fix(heal): coalesce duplicate MRF intents

.config/e2e-full-selection.txt

@@ -1,2 +1,2 @@
-sha256-darwin=52a05fdfae8bcf6f5828cc2b1e91b2a268139d3f7e1fc47d7b875b55fffb3995
-sha256-linux=a22d8af72e250595ac4445e8c880f3f9706202e09ed196e5b7baac632dead8d8
+sha256-darwin=9f767b37ed8b1c82da62ea441462d75487785c8086e56f08fb6f6cd89c6e2e52
+sha256-linux=fbdaf42b220958d4b1e8880e0f8b5a7992d38e21051bb60596dd4538424757d6

.config/nextest.toml

@@ -107,10 +107,9 @@ filter = 'package(e2e_test) & test(/^inline_fast_path_cluster_test::/)'
 test-group = 'e2e-inline-boundaries'
 
 # Vault KMS tests share the fixed dev-server port 8200. serial_test's #[serial]
-# does not cross nextest process boundaries, so keep every Vault-backed test in
-# one group.
+# does not cross nextest process boundaries, so keep these tests in one group.
 [[profile.default.overrides]]
-filter = 'package(e2e_test) & (test(/^kms::kms_vault_test::/) | test(/^kms::configured_roundtrip_test::test_configured_vault_kms_admin_and_versioned_cleanup$/))'
+filter = 'package(e2e_test) & test(/^kms::kms_vault_test::/)'
 test-group = 'e2e-vault'
 
 # ---------------------------------------------------------------------------
@@ -444,5 +443,5 @@ filter = 'package(e2e_test) & test(/^inline_fast_path_cluster_test::/)'
 test-group = 'e2e-inline-boundaries'
 
 [[profile.e2e-full.overrides]]
-filter = 'package(e2e_test) & (test(/^kms::kms_vault_test::/) | test(/^kms::configured_roundtrip_test::test_configured_vault_kms_admin_and_versioned_cleanup$/))'
+filter = 'package(e2e_test) & test(/^kms::kms_vault_test::/)'
 test-group = 'e2e-vault'

Cargo.lock

@@ -1858,9 +1858,9 @@ dependencies = [
 
 [[package]]
 name = "cc"
-version = "1.4.4"
+version = "1.4.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0ad534f4357a5264cce5019c989cf66a4f0dc4e0d1b1d15f8aacec0ff7360273"
+checksum = "509591b7bcd67f4ef775afad7662703b4935daaa6ec0e5605cfb1090b32a2b6d"
 dependencies = [
  "find-msvc-tools",
  "jobserver",
@@ -2522,6 +2522,12 @@ dependencies = [
  "subtle",
 ]
 
+[[package]]
+name = "cty"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b365fabc795046672053e29c954733ec3b05e4be654ab130fe8f1f94d7051f35"
+
 [[package]]
 name = "curve25519-dalek"
 version = "4.1.3"
@@ -5982,6 +5988,15 @@ version = "0.2.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b6d2cec3eae94f9f509c767b45932f1ada8350c4bdb85af2fcab4a3c14807981"
 
+[[package]]
+name = "libmimalloc-sys"
+version = "0.1.49"
+source = "git+https://github.com/xonatius/mimalloc_rust.git?rev=6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11#6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11"
+dependencies = [
+ "cc",
+ "cty",
+]
+
 [[package]]
 name = "libredox"
 version = "0.1.20"
@@ -6382,6 +6397,14 @@ dependencies = [
  "synstructure 0.13.2",
 ]
 
+[[package]]
+name = "mimalloc"
+version = "0.1.52"
+source = "git+https://github.com/xonatius/mimalloc_rust.git?rev=6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11#6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11"
+dependencies = [
+ "libmimalloc-sys",
+]
+
 [[package]]
 name = "mime"
 version = "0.3.17"
@@ -9139,11 +9162,13 @@ dependencies = [
  "insta",
  "jiff",
  "libc",
+ "libmimalloc-sys",
  "libsystemd",
  "matchit 0.9.2",
  "md-5 0.11.0",
  "metrics",
  "metrics-util",
+ "mimalloc",
  "mime_guess",
  "opentelemetry",
  "opentelemetry_sdk",
@@ -9179,8 +9204,6 @@ dependencies = [
  "rustfs-lock",
  "rustfs-log-analyzer",
  "rustfs-madmin",
- "rustfs-mimalloc",
- "rustfs-mimalloc-sys",
  "rustfs-notify",
  "rustfs-object-capacity",
  "rustfs-object-data-cache",
@@ -9852,24 +9875,6 @@ dependencies = [
  "tokio",
 ]
 
