feat(heal): add progress and trace observability (#6179)

* feat(heal): track erasure set progress baseline

Record erasure-set heal byte progress from per-object results and seed progress totals from complete usage-cache snapshots when available.

Keep usage-cache failures observational so heal execution continues without a baseline.

Co-Authored-By: heihutu <heihutu@gmail.com>

* feat(heal): skip filtered erasure set versions

Skip erasure-set versions written after the durable heal start time, and queue lifecycle-expired versions for expiry before skipping them.

Track new-version and ILM-expired skips separately so progress can explain completed baseline work without treating these skips as retry-blocking failures.

Co-Authored-By: heihutu <heihutu@gmail.com>

* feat(heal): wire abandoned data-dir cleanup check

Connect check_abandoned_parts through ECStore, pool, and set layers so heal can invoke the existing orphan data-dir reclaim path instead of returning NotImplemented.

Add dry-run support to the reclaim scan and cover dry-run plus scoped set behavior with regression tests.

Co-Authored-By: heihutu <heihutu@gmail.com>

* feat(obs): add heal scanner trace bus

Introduce an in-process broadcast trace bus with typed heal and scanner events, lazy event construction, and bounded lagged-subscriber behavior.

Cover zero-subscriber publishing, subscription delivery, drop accounting, and lagged receivers with focused common-crate tests.

Co-Authored-By: heihutu <heihutu@gmail.com>

* feat(obs): stream heal trace events from admin API

Wire the admin trace endpoint to the common trace bus for heal/scanner events, including kind, regex, and threshold filtering.

Co-Authored-By: heihutu <heihutu@gmail.com>

* feat(obs): emit heal trace events

Publish heal task lifecycle and abandoned-parts cleanup events through the common trace bus so the admin trace stream has live heal diagnostics.

Co-Authored-By: heihutu <heihutu@gmail.com>

* feat(obs): emit scanner trace events

Publish scanner folder, lifecycle action, and heal-candidate events through the common trace bus for live admin scanner diagnostics.

Co-Authored-By: heihutu <heihutu@gmail.com>

* fix(heal): route data usage loader through storage api

Keep ECStore data-usage facade access behind the heal storage_api boundary so architecture migration guards can validate the heal progress path.

Co-Authored-By: heihutu <heihutu@gmail.com>

* perf(heal): avoid lifecycle snapshots on ordinary heal pages

Only request lifecycle object snapshots when the heal pass has lifecycle expiry context. This keeps ordinary listing and disk-walk pages from cloning FileInfo/ObjectInfo payloads while preserving the skip path that queues expired versions.

Co-Authored-By: heihutu <heihutu@gmail.com>

* test(heal): update bug-fix mocks for lifecycle snapshots

Carry the lifecycle snapshot opt-in argument through the remaining heal bug-fix test mocks so all-targets clippy covers the updated storage trait.

Co-Authored-By: heihutu <heihutu@gmail.com>

* test(rustfs): sync heal storage mock signature

Update the rustfs storage RPC test mock for the lifecycle snapshot opt-in argument and cover it with rustfs all-targets clippy.

Co-Authored-By: heihutu <heihutu@gmail.com>

* test(e2e): allocate smoke ports across nextest processes

Serialize E2E port selection with a small /tmp allocator so nextest workers do not reuse the same just-released ephemeral port before RustFS binds it.

Co-Authored-By: heihutu <heihutu@gmail.com>

---------

Co-authored-by: heihutu <heihutu@gmail.com>

Cargo.lock

@@ -9280,6 +9280,7 @@ dependencies = [
  "s3s",
  "serde",
  "serde_json",
+ "smallvec",
  "tokio",
  "tonic",
  "tracing",

crates/common/Cargo.toml

@@ -42,6 +42,7 @@ chrono = { workspace = true, features = ["serde"] }
 jiff = { workspace = true, features = ["serde"] }
 metrics = { workspace = true }
 serde = { workspace = true, features = ["derive"] }
+smallvec = { workspace = true }
 rmp-serde = { workspace = true }
 s3s = { workspace = true, features = ["minio"] }
 tracing = { workspace = true }

crates/common/src/lib.rs

@@ -19,6 +19,7 @@ pub mod last_minute;
 pub mod metrics;
 mod readiness;
 pub mod table_catalog;
+pub mod trace_bus;
 
 pub use globals::*;
 pub use readiness::{GlobalReadiness, SystemStage};

crates/common/src/trace_bus.rs

@@ -0,0 +1,333 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use smallvec::SmallVec;
+use std::{
+    sync::{
+        Arc, OnceLock,
+        atomic::{AtomicUsize, Ordering},
+    },
+    time::{Duration, SystemTime},
+};
+use tokio::sync::broadcast;
+
+const DEFAULT_TRACE_BUS_CAPACITY: usize = 1024;
+const TRACE_ATTR_INLINE_CAPACITY: usize = 8;
+
+static GLOBAL_TRACE_BUS: OnceLock<TraceBus> = OnceLock::new();
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum TraceKind {
+    Heal,
+    Scanner,
+}
+
+impl TraceKind {
+    pub const fn as_str(self) -> &'static str {
+        match self {
+            Self::Heal => "heal",
+            Self::Scanner => "scanner",
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum TraceFunc {
+    HealTask,
+    HealBucket,
+    HealObject,
+    HealCheckAbandonedParts,
+    HealErasureSetPage,
+    ScannerFolder,
+    ScannerIlmAction,
+    ScannerHealCandidate,
+    Dropped,
+}
+
+impl TraceFunc {
+    pub const fn as_str(self) -> &'static str {
+        match self {
+            Self::HealTask => "heal.Task",
+            Self::HealBucket => "heal.Bucket",
+            Self::HealObject => "heal.Object",
+            Self::HealCheckAbandonedParts => "heal.CheckAbandonedParts",
+            Self::HealErasureSetPage => "heal.ErasureSetPage",
+            Self::ScannerFolder => "scanner.Folder",
+            Self::ScannerIlmAction => "scanner.IlmAction",
+            Self::ScannerHealCandidate => "scanner.HealCandidate",
+            Self::Dropped => "trace.Dropped",
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum TraceVal {
+    Bool(bool),
+    U64(u64),
+    I64(i64),
+    Str(Arc<str>),
+}
+
+impl From<bool> for TraceVal {
+    fn from(value: bool) -> Self {
+        Self::Bool(value)
+    }
+}
+
+impl From<u64> for TraceVal {
+    fn from(value: u64) -> Self {
+        Self::U64(value)
+    }
+}
+
+impl From<i64> for TraceVal {
+    fn from(value: i64) -> Self {
+        Self::I64(value)
+    }
+}
+
+impl From<&str> for TraceVal {
+    fn from(value: &str) -> Self {
+        Self::Str(Arc::from(value))
+    }
+}
+
+impl From<String> for TraceVal {
+    fn from(value: String) -> Self {
+        Self::Str(Arc::from(value))
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct TraceAttr {
+    pub key: &'static str,
+    pub value: TraceVal,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct TraceEvent {
+    pub kind: TraceKind,
+    pub func: TraceFunc,
+    pub time: SystemTime,
+    pub bucket: Option<Arc<str>>,
+    pub object: Option<Arc<str>>,
+    pub duration: Duration,
+    pub bytes: u64,
+    pub attrs: SmallVec<[TraceAttr; TRACE_ATTR_INLINE_CAPACITY]>,
+}
+
+impl TraceEvent {
+    pub fn new(kind: TraceKind, func: TraceFunc) -> Self {
+        Self {
+            kind,
+            func,
+            time: SystemTime::now(),
+            bucket: None,
+            object: None,
+            duration: Duration::ZERO,
+            bytes: 0,
+            attrs: SmallVec::new(),
+        }
+    }
+
+    pub fn with_bucket(mut self, bucket: impl Into<Arc<str>>) -> Self {
+        self.bucket = Some(bucket.into());
+        self
+    }
+
+    pub fn with_object(mut self, object: impl Into<Arc<str>>) -> Self {
+        self.object = Some(object.into());
+        self
+    }
+
+    pub fn with_duration(mut self, duration: Duration) -> Self {
+        self.duration = duration;
+        self
+    }
+
+    pub fn with_bytes(mut self, bytes: u64) -> Self {
+        self.bytes = bytes;
+        self
+    }
+
+    pub fn with_attr(mut self, key: &'static str, value: impl Into<TraceVal>) -> Self {
+        self.attrs.push(TraceAttr {
+            key,
+            value: value.into(),
+        });
+        self
+    }
+}
+
+#[derive(Debug)]
+pub struct TraceBus {
+    sender: broadcast::Sender<Arc<TraceEvent>>,
+    subscriber_count: Arc<AtomicUsize>,
+}
+
+impl TraceBus {
+    pub fn new(capacity: usize) -> Self {
+        let capacity = capacity.max(1);
+        let (sender, _receiver) = broadcast::channel(capacity);
+        Self {
+            sender,
+            subscriber_count: Arc::new(AtomicUsize::new(0)),
+        }
+    }
+
+    pub fn subscriber_count(&self) -> usize {
+        self.subscriber_count.load(Ordering::Acquire)
+    }
+
+    pub fn subscribe(&self) -> TraceSubscription {
+        let receiver = self.sender.subscribe();
+        self.subscriber_count.fetch_add(1, Ordering::AcqRel);
+        TraceSubscription {
+            receiver,
+            subscriber_count: Arc::clone(&self.subscriber_count),
+        }
+    }
+
+    pub fn emit(&self, build: impl FnOnce() -> TraceEvent) -> bool {
+        if self.subscriber_count() == 0 {
+            return false;
+        }
+
+        self.sender.send(Arc::new(build())).is_ok()
+    }
+}
+
+impl Default for TraceBus {
+    fn default() -> Self {
+        Self::new(DEFAULT_TRACE_BUS_CAPACITY)
+    }
+}
+
+#[derive(Debug)]
+pub struct TraceSubscription {
+    receiver: broadcast::Receiver<Arc<TraceEvent>>,
+    subscriber_count: Arc<AtomicUsize>,
+}
+
+impl TraceSubscription {
+    pub async fn recv(&mut self) -> Result<Arc<TraceEvent>, broadcast::error::RecvError> {
+        self.receiver.recv().await
+    }
+
+    pub fn try_recv(&mut self) -> Result<Arc<TraceEvent>, broadcast::error::TryRecvError> {
+        self.receiver.try_recv()
+    }
+}
+
+impl Drop for TraceSubscription {
+    fn drop(&mut self) {
+        self.subscriber_count.fetch_sub(1, Ordering::AcqRel);
+    }
+}
+
+pub fn global_trace_bus() -> &'static TraceBus {
+    GLOBAL_TRACE_BUS.get_or_init(TraceBus::default)
+}
+
+pub fn subscribe_trace_events() -> TraceSubscription {
+    global_trace_bus().subscribe()
+}
+
+pub fn trace_emit(build: impl FnOnce() -> TraceEvent) -> bool {
+    global_trace_bus().emit(build)
+}
+
+pub fn trace_subscriber_count() -> usize {
+    global_trace_bus().subscriber_count()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::sync::atomic::AtomicUsize;
+
+    #[test]
+    fn trace_emit_skips_builder_without_subscribers() {
+        let bus = TraceBus::new(4);
+        let built = AtomicUsize::new(0);
+
+        let sent = bus.emit(|| {
+            built.fetch_add(1, Ordering::Relaxed);
+            TraceEvent::new(TraceKind::Heal, TraceFunc::HealTask)
+        });
+
+        assert!(!sent);
+        assert_eq!(built.load(Ordering::Relaxed), 0);
+    }
+
+    #[tokio::test]
+    async fn trace_subscriber_receives_event() {
+        let bus = TraceBus::new(4);
+        let mut subscription = bus.subscribe();
+
+        assert!(bus.emit(|| {
+            TraceEvent::new(TraceKind::Heal, TraceFunc::HealObject)
+                .with_bucket("bucket")
+                .with_object("object")
+                .with_duration(Duration::from_millis(7))
+                .with_bytes(11)
+                .with_attr("dry", true)
+        }));
+
+        let event = subscription
+            .recv()
+            .await
+            .expect("subscriber should receive emitted trace event");
+
+        assert_eq!(event.kind, TraceKind::Heal);
+        assert_eq!(event.func, TraceFunc::HealObject);
+        assert_eq!(event.bucket.as_deref(), Some("bucket"));
+        assert_eq!(event.object.as_deref(), Some("object"));
+        assert_eq!(event.duration, Duration::from_millis(7));
+        assert_eq!(event.bytes, 11);
+        assert_eq!(
+            event.attrs.as_slice(),
+            &[TraceAttr {
+                key: "dry",
+                value: TraceVal::Bool(true)
+            }]
+        );
+    }
+
+    #[test]
+    fn trace_subscription_drop_decrements_count() {
+        let bus = TraceBus::new(4);
+        let subscription = bus.subscribe();
+
+        assert_eq!(bus.subscriber_count(), 1);
+        drop(subscription);
+        assert_eq!(bus.subscriber_count(), 0);
+    }
+
+    #[tokio::test]
+    async fn lagged_subscriber_drops_events_without_blocking_publishers() {
+        let bus = TraceBus::new(2);
+        let mut subscription = bus.subscribe();
+
+        for index in 0_u64..4 {
+            assert!(bus.emit(|| { TraceEvent::new(TraceKind::Scanner, TraceFunc::ScannerFolder).with_attr("index", index) }));
+        }
+
+        let err = subscription
+            .recv()
+            .await
+            .expect_err("receiver should observe lag instead of blocking publishers");
+        assert!(matches!(err, broadcast::error::RecvError::Lagged(_)));
+    }
+}

crates/e2e_test/src/common.rs

@@ -32,6 +32,7 @@ use rustfs_signer::sign_v4;
 use s3s::Body;
 use std::ffi::OsStr;
 use std::fs as stdfs;
+use std::io::ErrorKind;
 use std::path::{Path, PathBuf};
 use std::process::{Child, Command, Stdio};
 use std::sync::Once;
@@ -51,6 +52,11 @@ pub(crate) const FAST_DATA_USAGE_SCANNER_ENV: &[(&str, &str)] =
     &[("RUSTFS_SCANNER_CYCLE", "1"), ("RUSTFS_SCANNER_START_DELAY_SECS", "0")];
 pub const TEST_BUCKET: &str = "e2e-test-bucket";
 const RUSTFS_FULL_FEATURE: &str = "full";
+const TEST_PORT_MIN: u16 = 20_000;
+const TEST_PORT_RANGE: u16 = 40_000;
+const TEST_PORT_COUNTER_PATH: &str = "/tmp/rustfs_e2e_next_port";
+const TEST_PORT_LOCK_DIR: &str = "/tmp/rustfs_e2e_port_allocator.lock";
+const TEST_PORT_LOCK_STALE_AFTER: Duration = Duration::from_secs(30);
 
 fn capture_log_path(log_dir: &Path, temp_dir: &str) -> Option<PathBuf> {
     let temp_name = Path::new(temp_dir).file_name()?.to_string_lossy();
@@ -67,6 +73,64 @@ fn configured_capture_log_path(temp_dir: &str) -> Option<String> {
     capture_log_path(Path::new(&log_dir), temp_dir).map(|path| path.to_string_lossy().into_owned())
 }
 
+struct PortAllocatorGuard;
+
+impl PortAllocatorGuard {
+    async fn acquire() -> Result<Self, Box<dyn std::error::Error + Send + Sync>> {
+        loop {
+            match stdfs::create_dir(TEST_PORT_LOCK_DIR) {
+                Ok(()) => return Ok(Self),
+                Err(err) if err.kind() == ErrorKind::AlreadyExists => {
+                    remove_stale_port_allocator_lock();
+                    sleep(Duration::from_millis(10)).await;
+                }
+                Err(err) => return Err(err.into()),
+            }
+        }
+    }
+}
+
+impl Drop for PortAllocatorGuard {
+    fn drop(&mut self) {
+        let _ = stdfs::remove_dir(TEST_PORT_LOCK_DIR);
+    }
+}
+
+fn advance_test_port(port: u16) -> u16 {
+    let offset = (port - TEST_PORT_MIN + 1) % TEST_PORT_RANGE;
+    TEST_PORT_MIN + offset
+}
+
+fn seeded_test_port() -> u16 {
+    let offset = (Uuid::new_v4().as_u128() % u128::from(TEST_PORT_RANGE)) as u16;
+    TEST_PORT_MIN + offset
+}
+
+fn read_next_test_port() -> u16 {
+    stdfs::read_to_string(TEST_PORT_COUNTER_PATH)
+        .ok()
+        .and_then(|value| value.trim().parse::<u16>().ok())
+        .filter(|port| (TEST_PORT_MIN..TEST_PORT_MIN + TEST_PORT_RANGE).contains(port))
+        .unwrap_or_else(seeded_test_port)
+}
+
+fn remove_stale_port_allocator_lock() {
+    let Ok(metadata) = stdfs::metadata(TEST_PORT_LOCK_DIR) else {
+        return;
+    };
+    let Ok(modified) = metadata.modified() else {
+        return;
+    };
+    if modified.elapsed().is_ok_and(|elapsed| elapsed > TEST_PORT_LOCK_STALE_AFTER) {
+        let _ = stdfs::remove_dir(TEST_PORT_LOCK_DIR);
+    }
+}
+
+fn write_next_test_port(port: u16) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    stdfs::write(TEST_PORT_COUNTER_PATH, port.to_string())?;
+    Ok(())
+}
+
 pub(crate) fn capture_command_logs(
     command: &mut Command,
     log_path: Option<&str>,
@@ -508,10 +572,21 @@ impl RustFSTestEnvironment {
     /// Find an available port for the test
     pub async fn find_available_port() -> Result<u16, Box<dyn std::error::Error + Send + Sync>> {
         use std::net::TcpListener;
-        let listener = TcpListener::bind("127.0.0.1:0")?;
+        let _guard = PortAllocatorGuard::acquire().await?;
-        let port = listener.local_addr()?.port();
+        let mut next_port = read_next_test_port();
-        drop(listener);
+
-        Ok(port)
+        for _ in 0..TEST_PORT_RANGE {
+            let port = next_port;
+            next_port = advance_test_port(next_port);
+            write_next_test_port(next_port)?;
+
+            if let Ok(listener) = TcpListener::bind(("127.0.0.1", port)) {
+                drop(listener);
+                return Ok(port);
+            }
+        }
+
+        Err("no available E2E test port found".into())
     }
 
