struct CheckpointObjectOutcomeRecord {
-    pub object: String,
-    pub outcome: CheckpointObjectOutcome,
-    pub successful: u64,
-    pub failed: u64,
-    pub skipped: u64,
-    pub bytes: u64,
-    pub skipped_new_versions: u64,
-    pub skipped_ilm_expired: u64,
-    pub counter_unknown: bool,
-}
+/// Current on-disk schema version for `ResumeCheckpoint`. Same rationale as
+/// `CURRENT_RESUME_SCHEMA`: pre-per-version dedup identities are not comparable
+/// to the new `compose_key` identities, so a stale checkpoint is discarded.
+pub(super) const CURRENT_CHECKPOINT_SCHEMA: u32 = 5;
 
 /// resume checkpoint
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -77,30 +57,6 @@ pub struct ResumeCheckpoint {
     pub failed_objects: HashSet<String>,
     /// skipped objects
     pub skipped_objects: HashSet<String>,
-    /// Aggregate object ledger counters restored alongside the dedup sets.
-    #[serde(default)]
-    pub successful_objects: u64,
-    #[serde(default)]
-    pub failed_object_count: u64,
-    #[serde(default)]
-    pub skipped_object_count: u64,
-    #[serde(default)]
-    pub skipped_new_versions: u64,
-    #[serde(default)]
-    pub skipped_ilm_expired: u64,
-    #[serde(default)]
-    pub processed_bytes: u64,
-    #[serde(default)]
-    pub total_objects: u64,
-    #[serde(default)]
-    pub total_bytes: u64,
-    #[serde(default)]
-    pub baseline_generation: Option<u64>,
-    #[serde(default)]
-    pub baseline_known: bool,
-    /// Persistent telemetry fence for counter/byte overflow or corruption.
-    #[serde(default)]
-    pub counter_unknown: bool,
 }
 
 impl ResumeCheckpoint {
@@ -114,17 +70,6 @@ impl ResumeCheckpoint {
             processed_objects: HashSet::new(),
             failed_objects: HashSet::new(),
             skipped_objects: HashSet::new(),
-            successful_objects: 0,
-            failed_object_count: 0,
-            skipped_object_count: 0,
-            skipped_new_versions: 0,
-            skipped_ilm_expired: 0,
-            processed_bytes: 0,
-            total_objects: 0,
-            total_bytes: 0,
-            baseline_generation: None,
-            baseline_known: false,
-            counter_unknown: false,
         }
     }
 
@@ -146,34 +91,6 @@ impl ResumeCheckpoint {
         self.skipped_objects.insert(object);
     }
 
-    pub fn update_progress(&mut self, successful: u64, failed: u64, skipped: u64, bytes: u64) {
-        self.successful_objects = successful;
-        self.failed_object_count = failed;
-        self.skipped_object_count = skipped;
-        self.processed_bytes = bytes;
-    }
-
-    pub fn set_progress_baseline(&mut self, total_objects: u64, total_bytes: u64, generation: Option<u64>) {
-        self.total_objects = total_objects;
-        self.total_bytes = total_bytes;
-        self.baseline_generation = generation;
-        // The caller has already validated that this is a complete snapshot;
-        // preserve the distinction between a known empty scope and an old
-        // checkpoint that omitted all baseline fields.
-        self.baseline_known = true;
-    }
-
-    pub fn mark_counter_unknown(&mut self) {
-        self.counter_unknown = true;
-        self.checkpoint_time = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
-    }
-
-    pub fn set_skipped_version_counts(&mut self, new_versions: u64, ilm_expired: u64) {
-        self.skipped_new_versions = new_versions;
-        self.skipped_ilm_expired = ilm_expired;
-        self.checkpoint_time = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
-    }
-
     /// Advance past a fully-processed page: objects below `object_index` are
     /// skipped by position on resume, so the per-object sets no longer need
     /// their entries and would otherwise grow with the whole bucket.
@@ -190,17 +107,6 @@ impl ResumeCheckpoint {
         self.update_position(0, 0);
         self.processed_objects.clear();
         self.skipped_objects.clear();
-        self.successful_objects = 0;
-        self.failed_object_count = 0;
-        self.skipped_object_count = 0;
-        self.skipped_new_versions = 0;
-        self.skipped_ilm_expired = 0;
-        self.processed_bytes = 0;
-        self.total_objects = 0;
-        self.total_bytes = 0;
-        self.baseline_generation = None;
-        self.baseline_known = false;
-        self.counter_unknown = false;
         self.failed_objects.clear();
     }
 }
@@ -252,9 +158,10 @@ impl CheckpointManager {
             });
         }
 
-        // Older checkpoints can contain identities that are not comparable to
-        // the current keys or lack their corresponding aggregate counters.
-        // Discard the stale sets and position so the scan restarts cleanly.
+        // A checkpoint from an older schema stored latest-only dedup identities
+        // that are not comparable to the new per-version `compose_key`
+        // identities. Discard the stale sets and position, then stamp the
+        // current schema so the scan restarts cleanly.
         if checkpoint.schema_version > CURRENT_CHECKPOINT_SCHEMA {
             return Err(Error::TaskExecutionFailed {
                 message: format!(
@@ -278,17 +185,6 @@ impl CheckpointManager {
             checkpoint.processed_objects.clear();
             checkpoint.failed_objects.clear();
             checkpoint.skipped_objects.clear();
-            checkpoint.successful_objects = 0;
-            checkpoint.failed_object_count = 0;
-            checkpoint.skipped_object_count = 0;
-            checkpoint.skipped_new_versions = 0;
-            checkpoint.skipped_ilm_expired = 0;
-            checkpoint.processed_bytes = 0;
-            checkpoint.total_objects = 0;
-            checkpoint.total_bytes = 0;
-            checkpoint.baseline_generation = None;
-            checkpoint.baseline_known = false;
-            checkpoint.counter_unknown = false;
             checkpoint.current_bucket_index = 0;
             checkpoint.current_object_index = 0;
             checkpoint.schema_version = CURRENT_CHECKPOINT_SCHEMA;
@@ -329,7 +225,7 @@ impl CheckpointManager {
         self.save_checkpoint_throttled().await
     }
 
-    /// Persist a completed page position while retaining its identities.
+    /// Advance past a completed page and prune the per-object sets, then persist.
     pub async fn complete_page(&self, bucket_index: usize, object_index: usize) -> Result<()> {
         let mut checkpoint = self.checkpoint.write().await;
         checkpoint.complete_page(bucket_index, object_index);
@@ -337,35 +233,6 @@ impl CheckpointManager {
         self.save_checkpoint_throttled().await
     }
 
-    /// Persist the page position while retaining identities until the resume
-    /// cursor is durable.
-    pub async fn advance_page(&self, bucket_index: usize, object_index: usize) -> Result<()> {
-        let mut checkpoint = self.checkpoint.write().await;
-        checkpoint.update_position(bucket_index, object_index);
-        drop(checkpoint);
-        self.save_checkpoint().await
-    }
-
-    /// Remove the previous page's dedup identities only after its resume cursor
-    /// has been durably exposed.
-    pub async fn prune_completed_page(&self) -> Result<()> {
-        let mut checkpoint = self.checkpoint.write().await;
-        checkpoint.processed_objects.clear();
-        checkpoint.skipped_objects.clear();
-        checkpoint.failed_objects.clear();
-        drop(checkpoint);
-        self.save_checkpoint().await
-    }
-
-    /// Advance to the next bucket and clear the final page identities after the
-    /// resume state has durably recorded the completed bucket.
-    pub async fn complete_bucket(&self, next_bucket_index: usize) -> Result<()> {
-        let mut checkpoint = self.checkpoint.write().await;
-        checkpoint.complete_page(next_bucket_index, 0);
-        drop(checkpoint);
-        self.save_checkpoint().await
-    }
-
     /// Reset the checkpoint to the start of the scan for a retry, then persist.
     pub async fn reset_for_retry(&self) -> Result<()> {
         let mut checkpoint = self.checkpoint.write().await;
@@ -400,62 +267,6 @@ impl CheckpointManager {
         self.save_checkpoint_if_due().await
     }
 
-    /// Atomically persist an object's dedup identity with its aggregate result.
-    pub async fn record_object_outcome(&self, record: CheckpointObjectOutcomeRecord) -> Result<()> {
-        let CheckpointObjectOutcomeRecord {
-            object,
-            outcome,
-            successful,
-            failed,
-            skipped,
-            bytes,
-            skipped_new_versions,
-            skipped_ilm_expired,
-            counter_unknown,
-        } = record;
-        let mut checkpoint = self.checkpoint.write().await;
-        match outcome {
-            CheckpointObjectOutcome::Processed => checkpoint.add_processed_object(object),
-            CheckpointObjectOutcome::Failed => checkpoint.add_failed_object(object),
-            CheckpointObjectOutcome::Skipped => checkpoint.add_skipped_object(object),
-        }
-        checkpoint.update_progress(successful, failed, skipped, bytes);
-        checkpoint.set_skipped_version_counts(skipped_new_versions, skipped_ilm_expired);
-        if counter_unknown {
-            checkpoint.mark_counter_unknown();
-        }
-        drop(checkpoint);
-        self.save_checkpoint_if_due().await
-    }
-
-    pub async fn update_progress(&self, successful: u64, failed: u64, skipped: u64, bytes: u64) -> Result<()> {
-        let mut checkpoint = self.checkpoint.write().await;
-        checkpoint.update_progress(successful, failed, skipped, bytes);
-        drop(checkpoint);
-        self.save_checkpoint_if_due().await
-    }
-
-    pub async fn set_progress_baseline(&self, total_objects: u64, total_bytes: u64, generation: Option<u64>) -> Result<()> {
-        let mut checkpoint = self.checkpoint.write().await;
-        checkpoint.set_progress_baseline(total_objects, total_bytes, generation);
-        drop(checkpoint);
-        self.save_checkpoint_throttled().await
-    }
-
-    pub async fn mark_counter_unknown(&self) -> Result<()> {
-        let mut checkpoint = self.checkpoint.write().await;
-        checkpoint.mark_counter_unknown();
-        drop(checkpoint);
-        self.save_checkpoint().await
-    }
-
-    pub async fn set_skipped_version_counts(&self, new_versions: u64, ilm_expired: u64) -> Result<()> {
-        let mut checkpoint = self.checkpoint.write().await;
-        checkpoint.set_skipped_version_counts(new_versions, ilm_expired);
-        drop(checkpoint);
-        self.save_checkpoint_throttled().await
-    }
-
     async fn save_checkpoint_if_due(&self) -> Result<()> {
         let should_save = self.throttle.lock().map(|mut throttle| throttle.record()).unwrap_or(true);
         if !should_save {

crates/heal/src/heal/resume/tests.rs

@@ -1296,7 +1296,6 @@ async fn test_resume_state_progress() {
     assert_eq!(progress, 0.0); // total_objects is 0
 
     state.total_objects = 100;
-    state.baseline_known = true;
     let progress = state.get_progress_percentage();
     assert_eq!(progress, 10.0);
 }
@@ -1476,40 +1475,6 @@ fn test_checkpoint_object_sets_dedupe_and_prune() {
     assert!(checkpoint.failed_objects.is_empty());
 }
 