-[[package]]
-name = "rustfs-mimalloc"
-version = "0.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a406f4aa07084301d485beec873af6dccc8e3f8762da244743df92038b1db1a6"
-dependencies = [
- "rustfs-mimalloc-sys",
-]
-
-[[package]]
-name = "rustfs-mimalloc-sys"
-version = "0.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c3051b819175f58445d4c369a72f0ab88149f3885ba8bea2aff3be01f53fe7cd"
-dependencies = [
- "cc",
-]
-
 [[package]]
 name = "rustfs-notify"
 version = "1.0.0-rc.3"

Cargo.toml

@@ -350,8 +350,8 @@ russh-sftp = "2.4.0"
 dav-server = "0.11.0"
 
 # Performance Analysis and Memory Profiling
-rustfs-mimalloc = { version = "0.5.0" }
-rustfs-mimalloc-sys = { version = "0.5.0" }
+mimalloc = { version = "0.1.52", git = "https://github.com/xonatius/mimalloc_rust.git", rev = "6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11" }
+libmimalloc-sys = { version = "0.1.49", git = "https://github.com/xonatius/mimalloc_rust.git", rev = "6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11", features = ["extended"] }
 hotpath = { version = "0.23.3", default-features = false }
 # Snapshot testing for output format regression detection
 insta = { version = "1.48" }

crates/common/src/mrf_channel.rs

@@ -23,21 +23,32 @@
 //! unconsumed intents is the consumer's job (see `rustfs-heal`
 //! `heal::mrf_queue`), mirroring MinIO's `.heal/mrf/list.bin`.
 
+use std::collections::HashMap;
+use std::collections::hash_map::RandomState;
+use std::hash::{BuildHasher, Hash};
+use std::sync::atomic::AtomicU64;
+use std::sync::atomic::AtomicUsize;
 use std::sync::{
-    Arc, OnceLock,
+    Arc, Mutex, OnceLock,
     atomic::{AtomicBool, Ordering},
 };
+use std::time::{Duration, Instant};
 use tokio::sync::mpsc;
 use uuid::Uuid;
 
 /// Bounded capacity of the global MRF channel. Backpressure is resolved by
 /// dropping (and counting) intents, never by blocking the producer.
 const MRF_CHANNEL_CAPACITY: usize = 8192;
+const MRF_COALESCER_SHARDS: usize = 16;
+const MRF_COALESCER_MAX_KEYS: usize = 8192;
+const MRF_COALESCER_MAX_BYTES: usize = 16 * 1024 * 1024;
+const MRF_COALESCER_TTL: Duration = Duration::from_secs(60);
+const MRF_MAX_IDENTITY_COMPONENT: usize = 1024;
 
 /// Why an intent was produced. Drives the heal priority mapping on the
 /// consumer side (DecodeFailure -> Urgent, MetadataCorruption -> High,
 /// PartialWrite -> Normal).
-#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
 pub enum MrfKind {
     /// Erasure decode failed while serving a read (read path).
     DecodeFailure,
@@ -67,12 +78,52 @@ pub struct MrfIntent {
     /// Version the intent targets, as raw UUID bytes.
     pub version_id: Option<[u8; 16]>,
     pub kind: MrfKind,
+    /// Stable erasure-set scope when the producer has it. Kept optional so
+    /// metadata corruption and legacy producers do not invent a scope.
+    pub scope: Option<MrfScope>,
+    /// Generation of the node-local ingress lease. It is not persisted in
+    /// the journal; replayed records acquire a fresh lease when re-enqueued.
+    pub lease: Option<MrfIngressLease>,
     pub enqueued_at_ms: u64,
     /// Times this intent has already been offered to the heal manager.
     /// Dropped by the consumer once it reaches `MRF_MAX_ATTEMPTS`.
     pub attempts: u8,
 }
 