     /// Kill any existing RustFS processes

crates/ecstore/src/api/mod.rs

@@ -483,6 +483,7 @@ pub mod store_list {
 }
 
 pub mod storage {
+    pub use crate::core::pools::HealLifecycleExpiryContext;
     pub use crate::store::HealWalkVersion;
     pub use crate::store::{
         ECStore, all_local_disk, all_local_disk_path, find_local_disk_by_ref, init_local_disks,

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

@@ -19,7 +19,7 @@ pub mod core;
 pub mod evaluator;
 pub mod manual_transition_job;
 mod metadata_boundary;
-pub(crate) use metadata_boundary::get_expiry_configs;
+pub(crate) use metadata_boundary::{LifecycleExpiryConfigs, get_expiry_configs};
 mod object_lock_boundary;
 pub use self::core as lifecycle;
 mod replication_sink;

crates/ecstore/src/core/pools.rs

@@ -16,6 +16,7 @@ use crate::bucket::replication::replication_state_from_filemeta;
 use crate::bucket::versioning_sys::BucketVersioningSys;
 use crate::bucket::{
     lifecycle::{
+        LifecycleExpiryConfigs,
         bucket_lifecycle_audit::LcEventSrc,
         bucket_lifecycle_ops::{
             LifecycleOps, apply_expiry_on_transitioned_object, apply_expiry_rule_in, eval_action_from_lifecycle,
@@ -2335,6 +2336,10 @@ fn lifecycle_action_removes_data_movement_version(action: IlmAction) -> bool {
     )
 }
 
+fn lifecycle_action_skips_heal_version(action: IlmAction) -> bool {
+    action.delete()
+}
+
 fn resolve_data_movement_lifecycle_expiry_result(action: IlmAction, apply_actions: bool, applied: bool) -> Result<bool> {
     if !apply_actions || applied {
         return Ok(true);
@@ -2385,7 +2390,80 @@ pub(crate) async fn should_skip_lifecycle_for_data_movement(
     }
 }
 
+pub struct HealLifecycleExpiryContext {
+    configs: LifecycleExpiryConfigs,
+}
+
 impl ECStore {
+    pub async fn load_heal_lifecycle_expiry_context(&self, bucket: &str) -> Result<Option<HealLifecycleExpiryContext>> {
+        if bucket == RUSTFS_META_BUCKET {
+            return Ok(None);
+        }
+
+        let configs = get_expiry_configs(self, bucket).await?;
+        if configs.lifecycle.is_none() {
+            return Ok(None);
+        }
+
+        Ok(Some(HealLifecycleExpiryContext { configs }))
+    }
+
+    pub async fn enqueue_heal_lifecycle_expiry(
+        self: &Arc<Self>,
+        context: &HealLifecycleExpiryContext,
+        bucket: &str,
+        object: &str,
+        version_id: Option<&str>,
+        object_info: Option<&crate::object_api::ObjectInfo>,
+    ) -> Result<bool> {
+        let Some(lifecycle_config) = context.configs.lifecycle.as_ref() else {
+            return Ok(false);
+        };
+
+        let object_info = if let Some(object_info) = object_info {
+            if object_info.bucket != bucket || object_info.name != object {
+                return Ok(false);
+            }
+            let snapshot_version_id = object_info
+                .version_id
+                .filter(|version_id| !version_id.is_nil())
+                .map(|version_id| version_id.to_string());
+            if snapshot_version_id.as_deref() != version_id {
+                return Ok(false);
+            }
+            object_info.clone()
+        } else {
+            match self
+                .get_object_info(
+                    bucket,
+                    object,
+                    &ObjectOptions {
+                        version_id: version_id.map(str::to_string),
+                        versioned: version_id.is_some(),
+                        expected_bucket_incarnation_id: Some(context.configs.bucket_incarnation_id),
+                        ..Default::default()
+                    },
+                )
+                .await
+            {
+                Ok(object_info) => object_info,
+                Err(err) if is_err_object_not_found(&err) || is_err_version_not_found(&err) => return Ok(false),
+                Err(err) => return Err(err),
+            }
+        };
+
+        let event = eval_action_from_lifecycle(lifecycle_config, context.configs.object_lock.as_deref(), &object_info).await;
+        if !lifecycle_action_skips_heal_version(event.action) {
+            return Ok(false);
+        }
+
+        if lifecycle_delete_all_versions_blocked_by_replication(self.clone(), bucket, &object_info.name, event.action).await? {
+            return Ok(false);
+        }
+
+        Ok(apply_expiry_rule_in(self.clone(), &event, &LcEventSrc::Scanner, &object_info).await)
+    }
+
     async fn save_current_pool_meta(&self) -> Result<()> {
         let _save_guard = self.pool_meta_save_gate.lock().await;
         let snapshot = {
@@ -4287,6 +4365,19 @@ mod tests {
         ));
     }
 
+    #[test]
+    fn lifecycle_action_skips_heal_version_for_every_delete_action() {
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteVersionAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteRestoredAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteRestoredVersionAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteAllVersionsAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DelMarkerDeleteAllVersionsAction));
+        assert!(!lifecycle_action_skips_heal_version(IlmAction::TransitionAction));
+        assert!(!lifecycle_action_skips_heal_version(IlmAction::TransitionVersionAction));
+        assert!(!lifecycle_action_skips_heal_version(IlmAction::NoneAction));
+    }
+
     #[test]
     fn resolve_data_movement_lifecycle_expiry_result_allows_dry_run_skip() {
         let skip = resolve_data_movement_lifecycle_expiry_result(IlmAction::DeleteVersionAction, false, false)

crates/ecstore/src/core/sets.rs

@@ -1140,11 +1140,11 @@ impl crate::storage_api_contracts::heal::HealOperations for Sets {
 
         Err(Error::DiskNotFound)
     }
-    #[tracing::instrument(skip(self))]
+    #[tracing::instrument(level = "debug", skip(self, opts), fields(bucket = %bucket, object = %object, dry_run = opts.dry_run))]
-    async fn check_abandoned_parts(&self, _bucket: &str, _object: &str, _opts: &HealOpts) -> Result<()> {
+    async fn check_abandoned_parts(&self, bucket: &str, object: &str, opts: &HealOpts) -> Result<()> {
-        // Multipart orphan reconciliation is intentionally retained above the pool/set layers
+        self.get_disks_for_heal_object(object, opts)?
-        // until there is a concrete caller and a stable lower-level contract to implement.
+            .check_abandoned_parts(bucket, object, opts)
-        Err(StorageError::NotImplemented)
+            .await
     }
 }
 
@@ -1996,7 +1996,7 @@ mod tests {
     }
 
     #[tokio::test]
-    async fn sets_check_abandoned_parts_returns_typed_not_implemented_error() {
+    async fn sets_check_abandoned_parts_rejects_invalid_set_scope() {
         let format = FormatV3::new(1, 1);
         let sets = Sets {
             id: format.id,
@@ -2021,10 +2021,21 @@ mod tests {
         };
 
         let err = sets
-            .check_abandoned_parts("bucket", "object", &HealOpts::default())
+            .check_abandoned_parts(
+                "bucket",
+                "object",
+                &HealOpts {
+                    set: Some(1),
+                    ..Default::default()
+                },
+            )
             .await
-            .expect_err("abandoned-parts ownership should stay above the pool/set storage layers");
+            .expect_err("out-of-range abandoned-parts set scope must fail closed");
-        assert!(matches!(err, StorageError::NotImplemented));
+        assert!(
+            matches!(err, StorageError::InvalidArgument(_, ref field, ref reason)
+                if field == "set" && reason.contains("invalid heal set index 1")),
+            "unexpected invalid set error: {err:?}"
+        );
     }
 
     // Builds a single-set `Sets` over `SET_DRIVE_COUNT` local temp-dir disks,

crates/ecstore/src/set_disk/core/io_primitives.rs

@@ -4860,6 +4860,14 @@ impl SetDisks {
     /// is best-effort maintenance: individual delete failures are logged and
     /// skipped rather than propagated.
     pub(crate) async fn reclaim_orphan_data_dirs(&self, bucket: &str, object: &str) -> disk::error::Result<usize> {
+        self.reclaim_orphan_data_dirs_inner(bucket, object, false).await
+    }
+
+    pub(crate) async fn dry_run_reclaim_orphan_data_dirs(&self, bucket: &str, object: &str) -> disk::error::Result<usize> {
+        self.reclaim_orphan_data_dirs_inner(bucket, object, true).await
+    }
+
+    async fn reclaim_orphan_data_dirs_inner(&self, bucket: &str, object: &str, dry_run: bool) -> disk::error::Result<usize> {
         let disks = self.get_disks_internal().await;
 
         // Phase 1 (read-only): build the referenced-data-dir union and record the
@@ -4967,6 +4975,20 @@ impl SetDisks {
                     continue;
                 }
                 let stray = format!("{object}/{dir}");
+                if dry_run {
+                    removed += 1;
+                    debug!(
+                        target: "rustfs_ecstore::set_disk",
+                        event = "heal_abandoned_parts",
+                        component = "ecstore",
+                        subsystem = "heal",
+                        state = "dry_run_matched",
+                        result = "matched",
+                        bucket, object, data_dir = %dir,
+                        "Heal abandoned parts dry-run matched orphaned data directory"
+                    );
+                    continue;
+                }
                 match disk
                     .delete(
                         bucket,

crates/ecstore/src/set_disk/mod.rs

@@ -6998,6 +6998,100 @@ mod tests {
         assert!(object_dir.join(STORAGE_FORMAT_FILE).exists(), "metadata must be preserved");
     }
 
+    async fn recv_abandoned_parts_trace(
+        trace: &mut rustfs_common::trace_bus::TraceSubscription,
+        bucket: &str,
+        object: &str,
+        state: &str,
+    ) -> rustfs_common::trace_bus::TraceEvent {
+        for _ in 0..32 {
+            let event = tokio::time::timeout(std::time::Duration::from_secs(1), trace.recv())
+                .await
+                .expect("abandoned-parts trace event should arrive")
+                .expect("trace bus should stay open");
+            if event.kind == rustfs_common::trace_bus::TraceKind::Heal
+                && event.func == rustfs_common::trace_bus::TraceFunc::HealCheckAbandonedParts
+                && event.bucket.as_deref() == Some(bucket)
+                && event.object.as_deref() == Some(object)
+                && trace_attr_string(&event, "state").as_deref() == Some(state)
+            {
+                return (*event).clone();
+            }
+        }
+
+        panic!("expected abandoned-parts trace state {state} for {bucket}/{object}");
+    }
+
+    fn trace_attr_string(event: &rustfs_common::trace_bus::TraceEvent, key: &str) -> Option<String> {
+        event.attrs.iter().find_map(|attr| {
+            if attr.key != key {
+                return None;
+            }
+            Some(match &attr.value {
+                rustfs_common::trace_bus::TraceVal::Bool(value) => value.to_string(),
+                rustfs_common::trace_bus::TraceVal::U64(value) => value.to_string(),
+                rustfs_common::trace_bus::TraceVal::I64(value) => value.to_string(),
+                rustfs_common::trace_bus::TraceVal::Str(value) => value.to_string(),
+            })
+        })
+    }
+
+    #[tokio::test]
+    async fn check_abandoned_parts_dry_run_counts_without_deleting() {
+        let mut trace = rustfs_common::trace_bus::subscribe_trace_events();
+        let (dir, disk) = make_single_local_disk().await;
+        let live = Uuid::new_v4();
+        let orphan = Uuid::new_v4();
+
+        let object_dir = dir.path().join("bucket").join("obj");
+        write_object_meta_with_data_dirs(&object_dir, "bucket", "obj", &[live]).await;
+        fs::create_dir_all(object_dir.join(live.to_string()))
+            .await
+            .expect("live data dir should be created");
+        fs::create_dir_all(object_dir.join(orphan.to_string()))
+            .await
+            .expect("orphan data dir should be created");
+
+        let set = make_set_disks_with(vec![Some(disk)]).await;
+        set.check_abandoned_parts(
+            "bucket",
+            "obj",
+            &HealOpts {
+                dry_run: true,
+                no_lock: true,
+                ..Default::default()
+            },
+        )
+        .await
+        .expect("dry-run abandoned-parts check should succeed");
+        let dry_run_trace = recv_abandoned_parts_trace(&mut trace, "bucket", "obj", "dry_run_matched").await;
+        assert_eq!(trace_attr_string(&dry_run_trace, "dry_run").as_deref(), Some("true"));
+        assert_eq!(trace_attr_string(&dry_run_trace, "data_dirs").as_deref(), Some("1"));
+
+        assert!(object_dir.join(live.to_string()).exists(), "referenced data dir must be preserved");
+        assert!(object_dir.join(orphan.to_string()).exists(), "dry-run must not remove orphaned data dir");
+
+        set.check_abandoned_parts(
+            "bucket",
+            "obj",
+            &HealOpts {
+                no_lock: true,
+                ..Default::default()
+            },
+        )
+        .await
+        .expect("abandoned-parts check should reclaim stale data dir");
+        let reclaim_trace = recv_abandoned_parts_trace(&mut trace, "bucket", "obj", "reclaimed").await;
+        assert_eq!(trace_attr_string(&reclaim_trace, "dry_run").as_deref(), Some("false"));
+        assert_eq!(trace_attr_string(&reclaim_trace, "data_dirs").as_deref(), Some("1"));
+
+        assert!(
+            object_dir.join(live.to_string()).exists(),
+            "referenced data dir must remain after reclaim"
+        );
+        assert!(!object_dir.join(orphan.to_string()).exists(), "orphaned data dir must be removed");
+    }
+
     #[tokio::test]
     async fn reclaim_orphan_data_dirs_recovers_deferred_cleanup_after_restart() {
         let (dir, disk) = make_single_local_disk().await;
@@ -12233,11 +12327,18 @@ mod tests {
             .expect_err("unsupported copy_object_part should return a typed error");
         assert!(matches!(copy_part_err, StorageError::NotImplemented));
 
-        let abandoned_err = set_disks
+        set_disks
-            .check_abandoned_parts("bucket", "object", &HealOpts::default())
+            .check_abandoned_parts(
+                "bucket",
+                "object",
+                &HealOpts {
+                    dry_run: true,
+                    no_lock: true,
+                    ..Default::default()
+                },
+            )
             .await
-            .expect_err("abandoned-parts check should stay in the upper reconciliation layer");
+            .expect("abandoned-parts check should be callable on empty disk sets");
-        assert!(matches!(abandoned_err, StorageError::NotImplemented));
     }
 
     #[tokio::test]

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

@@ -16,6 +16,7 @@ use super::super::*;
 use crate::disk::disk_store::DiskStoreRenameDataExt;
 use crate::io_support::bitrot::object_mmap_read_enabled;
 use crate::storage_api_contracts::namespace::NamespaceLocking as _;
+use rustfs_common::trace_bus::{TraceEvent, TraceFunc, TraceKind, trace_emit};
 use tracing::trace;
 
 const LOG_COMPONENT_ECSTORE: &str = "ecstore";
@@ -2057,11 +2058,61 @@ impl crate::storage_api_contracts::heal::HealOperations for SetDisks {
         Err(Error::DiskNotFound)
     }
 
-    #[tracing::instrument(skip(self))]
+    #[tracing::instrument(level = "debug", skip(self, opts), fields(bucket = %bucket, object = %object, dry_run = opts.dry_run))]
-    async fn check_abandoned_parts(&self, _bucket: &str, _object: &str, _opts: &HealOpts) -> Result<()> {
+    async fn check_abandoned_parts(&self, bucket: &str, object: &str, opts: &HealOpts) -> Result<()> {
-        // Multipart orphan reconciliation is intentionally retained above the set layer
+        let started_at = std::time::Instant::now();
-        // until there is a concrete caller and a stable lower-level contract to implement.
+        let _write_lock_guard = if !opts.no_lock {
-        Err(StorageError::NotImplemented)
+            let ns_lock = self.new_ns_lock(bucket, object).await?;
+            Some(
+                ns_lock
+                    .get_write_lock(get_lock_acquire_timeout())
+                    .await
+                    .map_err(|e| self.map_namespace_lock_error(bucket, object, "write", e))?,
+            )
+        } else {
+            None
+        };
+
+        let removed = if opts.dry_run {
+            self.dry_run_reclaim_orphan_data_dirs(bucket, object).await?
+        } else {
+            self.reclaim_orphan_data_dirs(bucket, object).await?
+        };
+        let state = if opts.dry_run && removed > 0 {
+            "dry_run_matched"
+        } else if removed > 0 {
+            "reclaimed"
+        } else {
+            "checked"
+        };
+        let data_dirs = u64::try_from(removed).unwrap_or(u64::MAX);
+
+        trace_emit(|| {
+            TraceEvent::new(TraceKind::Heal, TraceFunc::HealCheckAbandonedParts)
+                .with_bucket(bucket)
+                .with_object(object)
+                .with_duration(started_at.elapsed())
+                .with_attr("state", state)
+                .with_attr("dry_run", opts.dry_run)
+                .with_attr("data_dirs", data_dirs)
+        });
+
+        if removed > 0 {
+            trace!(
+                event = "heal_abandoned_parts",
+                component = LOG_COMPONENT_ECSTORE,
+                subsystem = LOG_SUBSYSTEM_HEAL,
+                state = if opts.dry_run { "dry_run_matched" } else { "reclaimed" },
+                result = "ok",
+                bucket,
+                object,
+                dry_run = opts.dry_run,
+                data_dirs = removed,
+                "Heal abandoned parts checked object data directories"
+            );
+        }
+
+        Ok(())
     }
 }
 