-#[tokio::test]
-async fn checkpoint_page_commit_keeps_ledger_until_cursor_is_durable() {
-    let (_temp_dir, disk) = schema_test_disk().await;
-    let task_id = ResumeUtils::generate_task_id();
-    let checkpoint = CheckpointManager::new(disk.clone(), task_id.clone()).await.unwrap();
-
-    checkpoint
-        .record_object_outcome(CheckpointObjectOutcomeRecord {
-            object: "bucket/object:v1".to_string(),
-            outcome: CheckpointObjectOutcome::Processed,
-            successful: 1,
-            failed: 0,
-            skipped: 0,
-            bytes: 128,
-            skipped_new_versions: 0,
-            skipped_ilm_expired: 0,
-            counter_unknown: false,
-        })
-        .await
-        .unwrap();
-    checkpoint.advance_page(0, 1).await.unwrap();
-
-    let reloaded = CheckpointManager::load_from_disk(disk.clone(), &task_id).await.unwrap();
-    let snapshot = reloaded.get_checkpoint().await;
-    assert_eq!(snapshot.current_object_index, 1);
-    assert_eq!(snapshot.successful_objects, 1);
-    assert_eq!(snapshot.processed_bytes, 128);
-    assert!(snapshot.processed_objects.contains("bucket/object:v1"));
-
-    checkpoint.prune_completed_page().await.unwrap();
-    let reloaded = CheckpointManager::load_from_disk(disk, &task_id).await.unwrap();
-    assert!(reloaded.get_checkpoint().await.processed_objects.is_empty());
-}
-
 #[test]
 fn test_checkpoint_loads_legacy_vec_format() {
     // Checkpoints written before the HashSet migration stored the object
@@ -1603,14 +1568,14 @@ async fn test_resumestate_schema_v0_discarded_on_load() {
 }
 
 #[tokio::test]
-async fn test_checkpoint_schema_v5_discarded_on_load() {
+async fn test_checkpoint_schema_v4_discarded_on_load() {
     let (temp_dir, disk) = schema_test_disk().await;
 
-    // Schema v5 can persist failed identities without the aggregate counters
-    // that make those identities safe to deduplicate after an upgrade.
+    // The previous checkpoint schema is unsafe once its paired resume
+    // state is discarded: retaining either position would skip work.
     let task_id = "00000000-0000-4000-8000-000000000002";
     let legacy = r#"{
-            "schema_version": 5,
+            "schema_version": 4,
             "task_id": "00000000-0000-4000-8000-000000000002",
             "checkpoint_time": 1700000000,
             "current_bucket_index": 2,
@@ -1674,120 +1639,6 @@ async fn current_normal_resume_schema_preserves_progress() {
     temp_dir.close().expect("remove schema test directory");
 }
 
-#[test]
-fn progress_checkpoint_restores_bytes_and_generation() {
-    let mut checkpoint = ResumeCheckpoint::new("progress-checkpoint".to_string());
-    checkpoint.set_progress_baseline(9, 4096, Some(77));
-    checkpoint.update_progress(4, 1, 2, 2048);
-    checkpoint.set_skipped_version_counts(3, 1);
-    checkpoint.mark_counter_unknown();
-
-    let restored: ResumeCheckpoint =
-        serde_json::from_slice(&serde_json::to_vec(&checkpoint).expect("serialize checkpoint")).expect("deserialize checkpoint");
-    assert_eq!(restored.processed_bytes, 2048);
-    assert_eq!(restored.total_objects, 9);
-    assert_eq!(restored.total_bytes, 4096);
-    assert_eq!(restored.baseline_generation, Some(77));
-    assert!(restored.baseline_known);
-    assert_eq!(restored.skipped_new_versions, 3);
-    assert_eq!(restored.skipped_ilm_expired, 1);
-    assert!(restored.counter_unknown);
-}
-
-#[test]
-fn old_progress_schema_migrates_missing_fields_to_unknown() {
-    let state = ResumeState::new(
-        "legacy-progress".to_string(),
-        "erasure_set".to_string(),
-        "pool_0_set_0".to_string(),
-        Vec::new(),
-    );
-    let mut value = serde_json::to_value(state).expect("serialize legacy-compatible state");
-    let object = value.as_object_mut().expect("state must be an object");
-    for field in [
-        "processed_bytes",
-        "total_bytes",
-        "baseline_generation",
-        "baseline_known",
-        "skipped_new_versions",
-        "skipped_ilm_expired",
-    ] {
-        object.remove(field);
-    }
-    object.insert("total_objects".to_string(), serde_json::json!(10));
-    object.insert("processed_objects".to_string(), serde_json::json!(5));
-    let restored: ResumeState = serde_json::from_value(value).expect("deserialize old progress state");
-    assert_eq!(restored.processed_bytes, 0);
-    assert_eq!(restored.total_bytes, 0);
-    assert_eq!(restored.baseline_generation, None);
-    assert!(!restored.baseline_known, "missing baseline must remain unknown");
-    assert_eq!(restored.get_progress_percentage(), 0.0);
-    assert_eq!(restored.skipped_new_versions, 0);
-    assert_eq!(restored.skipped_ilm_expired, 0);
-}
-
-#[test]
-fn progress_counter_unknown_survives_resume_round_trip() {
-    let mut state = ResumeState::new(
-        "overflow-progress".to_string(),
-        "erasure_set".to_string(),
-        "pool_0_set_0".to_string(),
-        Vec::new(),
-    );
-    state.mark_counter_unknown();
-
-    let restored: ResumeState =
-        serde_json::from_slice(&serde_json::to_vec(&state).expect("serialize resume state")).expect("deserialize resume state");
-    assert!(restored.counter_unknown);
-}
-
-#[tokio::test]
-async fn checkpoint_progress_survives_a_torn_resume_summary_write() {
-    let (_temp_dir, disk) = schema_test_disk().await;
-    let task_id = ResumeUtils::generate_task_id();
-    let _resume = ResumeManager::new(
-        disk.clone(),
-        task_id.clone(),
-        "erasure_set".to_string(),
-        "pool_0_set_0".to_string(),
-        vec!["bucket".to_string()],
-    )
-    .await
-    .expect("resume state should persist");
-    let checkpoint = CheckpointManager::new(disk.clone(), task_id.clone())
-        .await
-        .expect("checkpoint should persist");
-
-    // This is the ordering used by the erasure-set loop: the checkpoint is
-    // durable before the summary write.  Stop here to model a crash in the
-    // inter-store window and verify that the recovery authority retains the
-    // telemetry fence and bytes.
-    checkpoint
-        .update_progress(3, 0, 0, 1024)
-        .await
-        .expect("checkpoint progress should persist");
-    checkpoint.mark_counter_unknown().await.expect("unknown fence should persist");
-    checkpoint
-        .update_position(0, 3)
-        .await
-        .expect("checkpoint position should persist");
-
-    let restored_checkpoint = CheckpointManager::load_from_disk(disk.clone(), &task_id)
-        .await
-        .expect("checkpoint should reload")
-        .get_checkpoint()
-        .await;
-    let restored_resume = ResumeManager::load_from_disk(disk, &task_id)
-        .await
-        .expect("resume summary should reload")
-        .get_state()
-        .await;
-    assert!(restored_checkpoint.counter_unknown);
-    assert_eq!(restored_checkpoint.processed_bytes, 1024);
-    assert_eq!(restored_checkpoint.current_object_index, 3);
-    assert!(!restored_resume.counter_unknown, "summary is intentionally the torn/older store");
-}
-
 #[tokio::test]
 async fn future_resume_and_checkpoint_schemas_are_rejected() {
     let (temp_dir, disk) = schema_test_disk().await;

crates/heal/src/heal/storage.rs

@@ -22,7 +22,6 @@ use serde::{Deserialize, Serialize};
 use std::sync::Arc;
 use tracing::{debug, error, warn};
 
-use super::progress::stable_generation;
 use super::storage_api::owner::{EcstoreHealLifecycleExpiryContext, ecstore_load_admin_data_usage_from_backend_cached};
 use super::storage_api::storage::{
     BucketInfo, BucketOperations, DiskSetSelector, HealOperations as _, ListOperations as _, ObjectIO as _,
@@ -35,9 +34,6 @@ pub use super::{HealObjectInfo, HealObjectOptions, HealPutObjReader};
 pub struct HealBucketUsageBaseline {
     pub objects_count: u64,
     pub bytes: u64,
-    /// Stable identity of the validated usage snapshot and selected scope.
-    /// `None` is retained for test/legacy providers that cannot expose one.
-    pub generation: Option<u64>,
 }
 
 pub struct HealLifecycleExpiryContext {
@@ -789,52 +785,11 @@ impl HealStorageAPI for ECStoreHealStorage {
         let mut baseline = HealBucketUsageBaseline::default();
         for bucket in buckets {
             if let Some(usage) = info.buckets_usage.get(bucket) {
-                baseline.objects_count = match baseline.objects_count.checked_add(usage.objects_count) {
-                    Some(total) => total,
-                    // A corrupt/overflowing usage snapshot is not a usable
-                    // denominator.  Leave progress indeterminate instead of
-                    // turning saturation into a plausible percentage.
-                    None => return Ok(None),
-                };
-                baseline.bytes = match baseline.bytes.checked_add(usage.size) {
-                    Some(total) => total,
-                    None => return Ok(None),
-                };
+                baseline.objects_count = baseline.objects_count.saturating_add(usage.objects_count);
+                baseline.bytes = baseline.bytes.saturating_add(usage.size);
             }
         }
 