+#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
+pub struct MrfScope {
+    pub pool_index: u32,
+    pub set_index: u32,
+}
+
+/// Opaque generation used to release exactly the admission that created an
+/// ingress entry. A generation prevents a late terminal callback from
+/// deleting a newer retry for the same identity (ABA).
+#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
+pub struct MrfIngressLease(u64);
+
+impl MrfIngressLease {
+    const fn new(value: u64) -> Self {
+        Self(value)
+    }
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub enum MrfDropReason {
+    Disabled,
+    Uninitialized,
+    Full,
+    OversizedIdentity,
+    CoalescerFull,
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub enum MrfIngressResult {
+    Enqueued,
+    Coalesced,
+    Dropped(MrfDropReason),
+}
+
 /// Consumer-side retry ceiling before an intent is given up on.
 pub const MRF_MAX_ATTEMPTS: u8 = 3;
 
@@ -87,6 +138,159 @@ impl MrfIntent {
 
 static GLOBAL_MRF_SENDER: OnceLock<mpsc::Sender<MrfIntent>> = OnceLock::new();
 
+#[derive(Clone, Debug, PartialEq, Eq, Hash)]
+struct MrfIdentityKey {
+    kind: MrfKind,
+    bucket: Arc<str>,
+    object: Arc<str>,
+    version_id: Option<[u8; 16]>,
+    scope: Option<MrfScope>,
+}
+
+#[derive(Debug)]
+struct IngressEntry {
+    lease: MrfIngressLease,
+    expires_at: Instant,
+    bytes: usize,
+}
+
+type MrfCoalescerShard = Mutex<HashMap<MrfIdentityKey, IngressEntry>>;
+type MrfCoalescer = Box<[MrfCoalescerShard]>;
+
+static MRF_COALESCER: OnceLock<MrfCoalescer> = OnceLock::new();
+static NEXT_MRF_LEASE: AtomicU64 = AtomicU64::new(1);
+static MRF_COALESCER_COUNT: AtomicUsize = AtomicUsize::new(0);
+static MRF_COALESCER_BYTES: AtomicUsize = AtomicUsize::new(0);
+static MRF_HASH_STATE: OnceLock<RandomState> = OnceLock::new();
+
+fn coalescer() -> &'static [MrfCoalescerShard] {
+    MRF_COALESCER.get_or_init(|| {
+        (0..MRF_COALESCER_SHARDS)
+            .map(|_| Mutex::new(HashMap::new()))
+            .collect::<Vec<_>>()
+            .into_boxed_slice()
+    })
+}
+
+fn key_shard(key: &MrfIdentityKey) -> usize {
+    let hash = MRF_HASH_STATE.get_or_init(RandomState::new).hash_one(key);
+    usize::try_from(hash).unwrap_or(0) % MRF_COALESCER_SHARDS
+}
+
+fn canonical_version(version_id: Option<Uuid>) -> Option<[u8; 16]> {
+    version_id
+        .filter(|version| !version.is_nil())
+        .map(|version| *version.as_bytes())
+}
+
+fn canonical_identity(
+    kind: MrfKind,
+    version_id: Option<[u8; 16]>,
+    scope: Option<MrfScope>,
+) -> (Option<[u8; 16]>, Option<MrfScope>) {
+    let version_id = version_id.filter(|bytes| *bytes != [0; 16]);
+    match kind {
+        MrfKind::MetadataCorruption => (None, None),
+        MrfKind::DecodeFailure | MrfKind::PartialWrite => (version_id, scope),
+    }
+}
+
+fn identity_estimated_bytes(key: &MrfIdentityKey) -> usize {
+    64usize
+        .saturating_add(key.bucket.len())
+        .saturating_add(key.object.len())
+        .saturating_add(key.version_id.map_or(0, |_| 16))
+        .saturating_add(key.scope.map_or(0, |_| 8))
+}
+
+fn reserve(counter: &AtomicUsize, limit: usize, amount: usize) -> bool {
+    let mut current = counter.load(Ordering::Relaxed);
+    loop {
+        let Some(next) = current.checked_add(amount) else {
+            return false;
+        };
+        if next > limit {