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

@@ -23,6 +23,7 @@
 //! per-version `SetDisks::heal_object`.
 
 use super::super::*;
+use crate::object_api::ObjectInfo;
 use std::collections::HashSet;
 use std::sync::Mutex;
 use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
@@ -39,12 +40,16 @@ const BACKGROUND_WALKDIR_STALL_TIMEOUT: Duration = Duration::from_secs(60);
 /// it must not gate healing logic — the delete-marker vs data path is chosen
 /// inside `ops/heal.rs` from the resolved latest metadata. `version_id` is
 /// normalized (nil/absent UUID => `None`).
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone)]
 pub struct HealWalkVersion {
     /// object key
     pub name: String,
     /// normalized version id (`None` when the version is nil/absent)
     pub version_id: Option<String>,
+    /// version modification time as Unix nanoseconds
+    pub mod_time_unix_nanos: Option<i128>,
+    /// object snapshot for lifecycle evaluation
+    pub lifecycle_object_info: Option<ObjectInfo>,
     /// whether this version is a delete marker (observability only)
     pub is_delete_marker: bool,
 }
@@ -63,6 +68,7 @@ struct HealWalkCollector {
     bucket: String,
     batch_objects: usize,
     version_budget: usize,
+    include_lifecycle_object_info: bool,
     objects: Mutex<Vec<HealWalkObject>>,
     decode_error: Mutex<Option<DiskError>>,
     version_total: AtomicUsize,
@@ -116,10 +122,25 @@ impl HealWalkCollector {
 
         let mut versions = Vec::with_capacity(fiv.versions.len() + fiv.free_versions.len());
         for fi in fiv.versions.iter().chain(fiv.free_versions.iter()) {
+            let version_uuid = fi.version_id.filter(|version_id| !version_id.is_nil());
+            let lifecycle_object_info = if self.include_lifecycle_object_info {
+                let mut lifecycle_fi = fi.clone();
+                lifecycle_fi.version_id = version_uuid;
+                Some(ObjectInfo::from_file_info(
+                    &lifecycle_fi,
+                    &self.bucket,
+                    &entry.name,
+                    version_uuid.is_some(),
+                ))
+            } else {
+                None
+            };
             versions.push(HealWalkVersion {
                 name: entry.name.clone(),
                 // Normalize: nil/absent version id => None.
-                version_id: fi.version_id.filter(|u| !u.is_nil()).map(|u| u.to_string()),
+                version_id: version_uuid.map(|u| u.to_string()),
+                mod_time_unix_nanos: fi.mod_time.map(|mod_time| mod_time.unix_timestamp_nanos()),
+                lifecycle_object_info,
                 is_delete_marker: fi.deleted,
             });
         }
@@ -173,11 +194,26 @@ impl HealWalkCollector {
                 }
             };
             for fi in fiv.versions.iter().chain(fiv.free_versions.iter()) {
-                let vid = fi.version_id.filter(|u| !u.is_nil()).map(|u| u.to_string());
+                let version_uuid = fi.version_id.filter(|version_id| !version_id.is_nil());
+                let vid = version_uuid.map(|u| u.to_string());
                 if seen.insert(vid.clone()) {
+                    let lifecycle_object_info = if self.include_lifecycle_object_info {
+                        let mut lifecycle_fi = fi.clone();
+                        lifecycle_fi.version_id = version_uuid;
+                        Some(ObjectInfo::from_file_info(
+                            &lifecycle_fi,
+                            &self.bucket,
+                            &entry.name,
+                            version_uuid.is_some(),
+                        ))
+                    } else {
+                        None
+                    };
                     versions.push(HealWalkVersion {
                         name: entry.name.clone(),
                         version_id: vid,
+                        mod_time_unix_nanos: fi.mod_time.map(|mod_time| mod_time.unix_timestamp_nanos()),
+                        lifecycle_object_info,
                         is_delete_marker: fi.deleted,
                     });
                 }
@@ -255,6 +291,7 @@ impl SetDisks {
         forward_to: Option<&str>,
         batch_objects: usize,
         version_budget: usize,
+        include_lifecycle_object_info: bool,
     ) -> disk::error::Result<(Vec<HealWalkVersion>, Option<String>, bool)> {
         assert!(batch_objects >= 2, "heal_walk_versions_page requires batch_objects >= 2");
 
@@ -264,6 +301,7 @@ impl SetDisks {
             bucket: bucket.to_string(),
             batch_objects,
             version_budget: version_budget.max(1),
+            include_lifecycle_object_info,
             objects: Mutex::new(Vec::new()),
             decode_error: Mutex::new(None),
             version_total: AtomicUsize::new(0),
@@ -347,6 +385,7 @@ mod tests {
             bucket: "bucket".to_string(),
             batch_objects: 2,
             version_budget: 2,
+            include_lifecycle_object_info: false,
             objects: Mutex::new(Vec::new()),
             decode_error: Mutex::new(None),
             version_total: AtomicUsize::new(0),
@@ -388,6 +427,8 @@ mod tests {
         HealWalkVersion {
             name: name.to_string(),
             version_id: Some(id.to_string()),
+            mod_time_unix_nanos: None,
+            lifecycle_object_info: None,
             is_delete_marker: dm,
         }
     }
@@ -491,6 +532,7 @@ mod tests {
             bucket: "bucket".to_string(),
             batch_objects: 1000,
             version_budget: 10_000,
+            include_lifecycle_object_info: false,
             objects: Mutex::new(Vec::new()),
             version_total: AtomicUsize::new(0),
             decode_error: Mutex::new(None),
@@ -567,7 +609,7 @@ mod tests {
         .expect("corrupt test metadata should be written");
 
         let error = set_disks
-            .heal_walk_versions_page(bucket, "", None, 2, 2)
+            .heal_walk_versions_page(bucket, "", None, 2, 2, false)
             .await
             .expect_err("semantic metadata corruption must fail the heal disk walk");
 

crates/ecstore/src/store/heal.rs

@@ -18,6 +18,7 @@ use tracing::trace;
 
 const LOG_COMPONENT_ECSTORE: &str = "ecstore";
 const LOG_SUBSYSTEM_HEAL: &str = "heal";
+const EVENT_HEAL_ABANDONED_PARTS: &str = "heal_abandoned_parts";
 const EVENT_HEAL_FORMAT_COMPLETED: &str = "heal_format_completed";
 const EVENT_HEAL_OBJECT_STARTED: &str = "heal_object_started";
 
@@ -256,13 +257,40 @@ impl ECStore {
 
     #[instrument(skip(self))]
     pub(super) async fn handle_check_abandoned_parts(&self, bucket: &str, object: &str, opts: &HealOpts) -> Result<()> {
-        let _ = (bucket, object, opts);
+        let object = encode_dir_object(object);
-        // Stale multipart reconciliation is already owned by the lifecycle-driven
+        let pools = self.get_pools_for_heal_object(opts)?;
-        // background cleanup path in `bucket_lifecycle_ops.rs`. There is currently
+
-        // no stable object-heal contract that should fan this request out through
+        let mut futures = Vec::with_capacity(pools.len());
-        // pool/set storage layers, so keep the placeholder explicit at the ECStore
+        for pool in pools.iter() {
-        // boundary instead of dispatching into lower layers.
+            futures.push(pool.check_abandoned_parts(bucket, &object, opts));
-        Err(StorageError::NotImplemented)
+        }
+
+        let mut first_error = None;
+        for result in join_all(futures).await {
+            if let Err(err) = result
+                && first_error.is_none()
+            {
+                first_error = Some(err);
+            }
+        }
+
+        if let Some(err) = first_error {
+            return Err(err);
+        }
+
+        trace!(
+            event = EVENT_HEAL_ABANDONED_PARTS,
+            component = LOG_COMPONENT_ECSTORE,
+            subsystem = LOG_SUBSYSTEM_HEAL,
+            state = "completed",
+            result = "ok",
+            bucket,
+            object,
+            dry_run = opts.dry_run,
+            "Heal abandoned parts completed"
+        );
+
+        Ok(())
     }
 }
 

crates/ecstore/src/store/heal_walk.rs

@@ -34,6 +34,7 @@ impl ECStore {
         forward_to: Option<&str>,
         batch_objects: usize,
         version_budget: usize,
+        include_lifecycle_object_info: bool,
     ) -> Result<(Vec<HealWalkVersion>, Option<String>, bool)> {
         if pool_idx >= self.pools.len() || set_idx >= self.pools[pool_idx].disk_set.len() {
             return Err(Error::other(format!(
@@ -43,7 +44,7 @@ impl ECStore {
         }
 
         self.pools[pool_idx].disk_set[set_idx]
-            .heal_walk_versions_page(bucket, prefix, forward_to, batch_objects, version_budget)
+            .heal_walk_versions_page(bucket, prefix, forward_to, batch_objects, version_budget, include_lifecycle_object_info)
             .await
             .map_err(Error::from)
     }

crates/heal/src/heal/channel.rs

@@ -767,6 +767,7 @@ mod tests {
             _bucket: &str,
             _prefix: &str,
             _continuation_token: Option<&str>,
+            _include_lifecycle_object_info: bool,
         ) -> crate::Result<(Vec<crate::heal::storage::HealListItem>, Option<String>, bool)> {
             Ok((vec![], None, false))
         }

crates/heal/src/heal/erasure_healer.rs

@@ -23,13 +23,14 @@ use crate::heal::{
 };
 use crate::{Error, Result};
 use futures::{StreamExt, stream::FuturesUnordered};
-use metrics::gauge;
+use metrics::{counter, gauge};
 use rustfs_common::heal_channel::{HealOpts, HealRequestSource, HealScanMode};
 use rustfs_madmin::heal_commands::HealResultItem;
 use std::sync::{
     Arc,
     atomic::{AtomicUsize, Ordering},
 };
+use std::time::{Duration, UNIX_EPOCH};
 use tokio::sync::{RwLock, Semaphore};
 use tracing::{debug, error, warn};
 
@@ -47,6 +48,21 @@ enum HealObjectOutcome {
     Failed,
 }
 
+fn result_object_size_u64(result: &HealResultItem) -> u64 {
+    u64::try_from(result.object_size).unwrap_or(u64::MAX)
+}
+
+const NEW_VERSION_SKIP_GRACE_SECS: u64 = 60;
+const NANOS_PER_SECOND: i128 = 1_000_000_000;
+
+fn should_skip_new_version(mod_time_unix_nanos: Option<i128>, started_at_secs: u64) -> bool {
+    let Some(mod_time_unix_nanos) = mod_time_unix_nanos else {
+        return false;
+    };
+    let cutoff_secs = started_at_secs.saturating_add(NEW_VERSION_SKIP_GRACE_SECS);
+    mod_time_unix_nanos > i128::from(cutoff_secs).saturating_mul(NANOS_PER_SECOND)
+}
+
 struct PageConcurrencyGuard {
     in_flight: Arc<AtomicUsize>,
     set_label: String,
@@ -492,6 +508,7 @@ impl ErasureSetHealer {
                     &mut skipped_objects,
                     resume_manager,
                     checkpoint_manager,
+                    state.start_time,
                 )
                 .await;
 
@@ -658,6 +675,7 @@ impl ErasureSetHealer {
         skipped_objects: &mut u64,
         resume_manager: &ResumeManager,
         checkpoint_manager: &CheckpointManager,
+        started_at_secs: u64,
     ) -> Result<()> {
         debug!(
             target: "rustfs::heal::erasure_healer",
@@ -710,6 +728,7 @@ impl ErasureSetHealer {
         // The end-of-pass summary reports the full failed/skipped counts.
         let mut transient_skip_samples_logged = 0_u64;
         let mut failure_samples_logged = 0_u64;
+        let mut bytes_processed = self.progress.read().await.bytes_processed;
 
         // backlog#920: select the per-erasure-set DISK-WALK union enumerator when
         // the scan is Deep OR the request came from AutoHeal — these are the paths
@@ -718,17 +737,25 @@ impl ErasureSetHealer {
         // which stays the default.
         let use_disk_walk =
             matches!(self.heal_opts.scan_mode, HealScanMode::Deep) || matches!(self.source, HealRequestSource::AutoHeal);
+        let lifecycle_expiry_context = self.storage.load_heal_lifecycle_expiry_context(bucket).await?;
+        let include_lifecycle_object_info = lifecycle_expiry_context.is_some();
 
         loop {
             self.verify_replacement_identity_fence("page scan").await?;
             // Get one page of object versions
             let (objects, next_token, is_truncated) = if use_disk_walk {
                 self.storage
-                    .list_versions_for_heal_page_disk_walk(set_disk_id, bucket, "", continuation_token.as_deref())
+                    .list_versions_for_heal_page_disk_walk(
+                        set_disk_id,
+                        bucket,
+                        "",
+                        continuation_token.as_deref(),
+                        include_lifecycle_object_info,
+                    )
                     .await?
             } else {
                 self.storage
-                    .list_objects_for_heal_page(bucket, "", continuation_token.as_deref())
+                    .list_objects_for_heal_page(bucket, "", continuation_token.as_deref(), include_lifecycle_object_info)
                     .await?
             };
             let page_is_empty = objects.is_empty();
@@ -736,6 +763,7 @@ impl ErasureSetHealer {
             let page_resume_index = *current_object_index;
             let semaphore = Arc::new(Semaphore::new(page_concurrency_limit));
             let mut page_tasks = FuturesUnordered::new();
+            let mut completed_in_page = 0usize;
 
             // Capture the last version identity of this page for the anti-loop guard.
             let page_last = objects.last().map(|item| (item.name.clone(), item.version_id.clone()));
@@ -751,6 +779,75 @@ impl ErasureSetHealer {
                     continue;
                 }
 
+                if should_skip_new_version(item.mod_time_unix_nanos, started_at_secs) {
+                    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 mut progress = self.progress.write().await;
+                        progress.record_skipped_new_version();
+                        progress.set_current_object(Some(format!("skipped_new: {bucket}/{}", item.name)));
+                        progress.update_progress(*processed_objects, *successful_objects, *failed_objects, bytes_processed);
+                    }
+                    debug!(
+                        target: "rustfs::heal::erasure_healer",
+                        event = EVENT_HEAL_ERASURE_OBJECT_STATE,
+                        component = LOG_COMPONENT_HEAL,
+                        subsystem = LOG_SUBSYSTEM_ERASURE_HEALER,
+                        set_disk_id,
+                        bucket,
+                        object = %item.name,
+                        version_id = ?item.version_id,
+                        state = "skipped_new_version",
+                        "Erasure set object version skipped because it was written after heal started"
+                    );
+                    if completed_in_page.is_multiple_of(100) {
+                        checkpoint_manager.update_position(bucket_index, page_resume_index).await?;
+                    }
+                    continue;
+                }
+
+                if let Some(context) = lifecycle_expiry_context.as_ref()
+                    && self
+                        .storage
+                        .enqueue_heal_lifecycle_expiry(
+                            context,
+                            bucket,
+                            &item.name,
+                            item.version_id.as_deref(),
+                            item.lifecycle_object_info.as_ref(),
+                        )
+                        .await?
+                {
+                    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 mut progress = self.progress.write().await;
+                        progress.record_skipped_ilm_expired();
+                        progress.set_current_object(Some(format!("skipped_ilm: {bucket}/{}", item.name)));
+                        progress.update_progress(*processed_objects, *successful_objects, *failed_objects, bytes_processed);
+                    }
+                    debug!(
+                        target: "rustfs::heal::erasure_healer",
+                        event = EVENT_HEAL_ERASURE_OBJECT_STATE,
+                        component = LOG_COMPONENT_HEAL,
+                        subsystem = LOG_SUBSYSTEM_ERASURE_HEALER,
+                        set_disk_id,
+                        bucket,
+                        object = %item.name,
+                        version_id = ?item.version_id,
+                        state = "skipped_ilm_expired",
+                        "Erasure set object version skipped because lifecycle expiry was queued"
+                    );
+                    if completed_in_page.is_multiple_of(100) {
+                        checkpoint_manager.update_position(bucket_index, page_resume_index).await?;
+                    }
+                    continue;
+                }
+
                 resume_manager
                     .set_current_item(Some(bucket.to_string()), Some(item.name.clone()))
                     .await?;
@@ -777,7 +874,7 @@ impl ErasureSetHealer {
 
                     let _permit = match permit {
                         Ok(permit) => permit,
-                        Err(err) => return (dedup_key, object_name, version_id, Err(err)),
+                        Err(err) => return (dedup_key, object_name, version_id, (0, Err(err))),
                     };
 