-        let identity = info.snapshot_identity();
-        let mut canonical = Vec::new();
-        match identity.last_update {
-            Some(last_update) => {
-                canonical.push(1);
-                canonical.extend_from_slice(
-                    &last_update
-                        .duration_since(std::time::UNIX_EPOCH)
-                        .unwrap_or_default()
-                        .as_nanos()
-                        .to_be_bytes(),
-                );
-            }
-            None => canonical.push(0),
-        }
-        for value in [identity.scanner_cycle, identity.scanner_epoch] {
-            match value {
-                Some(value) => {
-                    canonical.push(1);
-                    canonical.extend_from_slice(&value.to_be_bytes());
-                }
-                None => canonical.push(0),
-            }
-        }
-        let mut scope = buckets.to_vec();
-        scope.sort_unstable();
-        for bucket in scope {
-            canonical.extend_from_slice(&(bucket.len() as u64).to_be_bytes());
-            canonical.extend_from_slice(bucket.as_bytes());
-        }
-        baseline.generation = Some(stable_generation(&[&canonical]));
-
         Ok(Some(baseline))
     }
 

crates/heal/src/heal/task.rs

@@ -649,7 +649,7 @@ impl HealTask {
 
         let mut progress = self.progress.write().await;
         progress.set_current_object(Some(format!("skipped: {bucket}/{object}")));
-        progress.update_stage(1, 1);
+        progress.update_progress(0, 1, 0, 0);
         Ok(())
     }
 
@@ -733,7 +733,7 @@ impl HealTask {
             "Heal object skipped for data usage cache after transient error"
         );
         let mut progress = self.progress.write().await;
-        progress.update_stage(3, 3);
+        progress.update_progress(3, 3, 0, 0);
         true
     }
 
@@ -757,7 +757,7 @@ impl HealTask {
         );
         let mut progress = self.progress.write().await;
         progress.set_current_object(Some(format!("skipped: {bucket}/{object}")));
-        progress.update_stage(4, 4);
+        progress.update_progress(4, 4, 0, 0);
         true
     }
 
@@ -831,10 +831,6 @@ impl HealTask {
 
         match &result {
             Ok(_) => {
-                // A stage can reach its final step before the durable resume
-                // ledger and cleanup fences commit. Publish terminal 100 only
-                // after the enclosing operation has returned success.
-                self.progress.write().await.mark_completed();
                 let mut status = self.status.write().await;
                 *status = HealTaskStatus::Completed;
                 demote_to_debug_when!(self.heal_type.is_per_object(), info, target: "rustfs::heal::task", {

crates/heal/src/heal/task/heal_bucket.rs

@@ -13,7 +13,6 @@
 // limitations under the License.
 /// bucket/cluster/prefix heal: the recursive bucket-objects sweep and the erasure-set usage baseline
 use super::*;
-use crate::heal::progress::{add_bytes, increment_counter, stable_generation};
 
 impl HealTask {
     pub(super) async fn heal_bucket(&self, bucket: &str) -> Result<()> {
@@ -33,7 +32,7 @@ impl HealTask {
         {
             let mut progress = self.progress.write().await;
             progress.set_current_object(Some(format!("bucket: {bucket}")));
-            progress.update_stage(0, 3);
+            progress.update_progress(0, 3, 0, 0);
         }
 
         // Step 1: Check if bucket exists
@@ -67,7 +66,7 @@ impl HealTask {
 
         {
             let mut progress = self.progress.write().await;
-            progress.update_stage(1, 3);
+            progress.update_progress(1, 3, 0, 0);
         }
 
         // Step 2: Perform bucket heal using ecstore
@@ -123,7 +122,7 @@ impl HealTask {
 
                 if !self.options.recursive {
                     let mut progress = self.progress.write().await;
-                    progress.update_stage(3, 3);
+                    progress.update_progress(3, 3, 0, 0);
                 }
                 Ok(())
             }
@@ -143,7 +142,7 @@ impl HealTask {
                 );
                 {
                     let mut progress = self.progress.write().await;
-                    progress.update_stage(3, 3);
+                    progress.update_progress(3, 3, 0, 0);
                 }
                 Err(Error::TaskExecutionFailed {
                     message: format!("Failed to heal bucket {bucket}: {e}"),
@@ -246,7 +245,6 @@ impl HealTask {
         let mut scanned = 0u64;
         let mut healed = 0u64;
         let mut failed = 0u64;
-        let mut skipped = 0u64;
         let mut retryable_failed = 0u64;
         let mut permanent_failed = 0u64;
         let mut bytes = 0u64;
@@ -288,14 +286,16 @@ impl HealTask {
                 let mut retry = Vec::with_capacity(pending.len());
                 for item in pending {
                     self.check_control_flags().await?;
-                    let mut telemetry_unknown = false;
                     let object = item.name.as_str();
+                    if retry_attempt == 0 {
+                        scanned = scanned.saturating_add(1);
+                    }
                     {
                         let mut progress = self.progress.write().await;
                         progress.set_current_object(Some(format!("{bucket}/{object}")));
+                        progress.update_progress(scanned, healed, failed, bytes);
                     }
 
-                    let mut terminal_outcome = true;
                     let error = match self
                         .await_with_control(
                             self.storage
@@ -304,13 +304,13 @@ impl HealTask {
                         .await
                     {
                         Ok((result, None)) => {
-                            telemetry_unknown |= !increment_counter(&mut healed);
-                            telemetry_unknown |= !add_bytes(&mut bytes, u64::try_from(result.object_size).unwrap_or(u64::MAX));
+                            healed = healed.saturating_add(1);
+                            bytes = bytes.saturating_add(u64::try_from(result.object_size).unwrap_or_default());
                             self.record_result_item(result).await;
                             None
                         }
                         Ok((_, Some(err))) if is_missing_object_dir_heal_result(object, &err) => {
-                            telemetry_unknown |= !increment_counter(&mut healed);
+                            healed = healed.saturating_add(1);
                             debug!(
                                 target: "rustfs::heal::task",
                                 event = EVENT_HEAL_BUCKET_RESULT,
@@ -329,7 +329,6 @@ impl HealTask {
 
                     if let Some(err) = error {
                         if Self::should_skip_data_usage_cache_heal_error(bucket, object, &err) {
-                            telemetry_unknown |= !increment_counter(&mut skipped);
                             warn!(
                                 target: "rustfs::heal::task",
                                 event = EVENT_HEAL_BUCKET_RESULT,
@@ -343,7 +342,6 @@ impl HealTask {
                                 "Heal bucket object repair skipped due to transient metadata error"
                             );
                         } else if err.is_recoverable_heal() && retry_attempt < MAX_BUCKET_OBJECT_HEAL_RETRIES {
-                            terminal_outcome = false;
                             debug!(
                                 target: "rustfs::heal::task",
                                 event = EVENT_HEAL_BUCKET_RESULT,
@@ -359,7 +357,7 @@ impl HealTask {
                             );
                             retry.push(item);
                         } else {
-                            telemetry_unknown |= !increment_counter(&mut failed);
+                            failed = failed.saturating_add(1);
                             if err.is_recoverable_heal() {
                                 retryable_failed = retryable_failed.saturating_add(1);
                             } else {
@@ -385,19 +383,8 @@ impl HealTask {
                         }
                     }
 
-                    if terminal_outcome {
-                        telemetry_unknown |= !increment_counter(&mut scanned);
-                    }
-
-                    if !terminal_outcome {
-                        continue;
-                    }
-
                     let mut progress = self.progress.write().await;
-                    progress.update_object_progress(scanned, healed, failed, skipped, bytes);
-                    if telemetry_unknown {
-                        progress.mark_unknown();
-                    }
+                    progress.update_progress(scanned, healed, failed, bytes);
                 }
                 pending = retry;
                 retry_attempt = retry_attempt.saturating_add(1);
@@ -444,10 +431,7 @@ impl HealTask {
         Ok(())
     }
 
-    pub(super) async fn apply_erasure_set_usage_baseline(&self, buckets: &[String], _set_disk_id: &str) -> Result<()> {
-        if matches!(self.options.scan_mode, HealScanMode::Deep) || matches!(self.source, HealRequestSource::AutoHeal) {
-            return Ok(());
-        }
+    pub(super) async fn apply_erasure_set_usage_baseline(&self, buckets: &[String]) -> Result<()> {
         let baseline = match self
             .await_with_control(self.storage.erasure_set_usage_baseline(buckets))
             .await
@@ -458,18 +442,9 @@ impl HealTask {
             Err(_) => return Ok(()),
         };
 
-        let HealBucketUsageBaseline {
-            objects_count,
-            bytes,
-            generation,
-        } = baseline;
-        let generation = generation.map(|snapshot_generation| stable_generation(&[&snapshot_generation.to_be_bytes()]));
+        let HealBucketUsageBaseline { objects_count, bytes } = baseline;
         let mut progress = self.progress.write().await;
-        if let Some(generation) = generation {
-            progress.set_total_baseline_with_generation(objects_count, bytes, generation);
-        } else {
-            progress.set_total_baseline(objects_count, bytes);
-        }
+        progress.set_total_baseline(objects_count, bytes);
         Ok(())
     }
 }

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

@@ -32,7 +32,7 @@ impl HealTask {
         {
             let mut progress = self.progress.write().await;
             progress.set_current_object(Some(format!("erasure_set: {} ({} buckets)", set_disk_id, buckets.len())));
-            progress.update_stage(0, 4);
+            progress.update_progress(0, 4, 0, 0);
         }
 
         let is_auto_replacement = matches!(self.source, HealRequestSource::AutoHeal) && !self.heal_endpoints.is_empty();
@@ -158,7 +158,7 @@ impl HealTask {
             None
         };
 
-        self.apply_erasure_set_usage_baseline(&buckets, &set_disk_id).await?;
+        self.apply_erasure_set_usage_baseline(&buckets).await?;
 
         let healing_marker = format!("{set_disk_id}:{}", self.id);
         if let Some((disk, resume_manager, _)) = replacement_resume.as_ref() {
@@ -248,7 +248,7 @@ impl HealTask {
                         );
                         {
                             let mut progress = self.progress.write().await;
-                            progress.update_stage(4, 4);
+                            progress.update_progress(4, 4, 0, 0);
                         }
                         return Err(Error::TaskExecutionFailed {
                             message: format!("Failed to heal disk format for {set_disk_id}: {error}"),
@@ -304,7 +304,7 @@ impl HealTask {
                 );
                 {
                     let mut progress = self.progress.write().await;
-                    progress.update_stage(4, 4);
+                    progress.update_progress(4, 4, 0, 0);
                 }
                 return Err(Error::TaskExecutionFailed {
                     message: format!("Failed to heal disk format for {set_disk_id}: {e}"),
@@ -314,7 +314,7 @@ impl HealTask {
 