+            return false;
+        }
+        match counter.compare_exchange_weak(current, next, Ordering::Relaxed, Ordering::Relaxed) {
+            Ok(_) => return true,
+            Err(observed) => current = observed,
+        }
+    }
+}
+
+fn coalescer_admit(key: MrfIdentityKey) -> Result<MrfIngressLease, MrfIngressResult> {
+    let shard = key_shard(&key);
+    let mut entries = coalescer()[shard]
+        .lock()
+        .map_err(|_| MrfIngressResult::Dropped(MrfDropReason::CoalescerFull))?;
+    let now = Instant::now();
+    let before = entries.len();
+    let mut expired_bytes = 0usize;
+    entries.retain(|_, entry| {
+        if entry.expires_at > now {
+            true
+        } else {
+            expired_bytes = expired_bytes.saturating_add(entry.bytes);
+            false
+        }
+    });
+    let evicted = before.saturating_sub(entries.len());
+    if evicted > 0 {
+        MRF_COALESCER_COUNT.fetch_sub(evicted, Ordering::Relaxed);
+        MRF_COALESCER_BYTES.fetch_sub(expired_bytes, Ordering::Relaxed);
+        let evicted = u64::try_from(evicted).unwrap_or(u64::MAX);
+        metrics::counter!("rustfs_heal_mrf_coalescer_expired_total").increment(evicted);
+        metrics::counter!("rustfs_heal_mrf_coalescer_evictions_total").increment(evicted);
+    }
+    if entries.contains_key(&key) {
+        metrics::counter!("rustfs_heal_mrf_coalesced_total").increment(1);
+        return Err(MrfIngressResult::Coalesced);
+    }
+    let bytes = identity_estimated_bytes(&key);
+    let count_reserved = reserve(&MRF_COALESCER_COUNT, MRF_COALESCER_MAX_KEYS, 1);
+    let bytes_reserved = count_reserved && reserve(&MRF_COALESCER_BYTES, MRF_COALESCER_MAX_BYTES, bytes);
+    if !count_reserved || !bytes_reserved {
+        if count_reserved {
+            MRF_COALESCER_COUNT.fetch_sub(1, Ordering::Relaxed);
+        }
+        metrics::counter!("rustfs_heal_mrf_dropped_total", "reason" => "coalescer_full").increment(1);
+        return Err(MrfIngressResult::Dropped(MrfDropReason::CoalescerFull));
+    }
+    let lease = MrfIngressLease::new(NEXT_MRF_LEASE.fetch_add(1, Ordering::Relaxed));
+    if entries
+        .insert(
+            key,
+            IngressEntry {
+                lease,
+                expires_at: now + MRF_COALESCER_TTL,
+                bytes,
+            },
+        )
+        .is_some()
+    {
+        MRF_COALESCER_COUNT.fetch_sub(1, Ordering::Relaxed);
+        MRF_COALESCER_BYTES.fetch_sub(bytes, Ordering::Relaxed);
+        metrics::counter!("rustfs_heal_mrf_coalesced_total").increment(1);
+        return Err(MrfIngressResult::Coalesced);
+    }
+    Ok(lease)
+}
+
+fn coalescer_release(key: &MrfIdentityKey, lease: Option<MrfIngressLease>) {
+    let Some(lease) = lease else {
+        return;
+    };
+    if let Ok(mut entries) = coalescer()[key_shard(key)].lock() {
+        let should_remove = entries.get(key).is_some_and(|entry| entry.lease == lease);
+        if should_remove {
+            let bytes = entries.remove(key).map(|entry| entry.bytes).unwrap_or(0);
+            MRF_COALESCER_COUNT.fetch_sub(1, Ordering::Relaxed);
+            MRF_COALESCER_BYTES.fetch_sub(bytes, Ordering::Relaxed);
+        }
+    }
+}
+
 /// Delivery kill-switch, set from `RUSTFS_HEAL_MRF_ENABLE`. Producers check
 /// this before touching the channel so the disabled path stays allocation- and
 /// sync-free.