                     let _in_flight_guard = PageConcurrencyGuard::new(in_flight, set_label);
@@ -788,7 +885,7 @@ impl ErasureSetHealer {
                     // recorded as skipped-ok rather than failed. The delete-marker
                     // vs data path is chosen internally in ops/heal.rs.
                     let result = if cancel_token.is_cancelled() {
-                        Err(Error::TaskCancelled)
+                        (0, Err(Error::TaskCancelled))
                     } else {
                         match storage
                             .heal_object(&bucket_name, &object_name, version_id.as_deref(), &heal_opts)
@@ -797,8 +894,9 @@ impl ErasureSetHealer {
                             Ok((result, None))
                                 if target_outcomes_complete(&result, &target_endpoints) =>
                             {
+                                let object_size = result_object_size_u64(&result);
                                 if !replacement_commit_evidence_required {
-                                    Ok(true)
+                                    (object_size, Ok(true))
                                 } else {
                                     match storage
                                         .replacement_targets_have_version(
@@ -810,27 +908,42 @@ impl ErasureSetHealer {
                                         )
                                         .await
                                     {
-                                        Ok(true) => Ok(true),
+                                        Ok(true) => (object_size, Ok(true)),
-                                        Ok(false) => Err(Error::transient_skip(format!(
+                                        Ok(false) => (object_size, Err(Error::transient_skip(format!(
                                             "Skipped heal for {bucket_name}/{object_name} because replacement target readback did not confirm the committed version"
-                                        ))),
+                                        )))),
-                                        Err(err) => Err(Error::transient_skip(format!(
+                                        Err(err) => (object_size, Err(Error::transient_skip(format!(
                                             "Skipped heal for {bucket_name}/{object_name} because replacement target readback failed: {err}"
-                                        ))),
+                                        )))),
                                     }
                                 }
-                            }
+                            },
-                            Ok((_result, None)) if !target_endpoints.is_empty() => Err(Error::transient_skip(format!(
+                            Ok((result, None)) if !target_endpoints.is_empty() => (
-                                "Skipped heal for {bucket_name}/{object_name} because a replacement target was not committed"
+                                result_object_size_u64(&result),
-                            ))),
+                                Err(Error::transient_skip(format!(
-                            Ok((_result, None)) => Ok(true),
+                                    "Skipped heal for {bucket_name}/{object_name} because a replacement target was not committed"
-                            Ok((_, Some(err))) if is_missing_object_dir_heal_result(&object_name, &err) => Ok(false),
-                            Ok((_, Some(err))) | Err(err) => match Self::classify_heal_object_error(&err) {
-                                HealObjectOutcome::Absent => Ok(false),
-                                HealObjectOutcome::Transient => Err(Error::transient_skip(format!(
-                                    "Skipped heal for {bucket_name}/{object_name} due to transient error: {err}"
                                 ))),
-                                HealObjectOutcome::Failed => Err(err),
+                            ),
+                            Ok((result, None)) => (result_object_size_u64(&result), Ok(true)),
+                            Ok((result, Some(err))) if is_missing_object_dir_heal_result(&object_name, &err) => {
+                                (result_object_size_u64(&result), Ok(false))
+                            }
+                            Ok((result, Some(err))) => {
+                                let object_size = result_object_size_u64(&result);
+                                match Self::classify_heal_object_error(&err) {
+                                    HealObjectOutcome::Absent => (object_size, Ok(false)),
+                                    HealObjectOutcome::Transient => (object_size, Err(Error::transient_skip(format!(
+                                        "Skipped heal for {bucket_name}/{object_name} due to transient error: {err}"
+                                    )))),
+                                    HealObjectOutcome::Failed => (object_size, Err(err)),
+                                }
+                            }
+                            Err(err) => match Self::classify_heal_object_error(&err) {
+                                HealObjectOutcome::Absent => (0, Ok(false)),
+                                HealObjectOutcome::Transient => (0, Err(Error::transient_skip(format!(
+                                    "Skipped heal for {bucket_name}/{object_name} due to transient error: {err}"
+                                )))),
+                                HealObjectOutcome::Failed => (0, Err(err)),
                             },
                         }
                     };
@@ -839,11 +952,12 @@ impl ErasureSetHealer {
                 });
             }
 
-            let mut completed_in_page = 0usize;
             while let Some((key, object, version_id, result)) = page_tasks.next().await {
+                let (object_size, result) = result;
                 match result {
                     Ok(true) => {
                         *successful_objects += 1;
+                        bytes_processed = bytes_processed.saturating_add(object_size);
                         checkpoint_manager.add_processed_object(key).await?;
                         debug!(
                             target: "rustfs::heal::erasure_healer",
@@ -861,6 +975,7 @@ impl ErasureSetHealer {
                     Ok(false) => {
                         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,
@@ -877,6 +992,7 @@ impl ErasureSetHealer {
                     Err(err @ Error::TaskCancelled) | Err(err @ Error::TaskTimeout) => return Err(err),
                     Err(Error::TransientSkip { message }) => {
                         *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,
@@ -893,6 +1009,7 @@ impl ErasureSetHealer {
                     }
                     Err(err) => {
                         *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,
@@ -911,6 +1028,11 @@ impl ErasureSetHealer {
 
                 *processed_objects += 1;
                 completed_in_page += 1;
+                {
+                    let mut progress = self.progress.write().await;
+                    progress.set_current_object(Some(format!("{bucket}/{object}")));
+                    progress.update_progress(*processed_objects, *successful_objects, *failed_objects, bytes_processed);
+                }
 
                 if completed_in_page.is_multiple_of(100) {
                     checkpoint_manager.update_position(bucket_index, page_resume_index).await?;
@@ -964,7 +1086,9 @@ impl ErasureSetHealer {
         progress.objects_scanned = state.total_objects;
         progress.objects_healed = state.successful_objects;
         progress.objects_failed = state.failed_objects;
-        progress.bytes_processed = 0; // set to 0 for now, can be extended later
+        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());
     }
 }
@@ -1135,13 +1259,15 @@ mod resume_loop_tests {
     //! that emits programmable multi-version pages. These exercise the real loop
     //! logic (cursor seeding, per-version dedup, anti-loop guard, absence
     //! handling) — not merely a mock's own output.
-    use super::{ErasureSetHealer, target_outcomes_complete};
+    use super::{
+        ErasureSetHealer, NANOS_PER_SECOND, NEW_VERSION_SKIP_GRACE_SECS, should_skip_new_version, target_outcomes_complete,
+    };
     use crate::heal::progress::HealProgress;
     use crate::heal::resume::{
         CheckpointManager, RESUME_CHECKPOINT_FILE, ReplacementTargetIdentity, ResumeDeleteFailure, ResumeManager, ResumeUtils,
         compose_key,
     };
-    use crate::heal::storage::{DiskStatus, HealListItem, HealObjectInfo, HealStorageAPI};
+    use crate::heal::storage::{DiskStatus, HealLifecycleExpiryContext, HealListItem, HealObjectInfo, HealStorageAPI};
     use crate::heal::storage_api::status::BucketInfo;
     use crate::heal::{
         BUCKET_META_PREFIX, DiskOption, DiskStore, EcstoreError, Endpoint, HealDiskExt as _, RUSTFS_META_BUCKET, new_disk,
@@ -1149,7 +1275,7 @@ mod resume_loop_tests {
     use crate::{Error, Result};
     use rustfs_common::heal_channel::{HealOpts, HealRequestSource};
     use rustfs_madmin::heal_commands::{HealDriveInfo, HealResultItem, Infos};
-    use std::collections::{HashMap, VecDeque};
+    use std::collections::{HashMap, HashSet, VecDeque};
     use std::sync::atomic::{AtomicBool, Ordering};
     use std::sync::{Arc, Mutex};
     use tempfile::TempDir;
@@ -1160,10 +1286,37 @@ mod resume_loop_tests {
         HealListItem {
             name: name.to_string(),
             version_id: version.map(str::to_string),
+            mod_time_unix_nanos: None,
+            lifecycle_object_info: None,
             is_delete_marker: delete_marker,
         }
     }
 
+    fn item_with_mod_time(name: &str, version: Option<&str>, mod_time_secs: u64) -> HealListItem {
+        HealListItem {
+            name: name.to_string(),
+            version_id: version.map(str::to_string),
+            mod_time_unix_nanos: Some(i128::from(mod_time_secs).saturating_mul(NANOS_PER_SECOND)),
+            lifecycle_object_info: None,
+            is_delete_marker: false,
+        }
+    }
+
+    #[test]
+    fn new_version_filter_respects_grace_boundary() {
+        let started_at = 1_700_000_000;
+
+        assert!(!should_skip_new_version(None, started_at));
+        assert!(!should_skip_new_version(
+            Some(i128::from(started_at + NEW_VERSION_SKIP_GRACE_SECS).saturating_mul(NANOS_PER_SECOND)),
+            started_at,
+        ));
+        assert!(should_skip_new_version(
+            Some(i128::from(started_at + NEW_VERSION_SKIP_GRACE_SECS + 1).saturating_mul(NANOS_PER_SECOND)),
+            started_at,
+        ));
+    }
+
     #[test]
     fn target_outcomes_require_each_requested_endpoint_once_and_ok() {
         let result = HealResultItem {
@@ -1246,8 +1399,10 @@ mod resume_loop_tests {
         /// Target-specific physical readback evidence per `compose_key`; the
         /// fake models a healthy backend unless a test explicitly revokes it.
         replacement_commit_evidence: Mutex<HashMap<String, ReplacementCommitEvidence>>,
+        lifecycle_expired: Mutex<HashSet<String>>,
         /// every heal_object call recorded as (name, version_id)
         heal_calls: Mutex<Vec<(String, Option<String>)>>,
+        list_include_lifecycle_object_info: Mutex<Vec<bool>>,
         replacement_target_identity_sequences: Mutex<VecDeque<Vec<ReplacementTargetIdentity>>>,
         fail_listing: AtomicBool,
     }
@@ -1274,9 +1429,15 @@ mod resume_loop_tests {
                 .unwrap()
                 .insert(compose_key(name, version), ReplacementCommitEvidence::Error(message.to_string()));
         }
+        fn set_lifecycle_expired(&self, name: &str, version: Option<&str>) {
+            self.lifecycle_expired.lock().unwrap().insert(compose_key(name, version));
+        }
         fn calls(&self) -> Vec<(String, Option<String>)> {
             self.heal_calls.lock().unwrap().clone()
         }
+        fn list_include_lifecycle_object_info_calls(&self) -> Vec<bool> {
+            self.list_include_lifecycle_object_info.lock().unwrap().clone()
+        }
         fn fail_listing(&self) {
             self.fail_listing.store(true, Ordering::SeqCst);
         }
@@ -1330,6 +1491,23 @@ mod resume_loop_tests {
         async fn get_object_checksum(&self, _b: &str, _o: &str) -> Result<Option<String>> {
             Ok(None)
         }
+        async fn load_heal_lifecycle_expiry_context(&self, _bucket: &str) -> Result<Option<HealLifecycleExpiryContext>> {
+            Ok((!self.lifecycle_expired.lock().unwrap().is_empty()).then(HealLifecycleExpiryContext::test))
+        }
+        async fn enqueue_heal_lifecycle_expiry(
+            &self,
+            _context: &HealLifecycleExpiryContext,
+            _bucket: &str,
+            object: &str,
+            version_id: Option<&str>,
+            _object_info: Option<&HealObjectInfo>,
+        ) -> Result<bool> {
+            Ok(self
+                .lifecycle_expired
+                .lock()
+                .unwrap()
+                .contains(&compose_key(object, version_id)))
+        }
         async fn heal_object(
             &self,
             _bucket: &str,
@@ -1386,7 +1564,12 @@ mod resume_loop_tests {
             _bucket: &str,
             _prefix: &str,
             continuation_token: Option<&str>,
+            include_lifecycle_object_info: bool,
         ) -> Result<(Vec<HealListItem>, Option<String>, bool)> {
+            self.list_include_lifecycle_object_info
+                .lock()
+                .unwrap()
+                .push(include_lifecycle_object_info);
             if self.fail_listing.load(Ordering::SeqCst) {
                 return Err(Error::other("injected listing failure"));
             }
@@ -1476,6 +1659,7 @@ mod resume_loop_tests {
 
     /// Drive one bucket heal pass; returns (processed, successful, failed, skipped, result).
     async fn run(env: &Env) -> (u64, u64, u64, u64, Result<()>) {
+        let state = env.resume.get_state().await;
         let mut current_object_index = 0usize;
         let mut processed = 0u64;
         let mut successful = 0u64;
@@ -1494,6 +1678,7 @@ mod resume_loop_tests {
                 &mut skipped,
                 &env.resume,
                 &env.checkpoint,
+                state.start_time,
             )
             .await;
         (processed, successful, failed, skipped, result)
@@ -1559,6 +1744,7 @@ mod resume_loop_tests {
         let mut successful = 0;
         let mut failed = 0;
         let mut skipped = 0;
+        let started_at = env.resume.get_state().await.start_time;
 
         let error = healer
             .heal_bucket_with_resume(
@@ -1572,6 +1758,7 @@ mod resume_loop_tests {
                 &mut skipped,
                 &env.resume,
                 &env.checkpoint,
+                started_at,
             )
             .await
             .expect_err("a remounted target must not begin a new page scan");
@@ -1641,6 +1828,109 @@ mod resume_loop_tests {
         assert_eq!(skipped, 0);
     }
 
+    #[tokio::test]
+    async fn erasure_set_progress_accumulates_healed_object_bytes() {
+        let env = make_env().await;
+        env.storage.set_page(
+            None,
+            Page {
+                items: vec![item("first", Some("v1"), false), item("second", Some("v2"), false)],
+                next: None,
+                truncated: false,
+            },
+        );
+        env.storage.set_result(
+            "first",
+            Some("v1"),
+            HealResultItem {
+                object_size: 1024,
+                ..Default::default()
+            },
+        );
+        env.storage.set_result(
+            "second",
+            Some("v2"),
+            HealResultItem {
+                object_size: 2048,
+                ..Default::default()
+            },
+        );
+
+        let (processed, successful, failed, skipped, result) = run(&env).await;
+
+        result.expect("page heal should succeed");
+        assert_eq!(processed, 2);
+        assert_eq!(successful, 2);
+        assert_eq!(failed, 0);
+        assert_eq!(skipped, 0);
+        let progress = env.healer.progress.read().await;
+        assert_eq!(progress.objects_scanned, 2);
+        assert_eq!(progress.objects_healed, 2);
+        assert_eq!(progress.objects_failed, 0);
+        assert_eq!(progress.bytes_processed, 3072);
+        assert!(matches!(progress.current_object.as_deref(), Some("b/first" | "b/second")));
+    }
+
+    #[tokio::test]
+    async fn erasure_set_skips_versions_written_after_heal_started() {
+        let env = make_env().await;
+        let started_at = env.resume.get_state().await.start_time;
+        env.storage.set_page(
+            None,
+            Page {
+                items: vec![
+                    item_with_mod_time("old", Some("v1"), started_at + NEW_VERSION_SKIP_GRACE_SECS),
+                    item_with_mod_time("new", Some("v2"), started_at + NEW_VERSION_SKIP_GRACE_SECS + 1),
+                ],
+                next: None,
+                truncated: false,
+            },
+        );
+
+        let (processed, successful, failed, skipped, result) = run(&env).await;
+
+        result.expect("page heal should succeed");
+        assert_eq!(processed, 2);
+        assert_eq!(successful, 1);
+        assert_eq!(failed, 0);
+        assert_eq!(skipped, 0);
+        assert_eq!(env.storage.calls(), vec![("old".to_string(), Some("v1".to_string()))]);
+        let progress = env.healer.progress.read().await;
+        assert_eq!(progress.skipped_new_versions, 1);
+        assert_eq!(progress.objects_scanned, 2);
+        assert_eq!(progress.objects_healed, 1);
+        assert_eq!(progress.objects_failed, 0);
+    }
+
+    #[tokio::test]
+    async fn erasure_set_skips_versions_queued_for_lifecycle_expiry() {
+        let env = make_env().await;
+        env.storage.set_page(
+            None,
+            Page {
+                items: vec![item("expired", Some("v1"), false), item("kept", Some("v2"), false)],
+                next: None,
+                truncated: false,
+            },
+        );
+        env.storage.set_lifecycle_expired("expired", Some("v1"));
+
+        let (processed, successful, failed, skipped, result) = run(&env).await;
+
+        result.expect("page heal should succeed");
+        assert_eq!(processed, 2);
+        assert_eq!(successful, 1);
+        assert_eq!(failed, 0);
+        assert_eq!(skipped, 0);
+        assert_eq!(env.storage.calls(), vec![("kept".to_string(), Some("v2".to_string()))]);
+        assert_eq!(env.storage.list_include_lifecycle_object_info_calls(), vec![true]);
+        let progress = env.healer.progress.read().await;
+        assert_eq!(progress.skipped_ilm_expired, 1);
+        assert_eq!(progress.objects_scanned, 2);
+        assert_eq!(progress.objects_healed, 1);
+        assert_eq!(progress.objects_failed, 0);
+    }
+
     #[tokio::test]
     async fn bucket_listing_failure_does_not_mark_set_completed() {
         let env = make_env().await;

crates/heal/src/heal/manager.rs

@@ -2385,8 +2385,27 @@ impl HealManager {
             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;
+            }
         }
+        snapshot.refresh_progress_percentage();
+        snapshot.refresh_estimated_completion_time();
         Some(snapshot)
     }
 
@@ -3208,6 +3227,7 @@ impl HealManager {
                         } else {
                             completed_task.get_status().await
                         };
+                        let completed_progress = completed_task.get_progress().await;
                         let completed_status_entry = CompletedHealStatus {
                             heal_type: completed_task.heal_type.clone(),
                             status: completed_status.clone(),
@@ -3223,6 +3243,7 @@ impl HealManager {
                         match completed_status {
                             HealTaskStatus::Completed => {
                                 stats.update_task_completion(true);
+                                stats.add_healed_objects(completed_progress.objects_healed, completed_progress.bytes_processed);
                             }
                             HealTaskStatus::Retrying { .. } => {}
                             _ => {
@@ -3749,6 +3770,7 @@ mod tests {
             _bucket: &str,
             _prefix: &str,
             _continuation_token: Option<&str>,
+            _include_lifecycle_object_info: bool,
         ) -> Result<(Vec<crate::heal::storage::HealListItem>, Option<String>, bool)> {
             Ok((Vec::new(), None, false))
         }
@@ -5396,6 +5418,8 @@ mod tests {
         ));
         {
             let mut progress = first.progress.write().await;
+            progress.start_time = Some(SystemTime::now() - Duration::from_secs(20));
+            progress.set_total_baseline(12, 8192);
             progress.update_progress(7, 3, 1, 4096);
         }
 
@@ -5405,6 +5429,8 @@ mod tests {
         ));
         {
             let mut progress = second.progress.write().await;
+            progress.start_time = Some(SystemTime::now() - Duration::from_secs(10));
+            progress.set_total_baseline(8, 4096);
             progress.update_progress(11, 5, 2, 2048);
         }
 