         {
             let mut progress = self.progress.write().await;
-            progress.update_stage(1, 4);
+            progress.update_progress(1, 4, 0, 0);
         }
 
         // The rebuilt disks are formatted now: mark them as healing so
@@ -343,7 +343,7 @@ impl HealTask {
 
         {
             let mut progress = self.progress.write().await;
-            progress.update_stage(2, 4);
+            progress.update_progress(2, 4, 0, 0);
         }
 
         // Step 3: Heal bucket structure
@@ -427,7 +427,7 @@ impl HealTask {
 
         {
             let mut progress = self.progress.write().await;
-            progress.update_stage(3, 4);
+            progress.update_progress(3, 4, 0, 0);
         }
 
         // Step 4: Execute erasure set heal with resume
@@ -470,7 +470,9 @@ impl HealTask {
         };
 
         {
-            self.progress.write().await.update_stage(4, 4);
+            let mut progress = self.progress.write().await;
+            let bytes_processed = progress.bytes_processed;
+            progress.update_progress(4, 4, 0, bytes_processed);
         }
 
         match result {

crates/heal/src/heal/task/heal_metadata.rs

@@ -32,7 +32,7 @@ impl HealTask {
         {
             let mut progress = self.progress.write().await;
             progress.set_current_object(Some(format!("metadata: {bucket}/{object}")));
-            progress.update_stage(0, 3);
+            progress.update_progress(0, 3, 0, 0);
         }
 
         // Step 1: Check if object exists
@@ -74,7 +74,7 @@ impl HealTask {
 
         {
             let mut progress = self.progress.write().await;
-            progress.update_stage(1, 3);
+            progress.update_progress(1, 3, 0, 0);
         }
 
         // Step 2: Perform metadata heal using ecstore
@@ -122,7 +122,7 @@ impl HealTask {
                     );
                     {
                         let mut progress = self.progress.write().await;
-                        progress.update_stage(3, 3);
+                        progress.update_progress(3, 3, 0, 0);
                     }
                     return Err(Error::TaskExecutionFailed {
                         message: format!("Failed to heal metadata {bucket}/{object}: {e}"),
@@ -145,7 +145,7 @@ impl HealTask {
 
                 {
                     let mut progress = self.progress.write().await;
-                    progress.update_stage(3, 3);
+                    progress.update_progress(3, 3, 0, 0);
                 }
                 self.record_result_item(result).await;
                 Ok(())
@@ -167,7 +167,7 @@ impl HealTask {
                 );
                 {
                     let mut progress = self.progress.write().await;
-                    progress.update_stage(3, 3);
+                    progress.update_progress(3, 3, 0, 0);
                 }
                 Err(Error::TaskExecutionFailed {
                     message: format!("Failed to heal metadata {bucket}/{object}: {e}"),
@@ -194,7 +194,7 @@ impl HealTask {
         {
             let mut progress = self.progress.write().await;
             progress.set_current_object(Some(format!("ec_decode: {bucket}/{object}")));
-            progress.update_stage(0, 3);
+            progress.update_progress(0, 3, 0, 0);
         }
 
         // Step 1: Check if object exists
@@ -236,7 +236,7 @@ impl HealTask {
 
         {
             let mut progress = self.progress.write().await;
-            progress.update_stage(1, 3);
+            progress.update_progress(1, 3, 0, 0);
         }
 
         // Step 2: Perform EC decode heal using ecstore
@@ -284,7 +284,7 @@ impl HealTask {
                     );
                     {
                         let mut progress = self.progress.write().await;
-                        progress.update_stage(3, 3);
+                        progress.update_progress(3, 3, 0, 0);
                     }
                     return Err(Error::TaskExecutionFailed {
                         message: format!("Failed to heal EC decode {bucket}/{object}: {e}"),
@@ -309,7 +309,7 @@ impl HealTask {
 
                 {
                     let mut progress = self.progress.write().await;
-                    progress.update_object_progress(1, 1, 0, 0, object_size);
+                    progress.update_progress(3, 3, 0, object_size);
                 }
                 self.record_result_item(result).await;
                 Ok(())
@@ -331,7 +331,7 @@ impl HealTask {
                 );
                 {
                     let mut progress = self.progress.write().await;
-                    progress.update_stage(3, 3);
+                    progress.update_progress(3, 3, 0, 0);
                 }
                 Err(Error::TaskExecutionFailed {
                     message: format!("Failed to heal EC decode {bucket}/{object}: {e}"),

crates/heal/src/heal/task/heal_object.rs

@@ -36,7 +36,7 @@ impl HealTask {
         {
             let mut progress = self.progress.write().await;
             progress.set_current_object(Some(format!("{bucket}/{object}")));
-            progress.update_stage(0, 4);
+            progress.update_progress(0, 4, 0, 0);
         }
 
         // Step 1: Check if object exists and get metadata
@@ -132,7 +132,7 @@ impl HealTask {
 
         {
             let mut progress = self.progress.write().await;
-            progress.update_stage(1, 3);
+            progress.update_progress(1, 3, 0, 0);
         }
 
         // Step 2: directly call ecstore to perform heal
@@ -187,7 +187,7 @@ impl HealTask {
                         );
                         {
                             let mut progress = self.progress.write().await;
-                            progress.update_stage(3, 3);
+                            progress.update_progress(3, 3, 0, 0);
                         }
                         return Ok(());
                     }
@@ -207,7 +207,7 @@ impl HealTask {
 
                     {
                         let mut progress = self.progress.write().await;
-                        progress.update_stage(3, 3);
+                        progress.update_progress(3, 3, 0, 0);
                     }
 
                     if Self::should_return_typed_heal_error(&e) {
@@ -249,7 +249,7 @@ impl HealTask {
 
                 {
                     let mut progress = self.progress.write().await;
-                    progress.update_object_progress(1, 1, 0, 0, object_size);
+                    progress.update_progress(3, 3, 0, object_size);
                 }
                 self.record_result_item(result).await;
                 Ok(())
@@ -275,7 +275,7 @@ impl HealTask {
                     );
                     {
                         let mut progress = self.progress.write().await;
-                        progress.update_stage(3, 3);
+                        progress.update_progress(3, 3, 0, 0);
                     }
                     return Ok(());
                 }
@@ -295,7 +295,7 @@ impl HealTask {
 
                 {
                     let mut progress = self.progress.write().await;
-                    progress.update_stage(3, 3);
+                    progress.update_progress(3, 3, 0, 0);
                 }
 
                 if Self::should_return_typed_heal_error(&e) {
@@ -414,7 +414,7 @@ impl HealTask {
 
                 {
                     let mut progress = self.progress.write().await;
-                    progress.update_object_progress(1, 1, 0, 0, object_size);
+                    progress.update_progress(4, 4, 0, object_size);
                 }
                 self.record_result_item(result).await;
                 Ok(())

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

@@ -22,7 +22,6 @@ use std::sync::Mutex;
 use tempfile::TempDir;
 
 use super::super::storage_api::status::BucketInfo;
-use crate::heal::progress::HealProgressState;
 
 #[tokio::test]
 async fn retry_request_carries_remaining_timeout_budget() {
@@ -2125,7 +2124,6 @@ async fn erasure_set_heal_applies_usage_baseline_to_progress() {
         usage_baseline: Mutex::new(Some(HealBucketUsageBaseline {
             objects_count: 10,
             bytes: 8,
-            generation: Some(1),
         })),
         ..Default::default()
     });
@@ -2149,55 +2147,10 @@ async fn erasure_set_heal_applies_usage_baseline_to_progress() {
     let progress = task.get_progress().await;
     assert_eq!(progress.objects_total_count, 10);
     assert_eq!(progress.objects_total_size, 8);
-    assert!(progress.baseline_generation.is_some());
-    assert!(progress.baseline_known);
     assert_eq!(progress.bytes_processed, 2);
     assert!((progress.progress_percentage - 25.0).abs() < 0.001);
 }
 
-#[tokio::test]
-async fn erasure_set_disk_walk_keeps_cluster_usage_baseline_indeterminate() {
-    for (scan_mode, source) in [
-        (HealScanMode::Deep, HealRequestSource::Admin),
-        (HealScanMode::Normal, HealRequestSource::AutoHeal),
-    ] {
-        let temp = TempDir::new().expect("temporary directory should be created");
-        let disk = make_resume_disk(&temp).await;
-        let storage = Arc::new(MockStorage {
-            resume_disk: Mutex::new(Some(disk)),
-            usage_baseline: Mutex::new(Some(HealBucketUsageBaseline {
-                objects_count: 10,
-                bytes: 8,
-                generation: Some(1),
-            })),
-            ..Default::default()
-        });
-        let mut request = HealRequest::new(
-            HealType::ErasureSet {
-                buckets: vec!["bucket-a".to_string()],
-                set_disk_id: "pool_0_set_0".to_string(),
-            },
-            HealOptions {
-                scan_mode,
-                timeout: None,
-                ..Default::default()
-            },
-            HealPriority::Normal,
-        );
-        request.source = source;
-        let task = HealTask::from_request(request, storage);
-
-        task.heal_erasure_set(vec!["bucket-a".to_string()], "pool_0_set_0".to_string())
-            .await
-            .expect("erasure set heal should complete");
-
-        let progress = task.get_progress().await;
-        assert!(!progress.baseline_known);
-        assert_eq!(progress.baseline_generation, None);
-        assert_eq!(progress.progress_state, HealProgressState::Indeterminate);
-    }
-}
-
 #[tokio::test]
 async fn erasure_set_heal_ignores_usage_baseline_errors() {
     let temp = TempDir::new().expect("temporary directory should be created");

crates/heal/src/lib.rs

@@ -19,7 +19,7 @@ pub use error::{Error, Result};
 pub use heal::{
     HealManager, HealOperationsSnapshot, HealOptions, HealPriority, HealPriorityCounts, HealRequest, HealSourceCounts, HealType,
     channel::HealChannelProcessor,
-    progress::{HealProgress, aggregate_heal_progress},
+    progress::HealProgress,
     resume::{ReplacementRecoveryRecord, ReplacementRecoveryState, ResumeUtils},
 };
 use rustfs_concurrency::WorkloadAdmissionSnapshotProvider;

crates/kms/tests/vault_ha_failover_live.rs

@@ -16,14 +16,14 @@
 //!
 //! `scripts/test/vault_ha_kms_live.sh` owns the official Vault containers and
 //! kills the active node while this test continuously decrypts through a
-//! surviving standby. KV2 and Transit requests must remain successful, use a
-//! bounded number of attempts, and leave the circuit and in-flight gauges at
-//! zero after a new leader is elected.
+//! surviving standby. KV2 and Transit must recover after the bounded circuit
+//! interval, use a bounded number of attempts, and leave the circuit and
+//! in-flight gauges at zero after a new leader is elected.
 