@@ -122,21 +326,90 @@ pub fn init_mrf_channel() -> Result<mpsc::Receiver<MrfIntent>, &'static str> {
 /// This runs on IO error paths, so it stays synchronous and cheap: one
 /// bounded allocation for the two `Arc<str>` handles plus the channel slot.
 pub fn try_send_mrf_intent(kind: MrfKind, bucket: &str, object: &str, version_id: Option<Uuid>) -> bool {
+    matches!(
+        try_send_mrf_intent_typed(kind, bucket, object, version_id, None),
+        MrfIngressResult::Enqueued
+    )
+}
+
+/// Typed ingress result. `Coalesced` means an equivalent in-flight channel
+/// intent already exists; it is not a second executable or durable admission.
+pub fn try_send_mrf_intent_typed(
+    kind: MrfKind,
+    bucket: &str,
+    object: &str,
+    version_id: Option<Uuid>,
+    scope: Option<MrfScope>,
+) -> MrfIngressResult {
     if !mrf_delivery_enabled() {
-        return false;
+        return MrfIngressResult::Dropped(MrfDropReason::Disabled);
     }
     let Some(sender) = GLOBAL_MRF_SENDER.get() else {
-        return false;
+        return MrfIngressResult::Dropped(MrfDropReason::Uninitialized);
     };
-    let intent = MrfIntent {
+    if bucket.len() > MRF_MAX_IDENTITY_COMPONENT || object.len() > MRF_MAX_IDENTITY_COMPONENT {
+        return MrfIngressResult::Dropped(MrfDropReason::OversizedIdentity);
+    }
+    let (version_id, scope) = canonical_identity(kind, canonical_version(version_id), scope);
+    let key = MrfIdentityKey {
+        kind,
         bucket: Arc::from(bucket),
         object: Arc::from(object),
-        version_id: version_id.map(|vid| *vid.as_bytes()),
+        version_id,
+        scope,
+    };
+    let lease = match coalescer_admit(key.clone()) {
+        Ok(lease) => lease,
+        Err(result) => return result,
+    };
+    let intent = MrfIntent {
+        bucket: key.bucket.clone(),
+        object: key.object.clone(),
+        version_id: key.version_id,
         kind,
+        scope,
+        lease: Some(lease),
         enqueued_at_ms: unix_now_ms(),
         attempts: 0,
     };
-    sender.try_send(intent).is_ok()
+    match sender.try_send(intent) {
+        Ok(()) => MrfIngressResult::Enqueued,
+        Err(mpsc::error::TrySendError::Full(_)) => {
+            coalescer_release(&key, Some(lease));
+            metrics::counter!("rustfs_heal_mrf_dropped_total", "reason" => "channel_full").increment(1);
+            MrfIngressResult::Dropped(MrfDropReason::Full)
+        }
+        Err(mpsc::error::TrySendError::Closed(_)) => {
+            coalescer_release(&key, Some(lease));
+            MrfIngressResult::Dropped(MrfDropReason::Uninitialized)
+        }
+    }
+}
+
+/// Release the ingress key once the consumer owns the intent.
+pub fn release_mrf_intent(intent: &MrfIntent) {
+    release_mrf_identity(intent.kind, &intent.bucket, &intent.object, intent.version_id, intent.scope, intent.lease);
+}
+
+pub fn release_mrf_identity(
+    kind: MrfKind,
+    bucket: &str,
+    object: &str,
+    version_id: Option<[u8; 16]>,
+    scope: Option<MrfScope>,
+    lease: Option<MrfIngressLease>,
+) {
+    let (version_id, scope) = canonical_identity(kind, version_id, scope);
+    coalescer_release(
+        &MrfIdentityKey {
+            kind,
+            bucket: Arc::from(bucket),
+            object: Arc::from(object),
+            version_id,
+            scope,
+        },
+        lease,
+    );
 }
 