@@ -5419,7 +5445,11 @@ mod tests {
         assert_eq!(progress.objects_scanned, 18);
         assert_eq!(progress.objects_healed, 8);
         assert_eq!(progress.objects_failed, 3);
+        assert_eq!(progress.objects_total_count, 20);
+        assert_eq!(progress.objects_total_size, 12288);
         assert_eq!(progress.bytes_processed, 6144);
+        assert!((progress.progress_percentage - 50.0).abs() < 0.001);
+        assert!(progress.estimated_completion_time.is_some());
     }
 
     #[tokio::test]

crates/heal/src/heal/progress.rs

@@ -13,7 +13,7 @@
 // limitations under the License.
 
 use serde::{Deserialize, Serialize};
-use std::time::SystemTime;
+use std::time::{Duration, SystemTime};
 
 #[derive(Debug, Default, Clone, Serialize, Deserialize)]
 #[serde(rename_all = "camelCase")]
@@ -24,6 +24,14 @@ pub struct HealProgress {
     pub objects_healed: u64,
     /// Objects failed
     pub objects_failed: 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
+    pub skipped_ilm_expired: u64,
+    /// Baseline object count from the latest complete usage snapshot
+    pub objects_total_count: u64,
+    /// Baseline object bytes from the latest complete usage snapshot
+    pub objects_total_size: u64,
     /// Bytes processed
     pub bytes_processed: u64,
     /// Current object
@@ -54,10 +62,56 @@ impl HealProgress {
         self.bytes_processed = bytes;
         self.last_update_time = Some(SystemTime::now());
 
-        // calculate progress percentage
+        self.refresh_progress_percentage();
-        let total = scanned + healed + failed;
+        self.refresh_estimated_completion_time();
+    }
+
+    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.last_update_time = Some(SystemTime::now());
+        self.refresh_progress_percentage();
+        self.refresh_estimated_completion_time();
+    }
+
+    pub fn record_skipped_new_version(&mut self) {
+        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) {
+        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) -> u64 {
+        self.objects_healed
+            .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.objects_total_size > 0 {
+            self.progress_percentage = ((self.bytes_processed as f64 / self.objects_total_size as f64) * 100.0).min(100.0);
+            return;
+        }
+        if self.objects_total_count > 0 {
+            let completed = self.completed_for_baseline();
+            self.progress_percentage = ((completed as f64 / self.objects_total_count as f64) * 100.0).min(100.0);
+            return;
+        }
+
+        let total = self
+            .objects_scanned
+            .saturating_add(self.objects_healed)
+            .saturating_add(self.objects_failed);
         if total > 0 {
-            self.progress_percentage = (healed as f64 / total as f64) * 100.0;
+            self.progress_percentage = (self.objects_healed as f64 / total as f64) * 100.0;
         }
     }
 
@@ -66,9 +120,36 @@ impl HealProgress {
         self.last_update_time = Some(SystemTime::now());
     }
 
+    pub fn refresh_estimated_completion_time(&mut self) {
+        let Some(start_time) = self.start_time else {
+            self.estimated_completion_time = None;
+            return;
+        };
+        if self.is_completed() || !(0.0..100.0).contains(&self.progress_percentage) || self.bytes_processed == 0 {
+            self.estimated_completion_time = None;
+            return;
+        }
+
+        let elapsed = match SystemTime::now().duration_since(start_time) {
+            Ok(elapsed) if !elapsed.is_zero() => elapsed,
+            _ => {
+                self.estimated_completion_time = None;
+                return;
+            }
+        };
+        let estimated_total_secs = elapsed.as_secs_f64() * 100.0 / self.progress_percentage;
+        self.estimated_completion_time = start_time.checked_add(Duration::from_secs_f64(estimated_total_secs));
+    }
+
     pub fn is_completed(&self) -> bool {
-        self.progress_percentage >= 100.0
+        if self.progress_percentage >= 100.0 {
-            || self.objects_scanned > 0 && self.objects_healed + self.objects_failed >= self.objects_scanned
+            return true;
+        }
+        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 {
@@ -158,6 +239,10 @@ mod tests {
         assert_eq!(progress.objects_scanned, 0);
         assert_eq!(progress.objects_healed, 0);
         assert_eq!(progress.objects_failed, 0);
+        assert_eq!(progress.skipped_new_versions, 0);
+        assert_eq!(progress.skipped_ilm_expired, 0);
+        assert_eq!(progress.objects_total_count, 0);
+        assert_eq!(progress.objects_total_size, 0);
         assert_eq!(progress.bytes_processed, 0);
         assert_eq!(progress.progress_percentage, 0.0);
         assert!(progress.start_time.is_some());
@@ -181,6 +266,73 @@ mod tests {
         assert!(progress.last_update_time.is_some());
     }
 
+    #[test]
+    fn test_heal_progress_estimates_completion_time_from_progress() {
+        let mut progress = HealProgress::new();
+        progress.start_time = Some(SystemTime::now() - Duration::from_secs(10));
+
+        progress.update_progress(100, 25, 0, 4096);
+
+        let eta = progress
+            .estimated_completion_time
+            .expect("partial byte progress should estimate completion");
+        assert!(eta > SystemTime::now());
+    }
+
+    #[test]
+    fn test_heal_progress_uses_byte_baseline_for_percentage() {
+        let mut progress = HealProgress::new();
+        progress.set_total_baseline(10, 8192);
+
+        progress.update_progress(100, 25, 0, 4096);
+
+        assert!((progress.progress_percentage - 50.0).abs() < 0.001);
+    }
+
+    #[test]
+    fn test_heal_progress_uses_object_baseline_when_bytes_unknown() {
+        let mut progress = HealProgress::new();
+        progress.set_total_baseline(10, 0);
+
+        progress.update_progress(100, 3, 2, 0);
+
+        assert!((progress.progress_percentage - 50.0).abs() < 0.001);
+    }
+
+    #[test]
+    fn test_heal_progress_counts_skipped_versions_for_object_baseline() {
+        let mut progress = HealProgress::new();
+        progress.set_total_baseline(10, 0);
+
+        progress.update_progress(100, 3, 2, 0);
+        progress.record_skipped_new_version();
+
+        assert_eq!(progress.skipped_new_versions, 1);
+        assert!((progress.progress_percentage - 60.0).abs() < 0.001);
+    }
+
+    #[test]
+    fn test_heal_progress_does_not_estimate_completion_without_bytes() {
+        let mut progress = HealProgress::new();
+        progress.start_time = Some(SystemTime::now() - Duration::from_secs(10));
+
+        progress.update_progress(100, 25, 0, 0);
+
+        assert!(progress.estimated_completion_time.is_none());
+    }
+
+    #[test]
+    fn test_heal_progress_with_baseline_is_not_completed_by_processed_count() {
+        let mut progress = HealProgress::new();
+        progress.start_time = Some(SystemTime::now() - Duration::from_secs(10));
+        progress.set_total_baseline(10, 8192);
+
+        progress.update_progress(1, 1, 0, 1024);
+
+        assert!(!progress.is_completed());
+        assert!(progress.estimated_completion_time.is_some());
+    }
+
     #[test]
     fn test_heal_progress_update_progress_zero_total() {
         let mut progress = HealProgress::new();
@@ -251,6 +403,8 @@ mod tests {
         assert_eq!(json["objectsScanned"], 10);
         assert_eq!(json["objectsHealed"], 8);
         assert_eq!(json["objectsFailed"], 2);
+        assert_eq!(json["skippedNewVersions"], 0);
+        assert_eq!(json["skippedIlmExpired"], 0);
         assert_eq!(json["bytesProcessed"], 1024);
         assert_eq!(json["currentObject"], "test-bucket/test-object");
         assert!(json["progressPercentage"].is_number());

crates/heal/src/heal/storage.rs

@@ -22,6 +22,7 @@ use serde::{Deserialize, Serialize};
 use std::sync::Arc;
 use tracing::{debug, error, warn};
 
+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 _,
     ObjectOperations as _, StorageAdminApi,
@@ -29,6 +30,37 @@ use super::storage_api::storage::{
 use super::{DiskStore, ECStore, Endpoint, HealDiskExt as _, StorageError, resume::ReplacementTargetIdentity};
 pub use super::{HealObjectInfo, HealObjectOptions, HealPutObjReader};
 
+#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
+pub struct HealBucketUsageBaseline {
+    pub objects_count: u64,
+    pub bytes: u64,
+}
+
+pub struct HealLifecycleExpiryContext {
+    inner: HealLifecycleExpiryContextInner,
+}
+
+enum HealLifecycleExpiryContextInner {
+    Ecstore(EcstoreHealLifecycleExpiryContext),
+    #[allow(dead_code)]
+    Test,
+}
+
+impl HealLifecycleExpiryContext {
+    fn ecstore(inner: EcstoreHealLifecycleExpiryContext) -> Self {
+        Self {
+            inner: HealLifecycleExpiryContextInner::Ecstore(inner),
+        }
+    }
+
+    #[cfg(test)]
+    pub(crate) fn test() -> Self {
+        Self {
+            inner: HealLifecycleExpiryContextInner::Test,
+        }
+    }
+}
+
 const LOG_COMPONENT_HEAL: &str = "heal";
 const LOG_SUBSYSTEM_STORAGE: &str = "storage";
 const EVENT_HEAL_STORAGE_OBJECT_IO: &str = "heal_storage_object_io";
@@ -272,6 +304,10 @@ pub struct HealListItem {
     pub name: String,
     /// normalized version id (`None` when the version is nil/absent)
     pub version_id: Option<String>,
+    /// version modification time as Unix nanoseconds
+    pub mod_time_unix_nanos: Option<i128>,
+    /// object snapshot for lifecycle evaluation
+    pub lifecycle_object_info: Option<HealObjectInfo>,
     /// whether this version is a delete marker (observability only)
     pub is_delete_marker: bool,
 }
@@ -329,6 +365,28 @@ pub trait HealStorageAPI: Send + Sync {
     /// Get bucket info
     async fn get_bucket_info(&self, bucket: &str) -> Result<Option<BucketInfo>>;
 
+    /// Aggregate usage-cache baselines for the requested buckets.
+    async fn erasure_set_usage_baseline(&self, _buckets: &[String]) -> Result<Option<HealBucketUsageBaseline>> {
+        Ok(None)
+    }
+
+    /// Load per-bucket lifecycle expiry context for heal skips.
+    async fn load_heal_lifecycle_expiry_context(&self, _bucket: &str) -> Result<Option<HealLifecycleExpiryContext>> {
+        Ok(None)
+    }
+
+    /// Queue lifecycle expiry for a version that heal can skip.
+    async fn enqueue_heal_lifecycle_expiry(
+        &self,
+        _context: &HealLifecycleExpiryContext,
+        _bucket: &str,
+        _object: &str,
+        _version_id: Option<&str>,
+        _object_info: Option<&HealObjectInfo>,
+    ) -> Result<bool> {
+        Ok(false)
+    }
+
     /// Fix bucket metadata
     async fn heal_bucket_metadata(&self, bucket: &str) -> Result<()>;
 
@@ -409,6 +467,7 @@ pub trait HealStorageAPI: Send + Sync {
         bucket: &str,
         prefix: &str,
         continuation_token: Option<&str>,
+        include_lifecycle_object_info: bool,
     ) -> Result<(Vec<HealListItem>, Option<String>, bool)>;
 
     /// List versions for healing via a per-erasure-set DISK-WALK union enumerator
@@ -427,8 +486,10 @@ pub trait HealStorageAPI: Send + Sync {
         bucket: &str,
         prefix: &str,
         continuation_token: Option<&str>,
+        include_lifecycle_object_info: bool,
     ) -> Result<(Vec<HealListItem>, Option<String>, bool)> {
-        self.list_objects_for_heal_page(bucket, prefix, continuation_token).await
+        self.list_objects_for_heal_page(bucket, prefix, continuation_token, include_lifecycle_object_info)
+            .await
     }
 
     /// Get disk for resume functionality.
@@ -1021,6 +1082,85 @@ impl HealStorageAPI for ECStoreHealStorage {
         }
     }
 
+    async fn erasure_set_usage_baseline(&self, buckets: &[String]) -> Result<Option<HealBucketUsageBaseline>> {
+        if buckets.is_empty() {
+            return Ok(None);
+        }
+
+        let info = match ecstore_load_admin_data_usage_from_backend_cached(self.ecstore.clone()).await {
+            Ok(info) if info.is_complete_bucket_usage_snapshot() => info,
+            Ok(_) | Err(_) => return Ok(None),
+        };
+
+        let mut baseline = HealBucketUsageBaseline::default();
+        for bucket in buckets {
+            if let Some(usage) = info.buckets_usage.get(bucket) {
+                baseline.objects_count = baseline.objects_count.saturating_add(usage.objects_count);
+                baseline.bytes = baseline.bytes.saturating_add(usage.size);
+            }
+        }
+
+        Ok(Some(baseline))
+    }
+
+    async fn load_heal_lifecycle_expiry_context(&self, bucket: &str) -> Result<Option<HealLifecycleExpiryContext>> {
+        match self.ecstore.load_heal_lifecycle_expiry_context(bucket).await {
+            Ok(Some(context)) => Ok(Some(HealLifecycleExpiryContext::ecstore(context))),
+            Ok(None) => Ok(None),
+            Err(err) => {
+                debug!(
+                    target: "rustfs::heal::storage",
+                    event = EVENT_HEAL_STORAGE_ADMIN_OP,
+                    component = LOG_COMPONENT_HEAL,
+                    subsystem = LOG_SUBSYSTEM_STORAGE,
+                    operation = "load_heal_lifecycle_expiry_context",
+                    bucket,
+                    result = "failed",
+                    error = %err,
+                    "Heal storage lifecycle expiry context load failed"
+                );
+                Ok(None)
+            }
+        }
+    }
+
+    async fn enqueue_heal_lifecycle_expiry(
+        &self,
+        context: &HealLifecycleExpiryContext,
+        bucket: &str,
+        object: &str,
+        version_id: Option<&str>,
+        object_info: Option<&HealObjectInfo>,
+    ) -> Result<bool> {
+        let context = match &context.inner {
+            HealLifecycleExpiryContextInner::Ecstore(context) => context,
+            HealLifecycleExpiryContextInner::Test => return Ok(false),
+        };
+        match self
+            .ecstore
+            .enqueue_heal_lifecycle_expiry(context, bucket, object, version_id, object_info)
+            .await
+        {
+            Ok(queued) => Ok(queued),
+            Err(err) => {
+                debug!(
+                    target: "rustfs::heal::storage",
+                    event = EVENT_HEAL_STORAGE_ADMIN_OP,
+                    component = LOG_COMPONENT_HEAL,
+                    subsystem = LOG_SUBSYSTEM_STORAGE,
+                    operation = "enqueue_heal_lifecycle_expiry",
+                    bucket,
+                    object,
+                    version_id = ?version_id,
+                    result = "failed",
+                    error = %err,
+                    "Heal storage lifecycle expiry check failed"
+                );
+                Ok(false)
+            }
+        }
+    }
+
     async fn heal_bucket_metadata(&self, bucket: &str) -> Result<()> {
         debug!(
             target: "rustfs::heal::storage",
@@ -1436,7 +1576,7 @@ impl HealStorageAPI for ECStoreHealStorage {
 
         loop {
             let (page_objects, next_token, is_truncated) = self
-                .list_objects_for_heal_page(bucket, prefix, continuation_token.as_deref())
+                .list_objects_for_heal_page(bucket, prefix, continuation_token.as_deref(), false)
                 .await?;
 
             all_objects.extend(page_objects);
@@ -1471,6 +1611,7 @@ impl HealStorageAPI for ECStoreHealStorage {
         bucket: &str,
         prefix: &str,
         continuation_token: Option<&str>,
+        include_lifecycle_object_info: bool,
     ) -> Result<(Vec<HealListItem>, Option<String>, bool)> {
         debug!(
             target: "rustfs::heal::storage",
@@ -1522,10 +1663,19 @@ impl HealStorageAPI for ECStoreHealStorage {
         let page_objects: Vec<HealListItem> = list_info
             .objects
             .into_iter()
-            .map(|obj| HealListItem {
+            .map(|mut obj| {
-                name: obj.name,
+                obj.version_id = obj.version_id.filter(|u| !u.is_nil());
-                version_id: obj.version_id.filter(|u| !u.is_nil()).map(|u| u.to_string()),
+                let version_id = obj.version_id.map(|u| u.to_string());
-                is_delete_marker: obj.delete_marker,
+                let mod_time_unix_nanos = obj.mod_time.map(|mod_time| mod_time.unix_timestamp_nanos());
+                let is_delete_marker = obj.delete_marker;
+                let lifecycle_object_info = include_lifecycle_object_info.then(|| obj.clone());
+                HealListItem {
+                    name: obj.name,
+                    version_id,
+                    mod_time_unix_nanos,
+                    lifecycle_object_info,
+                    is_delete_marker,
+                }
             })
             .collect();
         let page_count = page_objects.len();
@@ -1562,6 +1712,7 @@ impl HealStorageAPI for ECStoreHealStorage {
         bucket: &str,
         prefix: &str,
         continuation_token: Option<&str>,
+        include_lifecycle_object_info: bool,
     ) -> Result<(Vec<HealListItem>, Option<String>, bool)> {
         // Per-page bounds for the disk-walk union enumerator. Objects are atomic
         // (never split across pages), so version_budget only bounds how many
@@ -1590,7 +1741,16 @@ impl HealStorageAPI for ECStoreHealStorage {
 
         let (versions, next_forward, is_truncated) = self
             .ecstore
-            .heal_walk_versions_page(pool_idx, set_idx, bucket, prefix, forward_to.as_deref(), BATCH_OBJECTS, VERSION_BUDGET)
+            .heal_walk_versions_page(
+                pool_idx,
+                set_idx,
+                bucket,
+                prefix,
+                forward_to.as_deref(),
+                BATCH_OBJECTS,
+                VERSION_BUDGET,
+                include_lifecycle_object_info,
+            )
             .await
             .map_err(|e| {
                 error!(
@@ -1614,6 +1774,8 @@ impl HealStorageAPI for ECStoreHealStorage {
             .map(|v| HealListItem {
                 name: v.name,
                 version_id: v.version_id,
+                mod_time_unix_nanos: v.mod_time_unix_nanos,
+                lifecycle_object_info: v.lifecycle_object_info,
                 is_delete_marker: v.is_delete_marker,
             })
             .collect();