 use std::collections::HashMap;
 use std::path::{Path, PathBuf};
-use std::sync::Arc;
 use std::sync::atomic::{AtomicBool, AtomicU64, Ordering};
+use std::sync::{Arc, Mutex};
 use std::time::Duration;
 
 use metrics_util::MetricKind;
@@ -43,6 +43,11 @@ const OPERATION_ATTEMPTS: &str = "rustfs_kms_backend_operation_attempts";
 const IN_FLIGHT: &str = "rustfs_kms_backend_in_flight";
 const CIRCUIT_OPEN: &str = "rustfs_kms_backend_circuit_open";
 const MAX_ATTEMPTS: u32 = 10;
+const ATTEMPT_TIMEOUT: Duration = Duration::from_secs(2);
+const HEALTHY_PROGRESS_TIMEOUT: Duration = Duration::from_secs(20);
+// The circuit remains open for 30s after five failed attempts.
+const POST_FAILOVER_PROGRESS_TIMEOUT: Duration = Duration::from_secs(35);
+const FAILOVER_ERROR_POLL_INTERVAL: Duration = Duration::from_millis(100);
 
 type MetricEntry = (
     metrics_util::CompositeKey,
@@ -64,7 +69,7 @@ fn config(backend: KmsBackend, backend_config: BackendConfig) -> KmsConfig {
         backend,
         backend_config,
         allow_insecure_dev_defaults: true,
-        timeout: Duration::from_secs(2),
+        timeout: ATTEMPT_TIMEOUT,
         retry_attempts: MAX_ATTEMPTS,
         enable_cache: false,
         ..KmsConfig::default()
@@ -164,14 +169,31 @@ fn retryable_failures(snapshot: &[MetricEntry], operation: &str) -> u64 {
         .sum()
 }
 
-async fn wait_for_count(counter: &AtomicU64, minimum: u64, description: &str) {
-    tokio::time::timeout(Duration::from_secs(20), async {
+async fn wait_for_count(
+    counter: &AtomicU64,
+    failure: &Mutex<Option<String>>,
+    minimum: u64,
+    description: &str,
+    timeout: Duration,
+) {
+    tokio::time::timeout(timeout, async {
         while counter.load(Ordering::SeqCst) < minimum {
+            if let Some(error) = failure.lock().expect("decrypt failure lock poisoned").as_ref() {
+                panic!(
+                    "{description} worker failed after {} successful decrypts: {error}",
+                    counter.load(Ordering::SeqCst)
+                );
+            }
             tokio::time::sleep(Duration::from_millis(25)).await;
         }
     })
     .await
-    .unwrap_or_else(|_| panic!("timed out waiting for {description}"));
+    .unwrap_or_else(|_| {
+        panic!(
+            "timed out after {timeout:?} waiting for {description}: completed {}, expected {minimum}",
+            counter.load(Ordering::SeqCst)
+        )
+    });
 }
 
 async fn wait_for_file(path: &Path, description: &str) {
@@ -189,7 +211,8 @@ async fn decrypt_loop<B: KmsBackendTrait + Send + Sync + 'static>(
     request: DecryptRequest,
     expected: Vec<u8>,
     completed: Arc<AtomicU64>,
-    failed: Arc<AtomicBool>,
+    allow_failover_errors: Arc<AtomicBool>,
+    failure: Arc<Mutex<Option<String>>>,
     stop: CancellationToken,
 ) {
     while !stop.is_cancelled() {
@@ -197,8 +220,18 @@ async fn decrypt_loop<B: KmsBackendTrait + Send + Sync + 'static>(
             Ok(response) if response.plaintext == expected => {
                 completed.fetch_add(1, Ordering::SeqCst);
             }
-            Ok(_) | Err(_) => {
-                failed.store(true, Ordering::SeqCst);
+            Ok(_) => {
+                *failure.lock().expect("decrypt failure lock poisoned") =
+                    Some("decrypt returned unexpected plaintext".to_string());
+                return;
+            }
+            Err(rustfs_kms::KmsError::BackendError { .. } | rustfs_kms::KmsError::OperationTimedOut { .. })
+                if allow_failover_errors.load(Ordering::SeqCst) =>
+            {
+                tokio::time::sleep(FAILOVER_ERROR_POLL_INTERVAL).await;
+            }
+            Err(error) => {
+                *failure.lock().expect("decrypt failure lock poisoned") = Some(error.to_string());
                 return;
             }
         }
@@ -296,7 +329,9 @@ async fn exercise_failover(snapshotter: &Snapshotter) {
     );
 
     let stop = CancellationToken::new();
-    let failed = Arc::new(AtomicBool::new(false));
+    let allow_failover_errors = Arc::new(AtomicBool::new(false));
+    let kv2_failure = Arc::new(Mutex::new(None));
+    let transit_failure = Arc::new(Mutex::new(None));
     let kv2_completed = Arc::new(AtomicU64::new(0));
     let transit_completed = Arc::new(AtomicU64::new(0));
     let kv2_worker = tokio::spawn(decrypt_loop(
@@ -304,7 +339,8 @@ async fn exercise_failover(snapshotter: &Snapshotter) {
         kv2_request,
         kv2_data_key.plaintext_key,
         Arc::clone(&kv2_completed),
-        Arc::clone(&failed),
+        Arc::clone(&allow_failover_errors),
+        Arc::clone(&kv2_failure),
         stop.clone(),
     ));
     let transit_worker = tokio::spawn(decrypt_loop(
@@ -312,12 +348,21 @@ async fn exercise_failover(snapshotter: &Snapshotter) {
         transit_request,
         transit_data_key.plaintext_key,
         Arc::clone(&transit_completed),
-        Arc::clone(&failed),
+        Arc::clone(&allow_failover_errors),
+        Arc::clone(&transit_failure),
         stop.clone(),
     ));
 
-    wait_for_count(&kv2_completed, 2, "two healthy KV2 decrypts").await;
-    wait_for_count(&transit_completed, 2, "two healthy Transit decrypts").await;
+    wait_for_count(&kv2_completed, &kv2_failure, 2, "two healthy KV2 decrypts", HEALTHY_PROGRESS_TIMEOUT).await;
+    wait_for_count(
+        &transit_completed,
+        &transit_failure,
+        2,
+        "two healthy Transit decrypts",
+        HEALTHY_PROGRESS_TIMEOUT,
+    )
+    .await;
+    allow_failover_errors.store(true, Ordering::SeqCst);
     std::fs::write(&marker, b"ready").expect("publish failover readiness marker");
 
     wait_for_file(&elected, "the replacement Vault leader").await;
@@ -326,18 +371,39 @@ async fn exercise_failover(snapshotter: &Snapshotter) {
 
     let kv2_after_election = kv2_completed.load(Ordering::SeqCst) + 2;
     let transit_after_election = transit_completed.load(Ordering::SeqCst) + 2;
-    wait_for_count(&kv2_completed, kv2_after_election, "post-failover KV2 decrypts").await;
-    wait_for_count(&transit_completed, transit_after_election, "post-failover Transit decrypts").await;
+    wait_for_count(
+        &kv2_completed,
+        &kv2_failure,
+        kv2_after_election,
+        "post-failover KV2 decrypts",
+        POST_FAILOVER_PROGRESS_TIMEOUT,
+    )
+    .await;
+    wait_for_count(
+        &transit_completed,
+        &transit_failure,
+        transit_after_election,
+        "post-failover Transit decrypts",
+        POST_FAILOVER_PROGRESS_TIMEOUT,
+    )
+    .await;
 
     stop.cancel();
     kv2_worker.await.expect("KV2 decrypt worker must join");
     transit_worker.await.expect("Transit decrypt worker must join");
-    assert!(!failed.load(Ordering::SeqCst), "no decrypt may fail or return different plaintext");
+    assert!(
+        kv2_failure.lock().expect("KV2 failure lock poisoned").is_none(),
+        "no KV2 decrypt may fail or return different plaintext"
+    );
+    assert!(
+        transit_failure.lock().expect("Transit failure lock poisoned").is_none(),
+        "no Transit decrypt may fail or return different plaintext"
+    );
 }
 
 #[test]
 #[ignore = "requires a real three-node Vault Raft cluster; run scripts/test/vault_ha_kms_live.sh"]
-fn vault_raft_leader_failure_preserves_kv2_and_transit_decrypts() {
+fn vault_raft_leader_failure_recovers_kv2_and_transit_decrypts() {
     let recorder = DebuggingRecorder::new();
     let snapshotter = recorder.snapshotter();
     metrics::with_local_recorder(&recorder, || {
@@ -349,11 +415,6 @@ fn vault_raft_leader_failure_preserves_kv2_and_transit_decrypts() {
     });
     let snapshot = snapshotter.snapshot().into_vec();
 