 fn unix_now_ms() -> u64 {
@@ -144,7 +417,8 @@ fn unix_now_ms() -> u64 {
     // failure would be a bug rather than something to handle here.
     std::time::SystemTime::now()
         .duration_since(std::time::UNIX_EPOCH)
-        .map(|d| d.as_millis() as u64)
+        .ok()
+        .and_then(|d| u64::try_from(d.as_millis()).ok())
         .unwrap_or(0)
 }
 
@@ -215,12 +489,60 @@ mod tests {
             object: Arc::from("object"),
             version_id: Some([0u8; 16]),
             kind: MrfKind::DecodeFailure,
+            scope: None,
+            lease: None,
             enqueued_at_ms: 0,
             attempts: 0,
         };
         assert!(intent.estimated_bytes() >= intent.bucket.len() + intent.object.len());
     }
 
+    #[test]
+    fn ingress_duplicate_identity_coalesces_and_releases_for_retry() {
+        let key = MrfIdentityKey {
+            kind: MrfKind::DecodeFailure,
+            bucket: Arc::from("ingress-test-bucket"),
+            object: Arc::from("ingress-test-object"),
+            version_id: Some([9; 16]),
+            scope: Some(MrfScope {
+                pool_index: 3,
+                set_index: 4,
+            }),
+        };
+        let lease = coalescer_admit(key.clone()).expect("first identity should be admitted");
+        for _ in 0..999 {
+            assert_eq!(coalescer_admit(key.clone()), Err(MrfIngressResult::Coalesced));
+        }
+        coalescer_release(&key, Some(lease));
+        let retry_lease = coalescer_admit(key.clone()).expect("released identity must admit a retry");
+        coalescer_release(&key, Some(retry_lease));
+    }
+
+    #[test]
+    fn ingress_identity_preserves_kind_scope_and_version_boundaries() {
+        let (nil_version, nil_scope) = canonical_identity(
+            MrfKind::DecodeFailure,
+            Some([0; 16]),
+            Some(MrfScope {
+                pool_index: 1,
+                set_index: 2,
+            }),
+        );
+        assert_eq!(nil_version, None, "nil UUID is the unversioned identity");
+        assert!(nil_scope.is_some());
+
+        let (metadata_version, metadata_scope) = canonical_identity(
+            MrfKind::MetadataCorruption,
+            Some([7; 16]),
+            Some(MrfScope {
+                pool_index: 1,
+                set_index: 2,
+            }),
+        );
+        assert_eq!(metadata_version, None);
+        assert_eq!(metadata_scope, None);
+    }
+
     #[tokio::test]
     async fn try_send_delivers_and_respects_capacity() {
         let mut receiver = init_mrf_channel().expect("first initialization should succeed");
@@ -230,6 +552,7 @@ mod tests {
         let intent = receiver.recv().await.expect("intent should arrive");
         assert_eq!(intent.kind, MrfKind::DecodeFailure);
         assert_eq!(intent.bucket.as_ref(), "b");
+        release_mrf_intent(&intent);
 
         // Disable delivery: producers become no-ops.
         set_mrf_delivery_enabled(false);
@@ -239,8 +562,8 @@ mod tests {
         // Fill the bounded channel past capacity: excess intents are dropped,
         // never blocking.
         let mut accepted = 0;
-        for _ in 0..(MRF_CHANNEL_CAPACITY + 64) {
-            if try_send_mrf_intent(MrfKind::PartialWrite, "b", "o", None) {
+        for index in 0..(MRF_CHANNEL_CAPACITY + 64) {
+            if try_send_mrf_intent(MrfKind::PartialWrite, "b", &format!("o-{index}"), None) {
                 accepted += 1;
             }
         }

crates/e2e_test/src/kms/configured_roundtrip_test.rs

@@ -432,6 +432,7 @@ async fn test_configured_local_kms_admin_and_versioned_cleanup() -> TestResult {
 }
 
 #[tokio::test]
+#[ignore = "requires a Vault binary"]
 async fn test_configured_vault_kms_admin_and_versioned_cleanup() -> TestResult {
     let mut env = VaultTestEnvironment::new().await?;
     env.start_vault().await?;

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

@@ -1412,8 +1412,11 @@ pub(in crate::set_disk) async fn submit_read_repair_heal_with_submitter(
 
     // Reservation won: this sighting owns the repair records for the object,
     // including the durable journal intent when the caller asked for one.
-    if let Some((kind, version_uuid)) = mrf_intent {
-        rustfs_common::mrf_channel::try_send_mrf_intent(kind, bucket, object, version_uuid);
+    if let Some((kind, version_uuid)) = mrf_intent
+        && let (Ok(pool_index), Ok(set_index)) = (u32::try_from(pool_index), u32::try_from(set_index))
+    {
+        let scope = rustfs_common::mrf_channel::MrfScope { pool_index, set_index };
+        let _ = rustfs_common::mrf_channel::try_send_mrf_intent_typed(kind, bucket, object, version_uuid, Some(scope));
     }
 
     let mut request = rustfs_common::heal_channel::create_heal_request_with_options(

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

@@ -2124,13 +2124,26 @@ impl SetDisks {
 
             let put_object_size = known_put_object_storage_size(data.size());
             let shard_file_size_raw = erasure.shard_file_size(put_object_size);
-            let is_inline_buffer = storage_class_config.should_inline(shard_file_size_raw, erasure.data_shards, opts.versioned);
+            let is_inline_buffer =
+                storage_class_config.should_inline(shard_file_size_raw, erasure.data_shards, opts.versioned);
 