crates/heal/src/heal/storage_api.rs

@@ -12,7 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-pub(crate) use rustfs_ecstore::api::data_usage::DATA_USAGE_CACHE_NAME as ECSTORE_DATA_USAGE_CACHE_NAME;
+pub(crate) use rustfs_ecstore::api::data_usage::{
+    DATA_USAGE_CACHE_NAME as ECSTORE_DATA_USAGE_CACHE_NAME,
+    load_admin_data_usage_from_backend_cached as ecstore_load_admin_data_usage_from_backend_cached,
+};
 pub(crate) use rustfs_ecstore::api::disk::endpoint::Endpoint as EcstoreEndpoint;
 pub(crate) use rustfs_ecstore::api::disk::error::{DiskError as EcstoreDiskError, Result as EcstoreDiskResult};
 pub(crate) use rustfs_ecstore::api::disk::{
@@ -25,7 +28,9 @@ pub(crate) use rustfs_ecstore::api::disk::{
 pub(crate) use rustfs_ecstore::api::disk::{DiskOption as EcstoreDiskOption, new_disk as ecstore_new_disk};
 pub(crate) use rustfs_ecstore::api::error::{Error as EcstoreErrorType, StorageError as EcstoreStorageError};
 pub(crate) use rustfs_ecstore::api::runtime::local_disk_map_read as ecstore_local_disk_map_read;
-pub(crate) use rustfs_ecstore::api::storage::ECStore as EcstoreStore;
+pub(crate) use rustfs_ecstore::api::storage::{
+    ECStore as EcstoreStore, HealLifecycleExpiryContext as EcstoreHealLifecycleExpiryContext,
+};
 use rustfs_storage_api as storage_contracts;
 
 pub(crate) mod owner {
@@ -34,8 +39,8 @@ pub(crate) mod owner {
     pub(crate) use super::{
         ECSTORE_BUCKET_META_PREFIX, ECSTORE_DATA_USAGE_CACHE_NAME, ECSTORE_HEALING_MARKER_PATH, ECSTORE_RUSTFS_META_BUCKET,
         EcstoreConditionalFileUpdate, EcstoreDeleteOptions, EcstoreDiskAPI, EcstoreDiskBytes, EcstoreDiskError,
-        EcstoreDiskResult, EcstoreDiskStore, EcstoreEndpoint, EcstoreErrorType, EcstoreStorageError, EcstoreStore,
+        EcstoreDiskResult, EcstoreDiskStore, EcstoreEndpoint, EcstoreErrorType, EcstoreHealLifecycleExpiryContext,
-        ecstore_local_disk_map_read,
+        EcstoreStorageError, EcstoreStore, ecstore_load_admin_data_usage_from_backend_cached, ecstore_local_disk_map_read,
     };
 
     #[cfg(test)]

crates/heal/src/heal/task.rs

@@ -19,11 +19,12 @@ use crate::heal::{
     resume::{
         CheckpointManager, ReplacementPhase, ReplacementTargetIdentity, ResumeManager, replacement_target_identities_match,
     },
-    storage::{HealStorageAPI, next_heal_listing_token},
+    storage::{HealBucketUsageBaseline, HealStorageAPI, next_heal_listing_token},
 };
 use crate::{Error, Result};
 use metrics::{counter, histogram};
 use rustfs_common::heal_channel::{HealOpts, HealRequestSource, HealScanMode};
+use rustfs_common::trace_bus::{TraceEvent, TraceFunc, TraceKind, trace_emit};
 use rustfs_madmin::heal_commands::HealResultItem;
 use rustfs_utils::path::SLASH_SEPARATOR;
 use serde::{Deserialize, Serialize};
@@ -178,6 +179,17 @@ pub enum HealPriority {
     Urgent = 3,
 }
 
+impl HealPriority {
+    fn as_str(self) -> &'static str {
+        match self {
+            Self::Low => "low",
+            Self::Normal => "normal",
+            Self::High => "high",
+            Self::Urgent => "urgent",
+        }
+    }
+}
+
 /// Heal options
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct HealOptions {
@@ -498,6 +510,61 @@ impl HealTask {
         }
     }
 
+    fn emit_trace_task_state(&self, state: &'static str, duration: Duration, error: Option<&Error>) {
+        trace_emit(|| {
+            let mut event = TraceEvent::new(TraceKind::Heal, TraceFunc::HealTask)
+                .with_duration(duration)
+                .with_attr("task_id", self.id.as_str())
+                .with_attr("heal_type", self.heal_type.log_kind())
+                .with_attr("state", state)
+                .with_attr("source", self.source.as_str())
+                .with_attr("priority", self.priority.as_str())
+                .with_attr("retry_attempts", u64::from(self.retry_attempts))
+                .with_attr("dry_run", self.options.dry_run);
+
+            event = match &self.heal_type {
+                HealType::Cluster => event,
+                HealType::Object {
+                    bucket,
+                    object,
+                    version_id,
+                } => {
+                    let event = event.with_bucket(bucket.as_str()).with_object(object.as_str());
+                    match version_id {
+                        Some(version_id) => event.with_attr("version_id", version_id.as_str()),
+                        None => event,
+                    }
+                }
+                HealType::Bucket { bucket } => event.with_bucket(bucket.as_str()),
+                HealType::Prefix { bucket, prefix } => event.with_bucket(bucket.as_str()).with_object(prefix.as_str()),
+                HealType::ErasureSet { buckets, set_disk_id } => {
+                    let bucket_count = u64::try_from(buckets.len()).unwrap_or(u64::MAX);
+                    event
+                        .with_attr("set_disk_id", set_disk_id.as_str())
+                        .with_attr("bucket_count", bucket_count)
+                }
+                HealType::Metadata { bucket, object } => event.with_bucket(bucket.as_str()).with_object(object.as_str()),
+                HealType::ECDecode {
+                    bucket,
+                    object,
+                    version_id,
+                } => {
+                    let event = event.with_bucket(bucket.as_str()).with_object(object.as_str());
+                    match version_id {
+                        Some(version_id) => event.with_attr("version_id", version_id.as_str()),
+                        None => event,
+                    }
+                }
+                HealType::MRF { meta_path } => event.with_object(meta_path.as_str()),
+            };
+
+            match error {
+                Some(error) => event.with_attr("error", error.to_string()),
+                None => event,
+            }
+        });
+    }
+
     async fn remaining_timeout(&self) -> Result<Option<Duration>> {
         if let Some(total) = self.options.timeout {
             let start_instant = { *self.task_start_instant.read().await };
@@ -717,6 +784,7 @@ impl HealTask {
             queue_delay = ?queue_delay,
             "Heal task started"
         });
+        self.emit_trace_task_state("started", Duration::ZERO, None);
 
         let result = match &self.heal_type {
             HealType::Cluster => self.heal_cluster().await,
@@ -805,6 +873,14 @@ impl HealTask {
             }
         }
 
+        let terminal_state = match &result {
+            Ok(_) => "completed",
+            Err(Error::TaskCancelled) => "cancelled",
+            Err(Error::TaskTimeout) => "timed_out",
+            Err(_) => "failed",
+        };
+        self.emit_trace_task_state(terminal_state, start_instant.elapsed(), result.as_ref().err());
+
         result
     }
 
@@ -1535,7 +1611,7 @@ impl HealTask {
             let (objects, next_token, is_truncated) = self
                 .await_with_control(
                     self.storage
-                        .list_objects_for_heal_page(bucket, prefix, continuation_token.as_deref()),
+                        .list_objects_for_heal_page(bucket, prefix, continuation_token.as_deref(), false),
                 )
                 .await?;
 
@@ -1697,6 +1773,23 @@ impl HealTask {
         Ok(())
     }
 
+    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
+        {
+            Ok(Some(baseline)) => baseline,
+            Ok(None) => return Ok(()),
+            Err(err @ Error::TaskCancelled) | Err(err @ Error::TaskTimeout) => return Err(err),
+            Err(_) => return Ok(()),
+        };
+
+        let HealBucketUsageBaseline { objects_count, bytes } = baseline;
+        let mut progress = self.progress.write().await;
+        progress.set_total_baseline(objects_count, bytes);
+        Ok(())
+    }
+
     async fn heal_metadata(&self, bucket: &str, object: &str) -> Result<()> {
         debug!(
             target: "rustfs::heal::task",
@@ -2298,6 +2391,8 @@ impl HealTask {
             None
         };
 
+        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() {
             let state = resume_manager.get_state().await;
@@ -2602,7 +2697,8 @@ impl HealTask {
 
         {
             let mut progress = self.progress.write().await;
-            progress.update_progress(4, 4, 0, 0);
+            let bytes_processed = progress.bytes_processed;
+            progress.update_progress(4, 4, 0, bytes_processed);
         }
 
         match result {
@@ -2658,6 +2754,7 @@ mod tests {
     use super::super::{DiskOption, DiskStore, Endpoint, HealDiskExt as _, new_disk};
     use super::*;
     use crate::heal::storage::{DiskStatus, HealListItem, HealObjectInfo};
+    use rustfs_common::trace_bus::{TraceEvent, TraceFunc, TraceKind, TraceSubscription, TraceVal, subscribe_trace_events};
     use rustfs_madmin::heal_commands::{HealDriveInfo, HealResultItem, Infos};
     use std::collections::{HashMap, VecDeque};
     use std::sync::Mutex;
@@ -3203,6 +3300,8 @@ mod tests {
         block_heal_object: Mutex<bool>,
         resume_disk: Mutex<Option<DiskStore>>,
         replacement_resume_disk: Mutex<Option<DiskStore>>,
+        usage_baseline: Mutex<Option<HealBucketUsageBaseline>>,
+        usage_baseline_error: Mutex<bool>,
     }
 
     #[test]
@@ -3265,11 +3364,69 @@ mod tests {
         assert_eq!(samples_logged, MAX_BUCKET_FAILURE_LOG_SAMPLES);
     }
 
+    #[tokio::test]
+    async fn execute_emits_heal_trace_task_state() {
+        let mut trace = subscribe_trace_events();
+        let storage = Arc::new(MockStorage::default());
+        let task = HealTask::from_request(
+            HealRequest::object("bucket-a".to_string(), "object-a".to_string(), Some("version-a".to_string())),
+            storage,
+        );
+
+        task.execute().await.expect("mock object heal should complete");
+
+        let started = recv_trace_task_state(&mut trace, &task.id, "started").await;
+        assert_eq!(started.kind, TraceKind::Heal);
+        assert_eq!(started.func, TraceFunc::HealTask);
+        assert_eq!(started.bucket.as_deref(), Some("bucket-a"));
+        assert_eq!(started.object.as_deref(), Some("object-a"));
+        assert_eq!(trace_attr_string(&started, "heal_type").as_deref(), Some("object"));
+        assert_eq!(trace_attr_string(&started, "source").as_deref(), Some("internal"));
+        assert_eq!(trace_attr_string(&started, "version_id").as_deref(), Some("version-a"));
+
+        let completed = recv_trace_task_state(&mut trace, &task.id, "completed").await;
+        assert_eq!(completed.kind, TraceKind::Heal);
+        assert_eq!(completed.func, TraceFunc::HealTask);
+        assert_eq!(trace_attr_string(&completed, "state").as_deref(), Some("completed"));
+    }
+
+    async fn recv_trace_task_state(trace: &mut TraceSubscription, task_id: &str, state: &str) -> TraceEvent {
+        for _ in 0..32 {
+            let event = tokio::time::timeout(Duration::from_secs(1), trace.recv())
+                .await
+                .expect("trace event should arrive")
+                .expect("trace bus should stay open");
+            if trace_attr_string(&event, "task_id").as_deref() == Some(task_id)
+                && trace_attr_string(&event, "state").as_deref() == Some(state)
+            {
+                return (*event).clone();
+            }
+        }
+
+        panic!("expected trace state {state} for task {task_id}");
+    }
+
+    fn trace_attr_string(event: &TraceEvent, key: &str) -> Option<String> {
+        event.attrs.iter().find_map(|attr| {
+            if attr.key != key {
+                return None;
+            }
+            Some(match &attr.value {
+                TraceVal::Bool(value) => value.to_string(),
+                TraceVal::U64(value) => value.to_string(),
+                TraceVal::I64(value) => value.to_string(),
+                TraceVal::Str(value) => value.to_string(),
+            })
+        })
+    }
+
     /// Build a latest, non-delete-marker heal list item with no version id.
     fn heal_item(name: &str) -> HealListItem {
         HealListItem {
             name: name.to_string(),
             version_id: None,
+            mod_time_unix_nanos: None,
+            lifecycle_object_info: None,
             is_delete_marker: false,
         }
     }
@@ -3357,6 +3514,13 @@ mod tests {
             }))
         }
 
+        async fn erasure_set_usage_baseline(&self, _buckets: &[String]) -> Result<Option<HealBucketUsageBaseline>> {
+            if *self.usage_baseline_error.lock().unwrap() {
+                return Err(Error::Other("usage baseline unavailable".to_string()));
+            }
+            Ok(*self.usage_baseline.lock().unwrap())
+        }
+
         async fn heal_bucket_metadata(&self, _bucket: &str) -> Result<()> {
             Ok(())
         }
@@ -3540,6 +3704,7 @@ mod tests {
             bucket: &str,
             prefix: &str,
             continuation_token: Option<&str>,
+            _include_lifecycle_object_info: bool,
         ) -> Result<(Vec<HealListItem>, Option<String>, bool)> {
             self.listed_prefixes.lock().unwrap().push(prefix.to_string());
             if *self.truncate_without_token.lock().unwrap() {
@@ -4654,6 +4819,73 @@ mod tests {
         assert!(storage.object_heal_opts.lock().unwrap().is_empty());
     }
 
+    #[tokio::test]
+    async fn erasure_set_heal_applies_usage_baseline_to_progress() {
+        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,
+            })),
+            ..Default::default()
+        });
+        let request = HealRequest::new(
+            HealType::ErasureSet {
+                buckets: vec!["bucket-a".to_string()],
+                set_disk_id: "pool_0_set_0".to_string(),
+            },
+            HealOptions {
+                timeout: None,
+                ..Default::default()
+            },
+            HealPriority::Normal,
+        );
+        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_eq!(progress.objects_total_count, 10);
+        assert_eq!(progress.objects_total_size, 8);
+        assert_eq!(progress.bytes_processed, 2);
+        assert!((progress.progress_percentage - 25.0).abs() < 0.001);
+    }
+
+    #[tokio::test]
+    async fn erasure_set_heal_ignores_usage_baseline_errors() {
+        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_error: Mutex::new(true),
+            ..Default::default()
+        });
+        let request = HealRequest::new(
+            HealType::ErasureSet {
+                buckets: vec!["bucket-a".to_string()],
+                set_disk_id: "pool_0_set_0".to_string(),
+            },
+            HealOptions {
+                timeout: None,
+                ..Default::default()
+            },
+            HealPriority::Normal,
+        );
+        let task = HealTask::from_request(request, storage);
+
+        task.heal_erasure_set(vec!["bucket-a".to_string()], "pool_0_set_0".to_string())
+            .await
+            .expect("usage baseline failures should not fail erasure set heal");
+
+        let progress = task.get_progress().await;
+        assert_eq!(progress.objects_total_count, 0);
+        assert_eq!(progress.objects_total_size, 0);
+    }
+
     #[tokio::test]
     async fn resumable_erasure_set_execution_is_cancelled_while_object_heal_is_pending() {
         let temp = TempDir::new().expect("temporary directory should be created");

crates/heal/src/lib.rs

@@ -445,6 +445,7 @@ mod tests {
             _bucket: &str,
             _prefix: &str,
             _continuation_token: Option<&str>,
+            _include_lifecycle_object_info: bool,
         ) -> Result<(Vec<HealListItem>, Option<String>, bool), Error> {
             Ok((Vec::new(), None, false))
         }

crates/heal/tests/heal_b5_versioned_regression_test.rs

@@ -176,7 +176,7 @@ async fn enumerate_all_versions(heal_storage: &Arc<ECStoreHealStorage>, bucket:
     let mut token: Option<String> = None;
     loop {
         let (page, next, truncated) = heal_storage
-            .list_objects_for_heal_page(bucket, "", token.as_deref())
+            .list_objects_for_heal_page(bucket, "", token.as_deref(), false)
             .await
             .expect("list_objects_for_heal_page failed");
         items.extend(page);

crates/heal/tests/heal_b920_subquorum_union_test.rs

@@ -166,7 +166,7 @@ async fn enumerate_b5(heal_storage: &Arc<ECStoreHealStorage>, bucket: &str) -> V
     let mut token: Option<String> = None;
     loop {
         let (page, next, truncated) = heal_storage
-            .list_objects_for_heal_page(bucket, "", token.as_deref())
+            .list_objects_for_heal_page(bucket, "", token.as_deref(), false)
             .await
             .expect("b5 list page failed");
         items.extend(page);
@@ -187,7 +187,7 @@ async fn enumerate_disk_walk(heal_storage: &Arc<ECStoreHealStorage>, bucket: &st
     let mut token: Option<String> = None;
     loop {
         let (page, next, truncated) = heal_storage
-            .list_versions_for_heal_page_disk_walk(SET_DISK_ID, bucket, "", token.as_deref())
+            .list_versions_for_heal_page_disk_walk(SET_DISK_ID, bucket, "", token.as_deref(), false)
             .await
             .expect("disk-walk list page failed");
         items.extend(page);
@@ -418,7 +418,7 @@ mod serial_tests {
         let mut pages = 0usize;
         loop {
             let (versions, next_forward, truncated) = ecstore
-                .heal_walk_versions_page(0, 0, bucket, "", forward.as_deref(), 2, 100_000)
+                .heal_walk_versions_page(0, 0, bucket, "", forward.as_deref(), 2, 100_000, false)
                 .await
                 .expect("heal_walk_versions_page failed");
             pages += 1;

crates/heal/tests/heal_bug_fixes_test.rs

@@ -242,6 +242,7 @@ fn test_heal_task_status_atomic_update() {
             _bucket: &str,
             _prefix: &str,
             _continuation_token: Option<&str>,
+            _include_lifecycle_object_info: bool,
         ) -> rustfs_heal::Result<(Vec<HealListItem>, Option<String>, bool)> {
             Ok((vec![], None, false))
         }
@@ -385,6 +386,7 @@ async fn test_heal_task_transient_object_exists_skip_avoids_recreate() {
             _bucket: &str,
             _prefix: &str,
             _continuation_token: Option<&str>,
+            _include_lifecycle_object_info: bool,
         ) -> rustfs_heal::Result<(Vec<HealListItem>, Option<String>, bool)> {
             Ok((Vec::new(), None, false))
         }

crates/madmin/src/service_commands.rs

@@ -43,7 +43,7 @@ pub struct ServiceTraceOpts {
 
 #[allow(dead_code)]
 impl ServiceTraceOpts {
-    fn trace_types(&self) -> TraceType {
+    pub fn trace_types(&self) -> TraceType {
         let mut tt = TraceType::default();
         tt.set_if(self.s3, &TraceType::S3);
         tt.set_if(self.internal, &TraceType::INTERNAL);
@@ -72,6 +72,14 @@ impl ServiceTraceOpts {
         tt
     }
 