-    assert_eq!(
-        counter_value(&snapshot, OPERATIONS_TOTAL, &[("outcome", "circuit_open")]),
-        0,
-        "a bounded leader election must not open the circuit"
-    );
     assert_eq!(
         counter_value(&snapshot, OPERATIONS_TOTAL, &[("outcome", "budget_exhausted")]),
         0,

crates/protos/src/generated/proto_gen/node_service.rs

@@ -1215,10 +1215,7 @@ pub struct ScannerActivityResponse {
     pub dirty_usage_pending: bool,
 }
 #[derive(Clone, Copy, PartialEq, Eq, Hash, ::prost::Message)]
-pub struct BackgroundHealStatusRequest {
-    #[prost(uint32, tag = "1")]
-    pub protocol_version: u32,
-}
+pub struct BackgroundHealStatusRequest {}
 #[derive(Clone, PartialEq, Eq, Hash, ::prost::Message)]
 pub struct BackgroundHealStatusResponse {
     #[prost(bool, tag = "1")]

crates/protos/src/lib.rs

@@ -170,7 +170,6 @@ pub fn internode_rpc_max_message_size() -> usize {
 pub const HEAL_CONTROL_RPC_MAX_MESSAGE_SIZE: usize = heal_control::RESULT_MAX_SIZE + 1024;
 pub const HEAL_CONTROL_PROTOCOL_VERSION: u32 = 3;
 pub const DYNAMIC_CONFIG_PROTOCOL_VERSION: u32 = 1;
-pub const BACKGROUND_HEAL_STATUS_PROTOCOL_VERSION: u32 = 2;
 pub const HEAL_CONTROL_CAPABILITY_PROBE_PREFIX: &[u8] = b"rustfs-heal-control-capability-v3\0";
 pub const REMOTE_VERSION_STATE_CAPABILITY_PROBE_PREFIX: &[u8] = b"rustfs-tier-remote-version-state-capability-v1\0";
 pub const CROSS_POOL_FENCE_CAPABILITY_PROBE_PREFIX: &[u8] = b"rustfs-cross-pool-fence-capability-v1\0";
@@ -2347,28 +2346,10 @@ pub async fn evict_failed_connection_with_log_level(addr: &str, log_level: Conne
 #[cfg(test)]
 mod tests {
     use super::*;
-    use prost::Message as _;
     use std::sync::Mutex;
 
     static INTERNODE_RPC_MSGPACK_ONLY_ENV_LOCK: Mutex<()> = Mutex::new(());
 
-    #[derive(Clone, PartialEq, prost::Message)]
-    struct BackgroundHealStatusRequestV1 {}
-
-    #[test]
-    fn background_heal_status_request_remains_rolling_upgrade_compatible() {
-        let current = proto_gen::node_service::BackgroundHealStatusRequest {
-            protocol_version: BACKGROUND_HEAL_STATUS_PROTOCOL_VERSION,
-        };
-        let encoded = current.encode_to_vec();
-        BackgroundHealStatusRequestV1::decode(encoded.as_slice()).expect("v1 server should ignore the version field");
-
-        let encoded = BackgroundHealStatusRequestV1 {}.encode_to_vec();
-        let decoded = proto_gen::node_service::BackgroundHealStatusRequest::decode(encoded.as_slice())
-            .expect("v2 server should accept a v1 request");
-        assert_eq!(decoded.protocol_version, 0);
-    }
-
     #[derive(Clone, Copy, Debug, Eq, Ord, PartialEq, PartialOrd)]
     struct CompatPayloadField {
         message: &'static str,

crates/protos/src/node.proto

@@ -846,9 +846,7 @@ message ScannerActivityResponse {
   bool dirty_usage_pending = 9;
 }
 
-message BackgroundHealStatusRequest {
-  uint32 protocol_version = 1;
-}
+message BackgroundHealStatusRequest {}
 
 message BackgroundHealStatusResponse {
   bool success = 1;

docs/architecture/decommission-compatibility.md

@@ -153,6 +153,99 @@ No migration step is required for these decisions because this note documents th
 current RustFS behavior. Changing either decision later requires an operator
 compatibility note and updated characterization tests.
 
+## Tier Free Versions During Decommission
+
+A tier free version is an internal xl.meta record (`rustfs_filemeta::FREE_VERSION`,
+flagged `XL_FLAG_FREE_VERSION`) shaped like a delete marker. It is created by
+`MetaObject::init_free_version` when a version whose remote transition completed is
+deleted locally: the visible version is removed and the record keeps the remote-tier
+identity (tier, object name, version id, state, destination id) needed for an
+idempotent remote delete. Free versions are not user-visible versions; `num_versions`
+and all listing/GET paths exclude them.
+
+### Lifecycle And Consumers
+
+Creation: any local delete that removes a version whose transition status is
+`complete` appends the record via `MetaObject::delete_version` →
+`init_free_version` (skipped only when `skip_tier_free_version` is set, as on
+data-movement copies). The same deletes also persist a durable tier-journal
+entry on every user-facing path: S3 single deletes (`execute_delete_object` →
+`delete_object_with_tier_delete_journal`), S3 batch deletes, lifecycle expiry,
+and lifecycle delete-all all prepare and commit a journal entry around the
+delete. A journal entry is omitted when the removed version's transition state
+decodes as `TransitionVersionState::Unknown`, or on internal journal-less
+delete paths that never touch transitioned user objects.
+
+Consumption while the record exists: the background recovery loop started by
+`init_background_expiry` (spawned by `spawn_tier_free_version_recovery_once`,
+enabled by default) scans disks for pending records and re-enqueues them; the
+usage scanner does the same; the lifecycle worker then deletes the remote tier
+object idempotently and only afterwards removes the local record. Heal walks
+include free-version records in metadata healing. Transition planning,
+replication, restore, GET, listings, and usage aggregation never depend on
+them.
+
+### Decommission Handling
+
+The exact decommission inventory loader (`load_file_info_versions_exact` via
+`get_all_file_info_versions`) keeps free-version records inline in `versions`.
+The migration loop handles them before lifecycle expiry and delete-marker
+shortcuts. It selects a target pool using the free-version-aware lookup, then
+writes the original free record to every target disk with the normal metadata
+write quorum. The free-version marker, local version id, transition identity,
+transition state, and destination id are preserved at the FileInfo/metadata
+boundary.
+
+The source record is physically removed only after the target write quorum has
+committed and the source cleanup preflight still matches the exact inventory.
+If the lifecycle worker has already completed the remote delete and removed the
+source record before decommission acquires the source lock, decommission records
+that identity as already consumed and treats the missing source record as safe.
+If target capacity, metadata validation, lock fencing, or quorum fails, the
+source record remains and the entry records `state = "free_version_retained"`
+with reason `tier_free_version_migration_failed`; the worker retries the
+operation on a later pass. A target record with the same version id is accepted
+only when its free-version identity matches; a conflicting ordinary version or
+different free record is an overwrite error. This makes retries idempotent and
+prevents a free record from replacing a user-visible version.
+
+### Reference-Audit Result
+
+After migration, user-facing GET/list/transition/replication/restore paths still
+exclude the record. Recovery, usage scanning, lifecycle tier cleanup, and heal
+continue to see it when they request free versions, so an unresolved remote
+delete remains actionable on the target pool. The committed tier journal remains
+an independent retry source where one exists; it is not used as a reason to drop
+the xl.meta record. In particular, `Unknown` transition state records are
+migrated unchanged rather than discarded: the lifecycle worker retains them if
+remote identity validation cannot make a delete request.
+
+Each migrated record emits `state = "free_version_migrated"` with reason
+`tier_free_version_migrated`. A record consumed before migration emits
+`state = "free_version_consumed"` with reason
+`tier_free_version_already_consumed`. Each failed record emits the retained state
+and failure reason above. The entry also emits a disposition summary with
+migrated, consumed, retained, and total counts. The final decommission sweep uses
+the exact loader, counts free records still present, and emits one retained
+record/reason for each unresolved free version before failing the sweep. This
+makes successful migration, completed cleanup, and retained cleanup obligations
+visible instead of silently omitting free records.
+
+No new S3-visible version or admin response field is needed: free versions remain
+internal and are never counted as user-visible versions. The structured
+`decommission_entry` events are the operational status surface for the
+free-version disposition; the existing decommission item/failed counters still
+report the enclosing object migration result.
+
+Regression guard:
+
+- `decommission_tier_free_version_preserves_remote_identity`
+- `decommission_tier_free_version_resume_requires_write_quorum`
+- `decommission_tier_free_version_commit_rejects_lost_fence`
+- `test_decommission_cleanup_preflight_accepts_migrated_free_version_consumed_from_source`
+- `decommission_entry_skips_cleanup_only_marker_when_free_version_is_present`
+- `decommission_entry_rejects_subquorum_free_version_conflict_and_retains_source`
+
 ## Regression Guard
 
 The queued multi-pool contract is guarded by:

rustfs/src/admin/handlers/heal.rs

@@ -20,7 +20,7 @@ use crate::admin::storage_api::bucket::utils::is_valid_object_prefix;
 use crate::server::ADMIN_PREFIX;
 use crate::server::RemoteAddr;
 use crate::storage::rpc::node_service::heal::{
-    HealControlCoordinator, NodeHealStatusSnapshot, capture_node_heal_status, decode_node_heal_status,
+    HealControlCoordinator, NodeHealProgress, NodeHealStatusSnapshot, capture_node_heal_status, decode_node_heal_status,
     decode_node_replacement_recovery_status, heal_control_coordinator, heal_topology_fingerprint,
 };
 use bytes::Bytes;
@@ -298,7 +298,14 @@ fn background_heal_runtime_state(
     }
 }
 
-type BackgroundHealProgress = rustfs_heal::HealProgress;
+#[derive(Debug, Serialize)]
+#[serde(rename_all = "camelCase")]
+struct BackgroundHealProgress {
+    objects_scanned: u64,
+    objects_healed: u64,
+    objects_failed: u64,
+    bytes_processed: u64,
+}
 
 #[derive(Debug)]
 struct ClusterHealStatusSnapshot {
@@ -337,10 +344,17 @@ fn add_operations(total: &mut rustfs_heal::HealOperationsSnapshot, next: rustfs_
     add_source_counts(&mut total.retrying_by_source, next.retrying_by_source);
 }
 
+fn add_progress(total: &mut BackgroundHealProgress, next: NodeHealProgress) {
+    total.objects_scanned = total.objects_scanned.saturating_add(next.objects_scanned);
+    total.objects_healed = total.objects_healed.saturating_add(next.objects_healed);
+    total.objects_failed = total.objects_failed.saturating_add(next.objects_failed);
+    total.bytes_processed = total.bytes_processed.saturating_add(next.bytes_processed);
+}
+
 fn aggregate_cluster_heal_status(snapshots: Vec<NodeHealStatusSnapshot>) -> ClusterHealStatusSnapshot {
     let mut info = BackgroundHealInfo::default();
     let mut operations = rustfs_heal::HealOperationsSnapshot::default();
-    let mut progress = Vec::new();
+    let mut progress = None;
     let mut any_services_enabled = false;
     let mut any_initialized = false;
 
@@ -357,12 +371,18 @@ fn aggregate_cluster_heal_status(snapshots: Vec<NodeHealStatusSnapshot>) -> Clus
         }
         add_operations(&mut operations, snapshot.operations);
         if let Some(next) = snapshot.progress {
-            progress.push(next);
+            add_progress(
+                progress.get_or_insert(BackgroundHealProgress {
+                    objects_scanned: 0,
+                    objects_healed: 0,
+                    objects_failed: 0,
+                    bytes_processed: 0,
+                }),
+                next,
+            );
         }
     }
 
-    let progress = rustfs_heal::aggregate_heal_progress(progress);
-
     let state = if operations.queue_length > 0 || operations.active_tasks > 0 || operations.retrying_tasks > 0 {
         HealRuntimeState::Active
     } else if any_initialized {
@@ -2181,19 +2201,10 @@ mod tests {
         };
 
         let progress = BackgroundHealProgress {
-            kind: rustfs_heal::heal::progress::HealProgressKind::ObjectSweep,
             objects_scanned: 7,
             objects_healed: 3,
             objects_failed: 1,
-            skipped_objects: 3,
-            objects_total_count: 10,
-            objects_total_size: 8192,
             bytes_processed: 4096,
-            progress_percentage: 50.0,
-            progress_state: rustfs_heal::heal::progress::HealProgressState::Running,
-            baseline_generation: Some(42),
-            baseline_known: true,
-            ..Default::default()
         };
 
         let encoded = encode_background_heal_status(
@@ -2209,14 +2220,7 @@ mod tests {
         assert_eq!(json["progress"]["objectsScanned"], 7);
         assert_eq!(json["progress"]["objectsHealed"], 3);
         assert_eq!(json["progress"]["objectsFailed"], 1);
-        assert_eq!(json["progress"]["skippedObjects"], 3);
-        assert_eq!(json["progress"]["objectsTotalCount"], 10);
-        assert_eq!(json["progress"]["objectsTotalSize"], 8192);
         assert_eq!(json["progress"]["bytesProcessed"], 4096);
-        assert_eq!(json["progress"]["progressState"], "running");
-        assert_eq!(json["progress"]["baselineGeneration"], 42);
-        assert_eq!(json["progress"]["baselineKnown"], true);
-        assert_eq!(json["progress"]["counterUnknown"], false);
     }
 
     #[test]
@@ -2238,18 +2242,10 @@ mod tests {
                 ..Default::default()
             },
             Some(NodeHealProgress {
-                kind: rustfs_heal::heal::progress::HealProgressKind::ObjectSweep,
                 objects_scanned: 3,
                 objects_healed: 1,
                 objects_failed: 0,
-                skipped_objects: 2,
-                objects_total_count: 6,
-                objects_total_size: 400,
                 bytes_processed: 100,
-                progress_state: rustfs_heal::heal::progress::HealProgressState::Running,
-                baseline_generation: Some(9),
-                baseline_known: true,
-                ..Default::default()
             }),
         );
         let peer = NodeHealStatusSnapshot::for_test(
@@ -2269,17 +2265,10 @@ mod tests {
                 ..Default::default()
             },
             Some(NodeHealProgress {
-                kind: rustfs_heal::heal::progress::HealProgressKind::ObjectSweep,
                 objects_scanned: 5,
                 objects_healed: 4,
                 objects_failed: 1,
-                objects_total_count: 4,
-                objects_total_size: 1600,
                 bytes_processed: 900,
-                progress_state: rustfs_heal::heal::progress::HealProgressState::Running,
-                baseline_generation: Some(9),
-                baseline_known: true,
-                ..Default::default()
             }),
         );
 
@@ -2298,15 +2287,7 @@ mod tests {
         assert_eq!(progress.objects_scanned, 8);
         assert_eq!(progress.objects_healed, 5);
         assert_eq!(progress.objects_failed, 1);
-        assert_eq!(progress.skipped_objects, 2);
-        assert_eq!(progress.objects_total_count, 10);
-        assert_eq!(progress.objects_total_size, 2000);
         assert_eq!(progress.bytes_processed, 1000);
-        assert_eq!(progress.progress_percentage, 50.0);
-        assert_eq!(progress.progress_state, rustfs_heal::heal::progress::HealProgressState::Running);
-        assert_eq!(progress.baseline_generation, Some(9));
-        assert!(progress.baseline_known);
-        assert!(!progress.counter_unknown);
 
         assert_eq!(peer_first.state, HealRuntimeState::Active);
         assert_eq!(peer_first.operations, local_first.operations);
@@ -2346,7 +2327,6 @@ mod tests {
             objects_healed: value,
             objects_failed: value,
             bytes_processed: value,
-            ..Default::default()
         };
         let saturated = NodeHealStatusSnapshot::for_test(
             true,

rustfs/src/storage/rpc/node_service.rs

@@ -1912,7 +1912,7 @@ impl Node for NodeService {
 
     async fn background_heal_status(
         &self,
-        request: Request<BackgroundHealStatusRequest>,
+        _request: Request<BackgroundHealStatusRequest>,
     ) -> Result<Response<BackgroundHealStatusResponse>, Status> {
         if self.resolve_object_store().is_none() {
             return Ok(Response::new(BackgroundHealStatusResponse {
@@ -1922,7 +1922,7 @@ impl Node for NodeService {
             }));
         }
         let snapshot = heal::capture_node_heal_status(rustfs_scanner::scanner::BackgroundHealInfo::default()).await;
-        match heal::encode_node_heal_status(&snapshot, request.into_inner().protocol_version) {
+        match heal::encode_node_heal_status(&snapshot) {
             Ok(bg_heal_state) => Ok(Response::new(BackgroundHealStatusResponse {
                 success: true,
                 bg_heal_state: bg_heal_state.into(),

rustfs/src/storage/rpc/node_service/heal.rs

@@ -25,8 +25,7 @@ use std::io::Cursor;
 
 use super::super::encode_msgpack_map;
 
-const NODE_HEAL_STATUS_PREVIOUS_VERSION: u8 = 1;
-const NODE_HEAL_STATUS_VERSION: u8 = 2;
+const NODE_HEAL_STATUS_VERSION: u8 = 1;
 const NODE_HEAL_STATUS_MAX_SIZE: usize = 64 * 1024;
 const NODE_REPLACEMENT_RECOVERY_STATUS_VERSION: u8 = 1;
 const NODE_REPLACEMENT_RECOVERY_STATUS_MAX_SIZE: usize = 64 * 1024;
@@ -173,38 +172,13 @@ pub(crate) fn heal_topology_fingerprint(endpoint_pools: &EndpointServerPools) ->
     Ok(hex_simd::encode_to_string(hasher.finalize(), hex_simd::AsciiCase::Lower))
 }
 
-pub(crate) type NodeHealProgress = rustfs_heal::HealProgress;
-
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, Default, PartialEq, Eq, Serialize, Deserialize)]
 #[serde(rename_all = "camelCase", deny_unknown_fields)]
-struct NodeHealProgressV1 {
-    objects_scanned: u64,
-    objects_healed: u64,
-    objects_failed: u64,
-    bytes_processed: u64,
-}
-
-impl From<&NodeHealProgress> for NodeHealProgressV1 {
-    fn from(progress: &NodeHealProgress) -> Self {
-        Self {
-            objects_scanned: progress.objects_scanned,
-            objects_healed: progress.objects_healed,
-            objects_failed: progress.objects_failed,
-            bytes_processed: progress.bytes_processed,
-        }
-    }
-}
-
-impl From<NodeHealProgressV1> for NodeHealProgress {
-    fn from(progress: NodeHealProgressV1) -> Self {
-        Self {
-            objects_scanned: progress.objects_scanned,
-            objects_healed: progress.objects_healed,
-            objects_failed: progress.objects_failed,
-            bytes_processed: progress.bytes_processed,
-            ..Default::default()
-        }
-    }
+pub(crate) struct NodeHealProgress {
+    pub objects_scanned: u64,
+    pub objects_healed: u64,
+    pub objects_failed: u64,
+    pub bytes_processed: u64,
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -246,48 +220,6 @@ pub(crate) struct NodeHealStatusSnapshot {
     pub progress: Option<NodeHealProgress>,
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
-#[serde(rename_all = "camelCase", deny_unknown_fields)]
-struct NodeHealStatusSnapshotV1 {
-    version: u8,
-    services_enabled: bool,
-    initialized: bool,
-    info: NodeHealInfo,
-    operations: HealOperationsSnapshot,
-    progress: Option<NodeHealProgressV1>,
-}
-
-impl From<&NodeHealStatusSnapshot> for NodeHealStatusSnapshotV1 {
-    fn from(snapshot: &NodeHealStatusSnapshot) -> Self {
-        Self {
-            version: NODE_HEAL_STATUS_PREVIOUS_VERSION,
-            services_enabled: snapshot.services_enabled,
-            initialized: snapshot.initialized,
-            info: snapshot.info.clone(),
-            operations: snapshot.operations,
-            progress: snapshot.progress.as_ref().map(NodeHealProgressV1::from),
-        }
-    }
-}
-
-impl From<NodeHealStatusSnapshotV1> for NodeHealStatusSnapshot {
-    fn from(snapshot: NodeHealStatusSnapshotV1) -> Self {
-        Self {
-            version: snapshot.version,
-            services_enabled: snapshot.services_enabled,
-            initialized: snapshot.initialized,
-            info: snapshot.info,
-            operations: snapshot.operations,
-            progress: snapshot.progress.map(NodeHealProgress::from),
-        }
-    }
-}
-
-#[derive(Deserialize)]
-struct NodeHealStatusVersion {
-    version: u8,
-}
-
 impl NodeHealStatusSnapshot {
     #[cfg(test)]
     pub(crate) fn for_test(
@@ -317,7 +249,14 @@ impl NodeHealStatusSnapshot {
 }
 
 pub(crate) async fn capture_node_heal_status(info: BackgroundHealInfo) -> NodeHealStatusSnapshot {
-    let progress = rustfs_heal::current_heal_progress_snapshot().await;
+    let progress = rustfs_heal::current_heal_progress_snapshot()
+        .await
+        .map(|progress| NodeHealProgress {
+            objects_scanned: progress.objects_scanned,
+            objects_healed: progress.objects_healed,
+            objects_failed: progress.objects_failed,
+            bytes_processed: progress.bytes_processed,
+        });
 
     NodeHealStatusSnapshot {
         version: NODE_HEAL_STATUS_VERSION,
@@ -329,44 +268,23 @@ pub(crate) async fn capture_node_heal_status(info: BackgroundHealInfo) -> NodeHe
     }
 }
 