+    pub fn only_errors(&self) -> bool {
+        self.only_errors
+    }
+
+    pub fn threshold(&self) -> Duration {
+        self.threshold
+    }
+
     pub fn parse_params(&mut self, uri: &Uri) -> Result<(), String> {
         let query_pairs: HashMap<_, _> = uri
             .query()

crates/scanner/src/scanner_folder.rs

@@ -41,6 +41,7 @@ use rustfs_common::metrics::{
     CloseDiskGuard, IlmAction, Metric, Metrics, ScannerReplicationRepairKind, ScannerSourceWorkUpdate, ScannerWorkSource,
     UpdateCurrentPathFn, current_path_updater, global_metrics,
 };
+use rustfs_common::trace_bus::{TraceEvent, TraceFunc, TraceKind, trace_emit, trace_subscriber_count};
 use rustfs_filemeta::{MetaCacheEntries, MetaCacheEntry, MetadataResolutionParams};
 use rustfs_utils::path::{SLASH_SEPARATOR, path_join_buf};
 use s3s::dto::{BucketLifecycleConfiguration, ObjectLockConfiguration};
@@ -430,6 +431,113 @@ fn non_negative_i64_to_u64(value: i64) -> u64 {
     value.max(0) as u64
 }
 
+fn trace_start_instant() -> Option<Instant> {
+    (trace_subscriber_count() > 0).then(Instant::now)
+}
+
+fn emit_scanner_folder_trace(root: &str, folder: &str, objects: u64, started_at: Option<Instant>, state: &'static str) {
+    let Some(started_at) = started_at else {
+        return;
+    };
+
+    trace_emit(|| {
+        let (bucket, prefix) = path2_bucket_object_with_base_path(root, folder);
+        TraceEvent::new(TraceKind::Scanner, TraceFunc::ScannerFolder)
+            .with_bucket(bucket)
+            .with_object(prefix)
+            .with_duration(started_at.elapsed())
+            .with_attr("state", state)
+            .with_attr("objects", objects)
+    });
+}
+
+fn emit_scanner_ilm_action_trace(
+    bucket: &str,
+    object: &str,
+    action: IlmAction,
+    count: u64,
+    queued: bool,
+    started_at: Option<Instant>,
+) {
+    let Some(started_at) = started_at else {
+        return;
+    };
+
+    let state = if queued { "queued" } else { "not_queued" };
+    trace_emit(|| {
+        TraceEvent::new(TraceKind::Scanner, TraceFunc::ScannerIlmAction)
+            .with_bucket(bucket)
+            .with_object(object)
+            .with_duration(started_at.elapsed())
+            .with_attr("state", state)
+            .with_attr("action", action.as_str())
+            .with_attr("count", count)
+            .with_attr("queued", queued)
+    });
+}
+
+struct ScannerHealCandidateTraceContext {
+    bucket: String,
+    object: Option<String>,
+    version_id: Option<String>,
+    scan_mode: Option<HealScanMode>,
+    started_at: Instant,
+}
+
+fn scanner_heal_candidate_trace_context(request: &HealChannelRequest) -> Option<ScannerHealCandidateTraceContext> {
+    let started_at = trace_start_instant()?;
+    Some(ScannerHealCandidateTraceContext {
+        bucket: request.bucket.clone(),
+        object: request.object_prefix.clone(),
+        version_id: request.object_version_id.clone(),
+        scan_mode: request.scan_mode,
+        started_at,
+    })
+}
+
+struct ScannerHealCandidateTrace<'a> {
+    candidate_type: &'static str,
+    bucket: &'a str,
+    object: Option<&'a str>,
+    version_id: Option<&'a str>,
+    priority: HealChannelPriority,
+    scan_mode: Option<HealScanMode>,
+    result: Result<HealAdmissionResult, &'a str>,
+    started_at: Instant,
+}
+
+fn emit_scanner_heal_candidate_trace(trace: ScannerHealCandidateTrace<'_>) {
+    trace_emit(|| {
+        let (state, admission, error) = match trace.result {
+            Ok(result) if result.is_admitted() => ("admitted", describe_heal_admission(result), None),
+            Ok(result) => ("not_admitted", describe_heal_admission(result), None),
+            Err(error) => ("submit_failed", "channel_error".to_string(), Some(error)),
+        };
+        let mut event = TraceEvent::new(TraceKind::Scanner, TraceFunc::ScannerHealCandidate)
+            .with_bucket(trace.bucket)
+            .with_duration(trace.started_at.elapsed())
+            .with_attr("state", state)
+            .with_attr("candidate_type", trace.candidate_type)
+            .with_attr("priority", heal_priority_label(trace.priority))
+            .with_attr("admission", admission);
+
+        if let Some(object) = trace.object {
+            event = event.with_object(object);
+        }
+        if let Some(version_id) = trace.version_id {
+            event = event.with_attr("version_id", version_id);
+        }
+        if let Some(scan_mode) = trace.scan_mode {
+            event = event.with_attr("scan_mode", scan_mode.as_str());
+        }
+        if let Some(error) = error {
+            event = event.with_attr("error", error);
+        }
+
+        event
+    });
+}
+
 fn apply_scanner_size_summary(into: &mut DataUsageEntry, summary: &SizeSummary) {
     into.size = into.size.saturating_add(summary.total_size);
     into.versions = into.versions.saturating_add(summary.versions);
@@ -677,9 +785,22 @@ async fn send_scanner_heal_request(
     request: HealChannelRequest,
 ) -> Result<HealAdmissionResult, ScannerError> {
     let priority = request.priority;
+    let trace_context = scanner_heal_candidate_trace_context(&request);
     match send_heal_request_with_admission(request).await {
         Ok(result) => {
             record_heal_candidate_admission(candidate_type, priority, result);
+            if let Some(trace_context) = trace_context.as_ref() {
+                emit_scanner_heal_candidate_trace(ScannerHealCandidateTrace {
+                    candidate_type,
+                    bucket: &trace_context.bucket,
+                    object: trace_context.object.as_deref(),
+                    version_id: trace_context.version_id.as_deref(),
+                    priority,
+                    scan_mode: trace_context.scan_mode,
+                    result: Ok(result),
+                    started_at: trace_context.started_at,
+                });
+            }
             Ok(result)
         }
         Err(err) => {
@@ -690,6 +811,18 @@ async fn send_scanner_heal_request(
                 "result" => "channel_error".to_string()
             )
             .increment(1);
+            if let Some(trace_context) = trace_context.as_ref() {
+                emit_scanner_heal_candidate_trace(ScannerHealCandidateTrace {
+                    candidate_type,
+                    bucket: &trace_context.bucket,
+                    object: trace_context.object.as_deref(),
+                    version_id: trace_context.version_id.as_deref(),
+                    priority,
+                    scan_mode: trace_context.scan_mode,
+                    result: Err(err.as_str()),
+                    started_at: trace_context.started_at,
+                });
+            }
             Err(ScannerError::Other(err))
         }
     }
@@ -905,7 +1038,9 @@ impl ScannerItem {
                             "Scanner lifecycle action dispatched"
                         );
                         let done_ilm = Metrics::time_ilm(event.action);
+                        let trace_started_at = trace_start_instant();
                         let queued = apply_expiry_rule(event, &LcEventSrc::Scanner, oi).await;
+                        emit_scanner_ilm_action_trace(&self.bucket, &oi.name, event.action, 1, queued, trace_started_at);
                         if record_scanner_ilm_action_if_queued(global_metrics(), event.action, 1, queued) {
                             done_ilm(1)();
                             remaining_versions = 0;
@@ -957,7 +1092,9 @@ impl ScannerItem {
                             "Scanner lifecycle action dispatched"
                         );
                         let done_ilm = Metrics::time_ilm(event.action);
+                        let trace_started_at = trace_start_instant();
                         let queued = apply_expiry_rule(event, &LcEventSrc::Scanner, oi).await;
+                        emit_scanner_ilm_action_trace(&self.bucket, &oi.name, event.action, 1, queued, trace_started_at);
                         if record_scanner_ilm_action_if_queued(global_metrics(), event.action, 1, queued) {
                             done_ilm(1)();
                             if !versioning_config.prefix_enabled(&self.object_path()) && event.action == IlmAction::DeleteAction {
@@ -995,7 +1132,9 @@ impl ScannerItem {
                             "Scanner lifecycle action dispatched"
                         );
                         let done_ilm = Metrics::time_ilm(event.action);
+                        let trace_started_at = trace_start_instant();
                         let queued = apply_transition_rule(event, &LcEventSrc::Scanner, oi).await;
+                        emit_scanner_ilm_action_trace(&self.bucket, &oi.name, event.action, 1, queued, trace_started_at);
                         if record_scanner_ilm_action_if_queued(global_metrics(), event.action, 1, queued) {
                             done_ilm(1)();
                         }
@@ -1019,7 +1158,21 @@ impl ScannerItem {
             let action = event.action;
             let count = u64::try_from(to_delete_objs.len()).unwrap_or(u64::MAX);
             let done_ilm = Metrics::time_ilm(action);
+            let trace_started_at = trace_start_instant();
             let queued = enqueue_runtime_newer_noncurrent(&self.bucket, to_delete_objs, event, &LcEventSrc::Scanner).await;
+            if let Some(trace_started_at) = trace_started_at {
+                let state = if queued { "queued" } else { "not_queued" };
+                trace_emit(|| {
+                    TraceEvent::new(TraceKind::Scanner, TraceFunc::ScannerIlmAction)
+                        .with_bucket(self.bucket.as_str())
+                        .with_object(self.object_path())
+                        .with_duration(trace_started_at.elapsed())
+                        .with_attr("state", state)
+                        .with_attr("action", action.as_str())
+                        .with_attr("count", count)
+                        .with_attr("queued", queued)
+                });
+            }
             if record_scanner_ilm_action_if_queued(global_metrics(), action, count, queued) {
                 done_ilm(count)();
                 remaining_versions = remaining_versions.saturating_sub(noncurrent_accounting.len());
@@ -1830,6 +1983,7 @@ impl FolderScanner {
         into: &mut DataUsageEntry,
     ) -> Result<(), ScannerError> {
         let done_folder = Metrics::time(Metric::ScanFolder);
+        let trace_started_at = trace_start_instant();
 
         if ctx.is_cancelled() {
             return Err(ScannerError::Other("Operation cancelled".to_string()));
@@ -2895,6 +3049,8 @@ impl FolderScanner {
         }
 
         done_folder();
+        let scanned_objects = u64::try_from(into.objects).unwrap_or(u64::MAX);
+        emit_scanner_folder_trace(&self.root, &folder.name, scanned_objects, trace_started_at, "completed");
 
         Ok(())
     }
@@ -4400,6 +4556,104 @@ mod tests {
         );
     }
 
+    #[tokio::test]
+    async fn scanner_trace_helpers_emit_expected_events() {
+        let mut trace = rustfs_common::trace_bus::subscribe_trace_events();
+
+        emit_scanner_folder_trace(
+            "/tmp/rustfs-scanner-trace",
+            "/tmp/rustfs-scanner-trace/bucket-a/folder-a",
+            7,
+            Some(Instant::now()),
+            "completed",
+        );
+        let folder = recv_scanner_trace_event(
+            &mut trace,
+            TraceFunc::ScannerFolder,
+            Some("bucket-a"),
+            Some("folder-a"),
+            Some("completed"),
+        )
+        .await;
+        assert_eq!(trace_attr_string(&folder, "objects").as_deref(), Some("7"));
+
+        emit_scanner_ilm_action_trace("bucket-a", "object-a", IlmAction::DeleteAction, 2, true, Some(Instant::now()));
+        let ilm = recv_scanner_trace_event(
+            &mut trace,
+            TraceFunc::ScannerIlmAction,
+            Some("bucket-a"),
+            Some("object-a"),
+            Some("queued"),
+        )
+        .await;
+        assert_eq!(trace_attr_string(&ilm, "action").as_deref(), Some("delete"));
+        assert_eq!(trace_attr_string(&ilm, "count").as_deref(), Some("2"));
+        assert_eq!(trace_attr_string(&ilm, "queued").as_deref(), Some("true"));
+
+        emit_scanner_heal_candidate_trace(ScannerHealCandidateTrace {
+            candidate_type: "object",
+            bucket: "bucket-a",
+            object: Some("object-a"),
+            version_id: Some("version-a"),
+            priority: HealChannelPriority::High,
+            scan_mode: Some(HealScanMode::Deep),
+            result: Ok(HealAdmissionResult::Merged),
+            started_at: Instant::now(),
+        });
+        let heal_candidate = recv_scanner_trace_event(
+            &mut trace,
+            TraceFunc::ScannerHealCandidate,
+            Some("bucket-a"),
+            Some("object-a"),
+            Some("admitted"),
+        )
+        .await;
+        assert_eq!(trace_attr_string(&heal_candidate, "candidate_type").as_deref(), Some("object"));
+        assert_eq!(trace_attr_string(&heal_candidate, "priority").as_deref(), Some("high"));
+        assert_eq!(trace_attr_string(&heal_candidate, "scan_mode").as_deref(), Some("deep"));
+        assert_eq!(trace_attr_string(&heal_candidate, "version_id").as_deref(), Some("version-a"));
+        assert_eq!(trace_attr_string(&heal_candidate, "admission").as_deref(), Some("merged"));
+    }
+
+    async fn recv_scanner_trace_event(
+        trace: &mut rustfs_common::trace_bus::TraceSubscription,
+        func: TraceFunc,
+        bucket: Option<&str>,
+        object: Option<&str>,
+        state: Option<&str>,
+    ) -> TraceEvent {
+        for _ in 0..32 {
+            let event = tokio::time::timeout(Duration::from_secs(1), trace.recv())
+                .await
+                .expect("scanner trace event should arrive")
+                .expect("trace bus should stay open");
+            if event.kind == TraceKind::Scanner
+                && event.func == func
+                && event.bucket.as_deref() == bucket
+                && event.object.as_deref() == object
+                && state.is_none_or(|state| trace_attr_string(&event, "state").as_deref() == Some(state))
+            {
+                return (*event).clone();
+            }
+        }
+
+        panic!("expected scanner trace event {func:?} for bucket {bucket:?} object {object:?}");
+    }
+
+    fn trace_attr_string(event: &TraceEvent, key: &str) -> Option<String> {
+        event.attrs.iter().find_map(|attr| {
+            if attr.key != key {
+                return None;
+            }
+            Some(match &attr.value {
+                rustfs_common::trace_bus::TraceVal::Bool(value) => value.to_string(),
+                rustfs_common::trace_bus::TraceVal::U64(value) => value.to_string(),
+                rustfs_common::trace_bus::TraceVal::I64(value) => value.to_string(),
+                rustfs_common::trace_bus::TraceVal::Str(value) => value.to_string(),
+            })
+        })
+    }
+
     #[test]
     fn test_build_high_priority_heal_admission_error_contains_context() {
         let err = build_high_priority_heal_admission_error(

rustfs/src/admin/handlers/profile_admin.rs

@@ -23,18 +23,24 @@ use futures::{Stream, StreamExt};
 use http::{HeaderMap, HeaderValue};
 use hyper::{Method, StatusCode};
 use matchit::Params;
+use regex::Regex;
+use rustfs_common::trace_bus::{TraceEvent, TraceKind, TraceVal, subscribe_trace_events};
 use rustfs_madmin::service_commands::ServiceTraceOpts;
+use rustfs_madmin::trace::TraceType;
 use rustfs_policy::policy::action::{Action, AdminAction};
 use s3s::header::CONTENT_TYPE;
 use s3s::stream::{ByteStream, DynByteStream};
 use s3s::{Body, S3Request, S3Response, S3Result, StdError, s3_error};
 use serde::Serialize;
+use std::collections::HashMap;
 use std::pin::Pin;
 use std::task::{Context, Poll};
-use std::time::Duration;
+use std::time::{Duration, SystemTime};
+use time::{OffsetDateTime, format_description::well_known::Rfc3339};
 use tokio::sync::mpsc;
 use tokio_stream::wrappers::ReceiverStream;
 use tracing::error;
+use url::form_urlencoded;
 