-pub(crate) fn encode_node_heal_status(snapshot: &NodeHealStatusSnapshot, protocol_version: u32) -> Result<Vec<u8>, String> {
-    let encoded = if protocol_version < rustfs_protos::BACKGROUND_HEAL_STATUS_PROTOCOL_VERSION {
-        encode_msgpack_map(&NodeHealStatusSnapshotV1::from(snapshot))
-    } else {
-        let mut snapshot = snapshot.clone();
-        snapshot.version = NODE_HEAL_STATUS_VERSION;
-        encode_msgpack_map(&snapshot)
-    };
-    encoded.map_err(|err| format!("failed to encode node heal status: {err}"))
+pub(crate) fn encode_node_heal_status(snapshot: &NodeHealStatusSnapshot) -> Result<Vec<u8>, String> {
+    encode_msgpack_map(snapshot).map_err(|err| format!("failed to encode node heal status: {err}"))
 }
 
 pub(crate) fn decode_node_heal_status(data: &[u8]) -> Result<NodeHealStatusSnapshot, String> {
     if data.len() > NODE_HEAL_STATUS_MAX_SIZE {
         return Err("node heal status exceeds size limit".to_string());
     }
-    let decode_version = || {
-        let mut deserializer = Deserializer::new(Cursor::new(data));
-        NodeHealStatusVersion::deserialize(&mut deserializer)
-            .map(|version| (version, deserializer))
-            .map_err(|err| format!("failed to decode node heal status: {err}"))
-    };
-    let (version, deserializer) = decode_version()?;
-    if usize::try_from(deserializer.get_ref().position()).ok() != Some(data.len()) {
-        return Err("node heal status contains trailing data".to_string());
-    }
-
     let mut deserializer = Deserializer::new(Cursor::new(data));
-    let snapshot = match version.version {
-        NODE_HEAL_STATUS_PREVIOUS_VERSION => {
-            NodeHealStatusSnapshotV1::deserialize(&mut deserializer).map(NodeHealStatusSnapshot::from)
-        }
-        NODE_HEAL_STATUS_VERSION => NodeHealStatusSnapshot::deserialize(&mut deserializer),
-        version => return Err(format!("unsupported node heal status version: {version}")),
-    }
-    .map_err(|err| format!("failed to decode node heal status: {err}"))?;
+    let snapshot = NodeHealStatusSnapshot::deserialize(&mut deserializer)
+        .map_err(|err| format!("failed to decode node heal status: {err}"))?;
     if usize::try_from(deserializer.get_ref().position()).ok() != Some(data.len()) {
         return Err("node heal status contains trailing data".to_string());
     }
+    if snapshot.version != NODE_HEAL_STATUS_VERSION {
+        return Err(format!("unsupported node heal status version: {}", snapshot.version));
+    }
     Ok(snapshot)
 }
 
@@ -469,10 +387,9 @@ pub(crate) fn decode_node_replacement_recovery_status(data: &[u8]) -> Result<Nod
 #[cfg(test)]
 mod tests {
     use super::{
-        NODE_HEAL_STATUS_MAX_SIZE, NODE_HEAL_STATUS_PREVIOUS_VERSION, NODE_HEAL_STATUS_VERSION, NodeHealProgress,
-        NodeHealStatusSnapshot, NodeReplacementRecoveryStatusSnapshot, decode_node_heal_status,
-        decode_node_replacement_recovery_status, encode_node_heal_status, encode_node_replacement_recovery_status,
-        heal_control_coordinator, heal_topology_fingerprint,
+        NODE_HEAL_STATUS_MAX_SIZE, NODE_HEAL_STATUS_VERSION, NodeHealProgress, NodeHealStatusSnapshot,
+        NodeReplacementRecoveryStatusSnapshot, decode_node_heal_status, decode_node_replacement_recovery_status,
+        encode_node_heal_status, encode_node_replacement_recovery_status, heal_control_coordinator, heal_topology_fingerprint,
     };
     use crate::storage::storage_api::{
         Endpoint,
@@ -616,72 +533,20 @@ mod tests {
                 ..Default::default()
             },
             Some(NodeHealProgress {
-                kind: rustfs_heal::heal::progress::HealProgressKind::ObjectSweep,
                 objects_scanned: 7,
                 objects_healed: 5,
                 objects_failed: 1,
-                skipped_objects: 1,
-                objects_total_count: 10,
-                objects_total_size: 2048,
                 bytes_processed: 1024,
-                progress_percentage: 50.0,
-                progress_state: rustfs_heal::heal::progress::HealProgressState::Running,
-                baseline_generation: Some(42),
-                baseline_known: true,
-                ..Default::default()
             }),
         );
 
-        let encoded = encode_node_heal_status(&snapshot, rustfs_protos::BACKGROUND_HEAL_STATUS_PROTOCOL_VERSION)
-            .expect("snapshot should encode");
+        let encoded = encode_node_heal_status(&snapshot).expect("snapshot should encode");
         let decoded = decode_node_heal_status(&encoded).expect("snapshot should decode");
         assert_eq!(decoded.version, NODE_HEAL_STATUS_VERSION);
         assert_eq!(decoded.operations.queue_length, 2);
         assert_eq!(decoded.progress, snapshot.progress);
     }
 
-    #[test]
-    fn node_heal_status_v1_encoding_preserves_rolling_compatibility() {
-        let snapshot = NodeHealStatusSnapshot::for_test(
-            true,
-            true,
-            BackgroundHealInfo::default(),
-            HealOperationsSnapshot::default(),
-            Some(NodeHealProgress {
-                objects_scanned: 7,
-                objects_healed: 5,
-                objects_failed: 1,
-                skipped_objects: 3,
-                bytes_processed: 1024,
-                progress_state: rustfs_heal::heal::progress::HealProgressState::Running,
-                baseline_known: true,
-                ..Default::default()
-            }),
-        );
-
-        let encoded =
-            encode_node_heal_status(&snapshot, u32::from(NODE_HEAL_STATUS_PREVIOUS_VERSION)).expect("v1 snapshot should encode");
-        let wire: serde_json::Value = rmp_serde::from_slice(&encoded).expect("v1 snapshot should decode as JSON");
-        let progress = wire["progress"].as_object().expect("v1 progress should be a map");
-        assert_eq!(wire["version"], NODE_HEAL_STATUS_PREVIOUS_VERSION);
-        assert_eq!(progress.len(), 4);
-        assert_eq!(progress["objectsScanned"], 7);
-        assert!(!progress.contains_key("skippedObjects"));
-
-        let decoded = decode_node_heal_status(&encoded).expect("v1 snapshot should decode");
-        let progress = decoded.progress.as_ref().expect("v1 progress should be present");
-        assert_eq!(decoded.version, NODE_HEAL_STATUS_PREVIOUS_VERSION);
-        assert_eq!(progress.objects_scanned, 7);
-        assert_eq!(progress.progress_state, rustfs_heal::heal::progress::HealProgressState::Unknown);
-        assert!(!progress.baseline_known);
-
-        let upgraded = encode_node_heal_status(&decoded, rustfs_protos::BACKGROUND_HEAL_STATUS_PROTOCOL_VERSION)
-            .expect("decoded v1 snapshot should upgrade to v2");
-        let upgraded = decode_node_heal_status(&upgraded).expect("upgraded snapshot should decode");
-        assert_eq!(upgraded.version, NODE_HEAL_STATUS_VERSION);
-        assert_eq!(upgraded.progress.as_ref(), Some(progress));
-    }
-
     #[test]
     fn node_heal_status_rejects_unknown_version() {
         let mut snapshot = NodeHealStatusSnapshot::for_test(
@@ -693,7 +558,7 @@ mod tests {
         );
         snapshot.version += 1;
 
-        let encoded = rmp_serde::to_vec_named(&snapshot).expect("snapshot should encode");
+        let encoded = encode_node_heal_status(&snapshot).expect("snapshot should encode");
         let err = decode_node_heal_status(&encoded).expect_err("unknown version should fail closed");
         assert!(err.contains("unsupported node heal status version"));
     }
@@ -714,22 +579,13 @@ mod tests {
                 "activeBySource": {"scanner": 0, "admin": 1, "autoHeal": 0, "internal": 0, "readRepair": 0},
                 "retryingBySource": {"scanner": 0, "admin": 0, "autoHeal": 0, "internal": 0, "readRepair": 0}
             },
-            "progress": {
-                "objectsScanned": 7,
-                "objectsHealed": 5,
-                "objectsFailed": 1,
-                "bytesProcessed": 1024
-            }
+            "progress": null
         });
         let encoded = rmp_serde::to_vec_named(&fixture).expect("fixture should encode");
         let decoded = decode_node_heal_status(&encoded).expect("fixed v1 fixture should decode");
         assert_eq!(decoded.info().bitrot_start_cycle, 9);
         assert_eq!(decoded.operations.queue_length, 2);
         assert_eq!(decoded.operations.queued_by_source.mrf, 0);
-        let progress = decoded.progress.expect("legacy progress should decode");
-        assert_eq!(progress.objects_scanned, 7);
-        assert!(!progress.baseline_known);
-        assert_eq!(progress.progress_state, rustfs_heal::heal::progress::HealProgressState::Unknown);
     }
 
     #[test]
@@ -775,8 +631,7 @@ mod tests {
             HealOperationsSnapshot::default(),
             None,
         );
-        let encoded = encode_node_heal_status(&snapshot, rustfs_protos::BACKGROUND_HEAL_STATUS_PROTOCOL_VERSION)
-            .expect("snapshot should encode");
+        let encoded = encode_node_heal_status(&snapshot).expect("snapshot should encode");
         let encoded_json: serde_json::Value = rmp_serde::from_slice(&encoded).expect("encoded snapshot should decode as JSON");
         assert_eq!(encoded_json["info"]["bitrotStartTime"], serde_json::json!("2023-11-14T22:13:20.123456Z"));
     }

scripts/test/vault_ha_kms_live.sh

@@ -241,7 +241,7 @@ env \
   RUSTFS_TEST_VAULT_FAILOVER_MARKER="$MARKER" \
   RUSTFS_TEST_VAULT_OLD_LEADER="$OLD_LEADER" \
   cargo test -p rustfs-kms --test vault_ha_failover_live \
-    vault_raft_leader_failure_preserves_kv2_and_transit_decrypts -- \
+    vault_raft_leader_failure_recovers_kv2_and_transit_decrypts -- \
     --ignored --nocapture --test-threads=1 &
 TEST_PID=$!