 #[derive(Serialize)]
 struct ProfileStatus {
@@ -206,16 +212,164 @@ impl Stream for TraceStream {
 
 impl ByteStream for TraceStream {}
 
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+struct TraceKindFilter {
+    heal: bool,
+    scanner: bool,
+}
+
+impl TraceKindFilter {
+    const ALL_SUPPORTED: Self = Self {
+        heal: true,
+        scanner: true,
+    };
+
+    fn from_request(uri: &hyper::Uri, trace_types: TraceType) -> S3Result<Self> {
+        let mut has_kind = false;
+        let mut filter = Self {
+            heal: false,
+            scanner: false,
+        };
+
+        for (key, value) in trace_query_pairs(uri) {
+            if key != "kind" {
+                continue;
+            }
+            has_kind = true;
+            for item in value.split(',') {
+                match item.trim().to_ascii_lowercase().as_str() {
+                    "heal" | "healing" => filter.heal = true,
+                    "scanner" => filter.scanner = true,
+                    "all" => return Ok(Self::ALL_SUPPORTED),
+                    _ => return Err(s3_error!(InvalidRequest, "invalid trace kind")),
+                }
+            }
+        }
+
+        if has_kind {
+            return Ok(filter);
+        }
+
+        if trace_types.mask() == 0 || trace_query_flag(uri, "all") {
+            return Ok(Self::ALL_SUPPORTED);
+        }
+
+        Ok(Self {
+            heal: trace_types.overlaps(&TraceType::HEALING),
+            scanner: trace_types.overlaps(&TraceType::SCANNER),
+        })
+    }
+
+    const fn matches(self, kind: TraceKind) -> bool {
+        match kind {
+            TraceKind::Heal => self.heal,
+            TraceKind::Scanner => self.scanner,
+        }
+    }
+}
+
+#[derive(Debug)]
+struct TraceStreamFilter {
+    kinds: TraceKindFilter,
+    regex: Option<Regex>,
+    threshold: Duration,
+}
+
+impl TraceStreamFilter {
+    fn from_request(uri: &hyper::Uri, opts: &ServiceTraceOpts) -> S3Result<Self> {
+        if opts.only_errors() {
+            return Err(s3_error!(
+                InvalidRequest,
+                "trace error-only filter is not supported for heal/scanner trace"
+            ));
+        }
+
+        Ok(Self {
+            kinds: TraceKindFilter::from_request(uri, opts.trace_types())?,
+            regex: trace_regex_filter(uri)?,
+            threshold: opts.threshold(),
+        })
+    }
+
+    fn matches_kind(&self, kind: TraceKind) -> bool {
+        self.kinds.matches(kind)
+    }
+
+    fn matches_record(&self, record: &TraceWireRecord) -> bool {
+        record.duration >= self.threshold && self.regex.as_ref().is_none_or(|regex| record.matches_regex(regex))
+    }
+}
+
+#[derive(Serialize)]
+struct TraceWireRecord {
+    #[serde(rename = "type")]
+    trace_type: u64,
+    #[serde(rename = "nodename")]
+    node_name: String,
+    #[serde(rename = "funcname")]
+    func_name: String,
+    #[serde(rename = "time")]
+    time: String,
+    #[serde(rename = "path")]
+    path: String,
+    #[serde(rename = "dur")]
+    duration: Duration,
+    #[serde(rename = "bytes", skip_serializing_if = "Option::is_none")]
+    bytes: Option<i64>,
+    #[serde(rename = "msg", skip_serializing_if = "Option::is_none")]
+    message: Option<String>,
+    #[serde(rename = "custom", skip_serializing_if = "Option::is_none")]
+    custom: Option<HashMap<String, String>>,
+}
+
+impl TraceWireRecord {
+    fn from_event(node_name: &str, event: &TraceEvent) -> Self {
+        Self {
+            trace_type: trace_type_mask(event.kind),
+            node_name: node_name.to_owned(),
+            func_name: event.func.as_str().to_owned(),
+            time: trace_time_string(event.time),
+            path: trace_path(event),
+            duration: event.duration,
+            bytes: trace_bytes(event.bytes),
+            message: None,
+            custom: trace_custom_attrs(event),
+        }
+    }
+
+    fn dropped(node_name: &str, dropped: u64) -> Self {
+        let mut custom = HashMap::new();
+        custom.insert("dropped_events".to_string(), dropped.to_string());
+
+        Self {
+            trace_type: 0,
+            node_name: node_name.to_owned(),
+            func_name: "trace.Dropped".to_string(),
+            time: trace_time_string(SystemTime::now()),
+            path: String::new(),
+            duration: Duration::ZERO,
+            bytes: None,
+            message: Some("trace subscriber lagged".to_string()),
+            custom: Some(custom),
+        }
+    }
+
+    fn matches_regex(&self, regex: &Regex) -> bool {
+        regex.is_match(&self.func_name)
+            || regex.is_match(&self.path)
+            || self.message.as_ref().is_some_and(|message| regex.is_match(message))
+            || self
+                .custom
+                .as_ref()
+                .is_some_and(|custom| custom.iter().any(|(key, value)| regex.is_match(key) || regex.is_match(value)))
+    }
+}
+
 /// `GET /v3/trace` — stream real-time server trace events.
 ///
-/// RustFS emits diagnostics through the `tracing` pipeline but does not expose
+/// RustFS currently publishes heal and scanner diagnostics through the common
-/// an in-process subscriber that can fan trace events out to an admin client
+/// trace bus. The admin endpoint exposes those events as MinIO-shaped NDJSON
-/// (there is no request-trace broadcast channel). Rather than return an opaque
+/// records while keeping unsupported trace classes filtered out.
-/// `501` — which would make `mc admin trace` fail to connect — this honors the
-/// streaming NDJSON contract: it validates the requested trace filters, opens
-/// the stream, emits a single capability record explaining that live tracing is
-/// not wired, then holds the connection open with keep-alives. No fabricated
-/// trace records are ever sent.
 pub struct TraceHandler {}
 
 #[async_trait::async_trait]
@@ -228,24 +382,13 @@ impl Operation for TraceHandler {
         let mut opts = ServiceTraceOpts::default();
         opts.parse_params(&req.uri)
             .map_err(|_| s3_error!(InvalidRequest, "invalid trace parameters"))?;
+        let filter = TraceStreamFilter::from_request(&req.uri, &opts)?;
 
         let node_name = sysinfo::System::host_name().unwrap_or_else(|| "rustfs".to_string());
-        let (tx, rx) = mpsc::channel::<Result<Bytes, StdError>>(8);
+        let mut subscription = subscribe_trace_events();
+        let (tx, rx) = mpsc::channel::<Result<Bytes, StdError>>(64);
 
         spawn_traced(async move {
-            let notice = serde_json::json!({
-                "nodename": node_name,
-                "funcname": "admin.Trace",
-                "msg": "RustFS does not expose an in-process trace-event subscriber; live tracing is not yet available",
-                "err": "trace_streaming_unsupported",
-            });
-            if let Ok(mut encoded) = serde_json::to_vec(&notice) {
-                encoded.push(b'\n');
-                if tx.send(Ok(Bytes::from(encoded))).await.is_err() {
-                    return;
-                }
-            }
-
             let mut ticker = tokio::time::interval(Duration::from_secs(15));
             ticker.tick().await;
             loop {
@@ -256,6 +399,26 @@ impl Operation for TraceHandler {
                             break;
                         }
                     }
+                    received = subscription.recv() => {
+                        match received {
+                            Ok(event) => {
+                                if !filter.matches_kind(event.kind) {
+                                    continue;
+                                }
+                                let record = TraceWireRecord::from_event(&node_name, &event);
+                                if filter.matches_record(&record) && send_trace_record(&tx, &record).await.is_err() {
+                                    break;
+                                }
+                            }
+                            Err(tokio::sync::broadcast::error::RecvError::Lagged(dropped)) => {
+                                let record = TraceWireRecord::dropped(&node_name, dropped);
+                                if send_trace_record(&tx, &record).await.is_err() {
+                                    break;
+                                }
+                            }
+                            Err(tokio::sync::broadcast::error::RecvError::Closed) => break,
+                        }
+                    }
                 }
             }
         });
@@ -269,17 +432,115 @@ impl Operation for TraceHandler {
     }
 }
 
+async fn send_trace_record(tx: &mpsc::Sender<Result<Bytes, StdError>>, record: &TraceWireRecord) -> Result<(), ()> {
+    let Some(encoded) = encode_ndjson(record) else {
+        return Ok(());
+    };
+    tx.send(Ok(encoded)).await.map_err(|_| ())
+}
+
+fn encode_ndjson(value: &impl Serialize) -> Option<Bytes> {
+    let mut encoded = serde_json::to_vec(value).ok()?;
+    encoded.push(b'\n');
+    Some(Bytes::from(encoded))
+}
+
+fn trace_query_pairs(uri: &hyper::Uri) -> impl Iterator<Item = (String, String)> + '_ {
+    uri.query()
+        .into_iter()
+        .flat_map(|query| form_urlencoded::parse(query.as_bytes()))
+        .map(|(key, value)| (key.into_owned(), value.into_owned()))
+}
+
+fn trace_query_flag(uri: &hyper::Uri, flag: &str) -> bool {
+    trace_query_pairs(uri).any(|(key, value)| key == flag && value == "true")
+}
+
+fn trace_regex_filter(uri: &hyper::Uri) -> S3Result<Option<Regex>> {
+    trace_query_pairs(uri)
+        .find_map(|(key, value)| {
+            if key == "filter" && !value.is_empty() {
+                Some(value)
+            } else {
+                None
+            }
+        })
+        .map(|pattern| Regex::new(&pattern).map_err(|_| s3_error!(InvalidRequest, "invalid trace filter")))
+        .transpose()
+}
+
+fn trace_type_mask(kind: TraceKind) -> u64 {
+    match kind {
+        TraceKind::Heal => TraceType::HEALING.mask(),
+        TraceKind::Scanner => TraceType::SCANNER.mask(),
+    }
+}
+
+fn trace_time_string(time: SystemTime) -> String {
+    match OffsetDateTime::from(time).format(&Rfc3339) {
+        Ok(value) => value,
+        Err(_) => "1970-01-01T00:00:00Z".to_string(),
+    }
+}
+
+fn trace_path(event: &TraceEvent) -> String {
+    match (event.bucket.as_deref(), event.object.as_deref()) {
+        (Some(bucket), Some(object)) if !object.is_empty() => format!("{bucket}/{object}"),
+        (Some(bucket), _) => bucket.to_owned(),
+        (None, Some(object)) => object.to_owned(),
+        (None, None) => String::new(),
+    }
+}
+
+fn trace_bytes(bytes: u64) -> Option<i64> {
+    if bytes == 0 {
+        return None;
+    }
+
+    match i64::try_from(bytes) {
+        Ok(value) => Some(value),
+        Err(_) => Some(i64::MAX),
+    }
+}
+
+fn trace_custom_attrs(event: &TraceEvent) -> Option<HashMap<String, String>> {
+    if event.attrs.is_empty() {
+        return None;
+    }
+
+    Some(
+        event
+            .attrs
+            .iter()
+            .map(|attr| (attr.key.to_string(), trace_value_string(&attr.value)))
+            .collect(),
+    )
+}
+
+fn trace_value_string(value: &TraceVal) -> String {
+    match value {
+        TraceVal::Bool(value) => value.to_string(),
+        TraceVal::U64(value) => value.to_string(),
+        TraceVal::I64(value) => value.to_string(),
+        TraceVal::Str(value) => value.to_string(),
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::{
         ProfileControlHandler, ProfileHandler, ProfileStatusHandler, ProfilingDownloadHandler, ProfilingStartHandler,
-        TraceHandler,
+        TraceHandler, TraceKindFilter, TraceStreamFilter, TraceWireRecord,
     };
     use crate::admin::router::Operation;
     use http::{Extensions, HeaderMap, Uri};
     use hyper::Method;
     use matchit::Params;
-    use s3s::{Body, S3ErrorCode, S3Request};
+    use rustfs_common::trace_bus::{TraceEvent, TraceFunc, TraceKind};
+    use rustfs_madmin::service_commands::ServiceTraceOpts;
+    use rustfs_madmin::trace::TraceType;
+    use s3s::{Body, S3ErrorCode, S3Request, S3Result};
+    use std::time::{Duration, UNIX_EPOCH};
 
     fn build_profile_request(uri: &'static str) -> S3Request<Body> {
         S3Request {
@@ -295,6 +556,13 @@ mod tests {
         }
     }
 
+    fn build_trace_stream_filter(uri: &'static str) -> S3Result<TraceStreamFilter> {
+        let uri = Uri::from_static(uri);
+        let mut opts = ServiceTraceOpts::default();
+        opts.parse_params(&uri).expect("test trace params should parse");
+        TraceStreamFilter::from_request(&uri, &opts)
+    }
+
     #[tokio::test]
     async fn profile_handler_rejects_missing_credentials() {
         let result = ProfileHandler {}
@@ -358,4 +626,108 @@ mod tests {
             .expect_err("trace must reject anonymous requests");
         assert_eq!(err.code(), &S3ErrorCode::AccessDenied);
     }
+
+    #[test]
+    fn trace_kind_filter_supports_kind_query() {
+        let uri = Uri::from_static("/rustfs/admin/v3/trace?kind=heal");
+        let filter = TraceKindFilter::from_request(&uri, TraceType::default()).expect("kind filter should parse");
+
+        assert!(filter.matches(TraceKind::Heal));
+        assert!(!filter.matches(TraceKind::Scanner));
+    }
+
+    #[test]
+    fn trace_kind_filter_defaults_to_supported_events_without_type_flags() {
+        let uri = Uri::from_static("/rustfs/admin/v3/trace");
+        let filter = TraceKindFilter::from_request(&uri, TraceType::default()).expect("empty filter should parse");
+
+        assert!(filter.matches(TraceKind::Heal));
+        assert!(filter.matches(TraceKind::Scanner));
+    }
+
+    #[test]
+    fn trace_kind_filter_rejects_unknown_kind() {
+        let uri = Uri::from_static("/rustfs/admin/v3/trace?kind=s3");
+        let err = TraceKindFilter::from_request(&uri, TraceType::default()).expect_err("unknown kind should fail");
+
+        assert_eq!(err.code(), &S3ErrorCode::InvalidRequest);
+    }
+
+    #[test]
+    fn trace_stream_filter_matches_regex_against_path_and_attrs() {
+        let filter = build_trace_stream_filter("/rustfs/admin/v3/trace?kind=heal&filter=data/.%2Bxl.meta")
+            .expect("regex filter should parse");
+        let event = TraceEvent::new(TraceKind::Heal, TraceFunc::HealObject)
+            .with_bucket("data")
+            .with_object("dir/xl.meta")
+            .with_attr("dry_run", true);
+        let record = TraceWireRecord::from_event("node-a", &event);
+
+        assert!(filter.matches_kind(event.kind));
+        assert!(filter.matches_record(&record));
+    }
+
+    #[test]
+    fn trace_stream_filter_rejects_invalid_regex() {
+        let err = build_trace_stream_filter("/rustfs/admin/v3/trace?kind=heal&filter=[").expect_err("invalid regex should fail");
+
+        assert_eq!(err.code(), &S3ErrorCode::InvalidRequest);
+    }
+
+    #[test]
+    fn trace_stream_filter_applies_threshold() {
+        let filter =
+            build_trace_stream_filter("/rustfs/admin/v3/trace?kind=heal&threshold=10ms").expect("threshold should parse");
+        let short = TraceWireRecord::from_event(
+            "node-a",
+            &TraceEvent::new(TraceKind::Heal, TraceFunc::HealObject).with_duration(Duration::from_millis(9)),
+        );
+        let long = TraceWireRecord::from_event(
+            "node-a",
+            &TraceEvent::new(TraceKind::Heal, TraceFunc::HealObject).with_duration(Duration::from_millis(10)),
+        );
+
+        assert!(!filter.matches_record(&short));
+        assert!(filter.matches_record(&long));
+    }
+
+    #[test]
+    fn trace_stream_filter_rejects_error_only_filter() {
+        let err = build_trace_stream_filter("/rustfs/admin/v3/trace?kind=heal&err=true").expect_err("err filter should fail");
+
+        assert_eq!(err.code(), &S3ErrorCode::InvalidRequest);
+    }
+
+    #[test]
+    fn trace_wire_record_contains_madmin_trace_fields() {
+        let event = TraceEvent::new(TraceKind::Heal, TraceFunc::HealObject)
+            .with_bucket("bucket")
+            .with_object("object")
+            .with_duration(Duration::from_millis(3))
+            .with_bytes(17)
+            .with_attr("dry", true);
+        let mut record = TraceWireRecord::from_event("node-a", &event);
+        record.time = "1970-01-01T00:00:00Z".to_string();
+
+        let value = serde_json::to_value(&record).expect("trace record should serialize");
+
+        assert_eq!(value["type"], TraceType::HEALING.mask());
+        assert_eq!(value["nodename"], "node-a");
+        assert_eq!(value["funcname"], "heal.Object");
+        assert_eq!(value["time"], "1970-01-01T00:00:00Z");
+        assert_eq!(value["path"], "bucket/object");
+        assert_eq!(value["bytes"], 17);
+        assert_eq!(value["custom"]["dry"], "true");
+    }
+
+    #[test]
+    fn trace_wire_record_formats_epoch_time() {
+        let event = TraceEvent {
+            time: UNIX_EPOCH,
+            ..TraceEvent::new(TraceKind::Scanner, TraceFunc::ScannerFolder)
+        };
+        let record = TraceWireRecord::from_event("node-a", &event);
+
+        assert_eq!(record.time, "1970-01-01T00:00:00Z");
+    }
 }

rustfs/src/storage/rpc/node_service.rs

@@ -2420,6 +2420,7 @@ mod tests {
             _bucket: &str,
             _prefix: &str,
             _continuation_token: Option<&str>,
+            _include_lifecycle_object_info: bool,
         ) -> rustfs_heal::Result<(Vec<rustfs_heal::heal::storage::HealListItem>, Option<String>, bool)> {
             Ok((Vec::new(), None, false))
         }