             let collect_stage_timing = rustfs_io_metrics::put_stage_metrics_enabled() || issue3031_diag_enabled();
             let shard_file_size = shard_file_size_raw;
             let shard_size = erasure.shard_size();
             let write_path = classify_put_write_path(is_inline_buffer, put_object_size, fi.erasure.block_size);
             let direct_inline_commit = matches!(write_path, SmallWritePath::Inline);
+            {
+                use std::io::Write;
+                let msg = format!(
+                    "INLINE_DEBUG: bucket={} obj={} size={} shard_fs={} ds={} bs={} inline={} direct={} path={} iblock={} ver={}\n",
+                    bucket, object, put_object_size, shard_file_size_raw, erasure.data_shards, fi.erasure.block_size,
+                    is_inline_buffer, direct_inline_commit, write_path.metric_label(), storage_class_config.inline_block(), opts.versioned
+                );
+                if let Ok(mut f) = std::fs::OpenOptions::new().create(true).append(true).open("/tmp/rustfs_inline_debug.log") {
+                    let _ = f.write_all(msg.as_bytes());
+                }
+                let _ = std::io::stderr().write_all(msg.as_bytes());
+            }
             rustfs_io_metrics::record_put_object_path(write_path.metric_label());
             let writer_setup_stage_start = collect_stage_timing.then(Instant::now);
             let (mut writers, errors) = if direct_inline_commit {
@@ -6637,12 +6650,23 @@ impl crate::storage_api_contracts::object::ObjectOperations for SetDisks {
     async fn add_partial(&self, bucket: &str, object: &str, version_id: &str) -> Result<()> {
         // MRF journal intent: partial-write recovery must survive a restart
         // (HS-01); the heal request below remains the in-memory fast path.
-        rustfs_common::mrf_channel::try_send_mrf_intent(
-            rustfs_common::mrf_channel::MrfKind::PartialWrite,
-            bucket,
-            object,
-            uuid::Uuid::try_parse(version_id).ok(),
-        );
+        let version_uuid = if version_id.is_empty() {
+            Some(None)
+        } else {
+            uuid::Uuid::try_parse(version_id).ok().map(Some)
+        };
+        if let Some(version_uuid) = version_uuid
+            && let (Ok(pool_index), Ok(set_index)) = (u32::try_from(self.pool_index), u32::try_from(self.set_index))
+        {
+            let scope = rustfs_common::mrf_channel::MrfScope { pool_index, set_index };
+            let _ = rustfs_common::mrf_channel::try_send_mrf_intent_typed(
+                rustfs_common::mrf_channel::MrfKind::PartialWrite,
+                bucket,
+                object,
+                version_uuid,
+                Some(scope),
+            );
+        }
         let mut request = rustfs_common::heal_channel::create_heal_request_with_options(
             bucket.to_string(),
             Some(object.to_string()),

crates/ecstore/src/store/object.rs

@@ -3194,7 +3194,7 @@ impl ECStore {
 
         // Default return value
         let mut del_objects = vec![DeletedObject::default(); objects.len()];
-        let accounting = vec![None; objects.len()];
+        let mut accounting = vec![None; objects.len()];
 
         let mut del_errs = Vec::with_capacity(objects.len());
         for _ in 0..objects.len() {

crates/ecstore/src/store/rebalance/support.rs

@@ -271,7 +271,7 @@ pub(super) fn resolve_latest_object_info_candidates(
             .filter(|candidate| latest_candidate_mod_time(candidate) == Some(latest_mod_time))
             .collect::<Vec<_>>();
 
-        latest_candidates.sort_by_key(|candidate| std::cmp::Reverse(candidate.idx));
+        latest_candidates.sort_by(|left, right| right.idx.cmp(&left.idx));
 
         let Some(winner) = latest_candidates.first() else {
             return Err(Error::ErasureReadQuorum);

crates/heal/src/heal/manager.rs

@@ -119,6 +119,9 @@ struct MrfRepairNoticeTarget {
     bucket: Arc<str>,
     object: Arc<str>,
     version_id: Option<[u8; 16]>,
+    kind: rustfs_common::mrf_channel::MrfKind,
+    scope: Option<rustfs_common::mrf_channel::MrfScope>,
+    lease: Option<rustfs_common::mrf_channel::MrfIngressLease>,
 }
 
 #[derive(Debug, Clone)]
@@ -889,7 +892,19 @@ impl HealManager {
     }
 
     fn remove_mrf_repair_notice_targets_for_task(&self, task_id: &str) {
-        lock_mrf_repair_notice_targets(&self.mrf_repair_notice_targets).remove(task_id);
+        let targets = lock_mrf_repair_notice_targets(&self.mrf_repair_notice_targets).remove(task_id);
+        if let Some(targets) = targets {
+            for target in targets {
+                rustfs_common::mrf_channel::release_mrf_identity(
+                    target.kind,
+                    &target.bucket,
+                    &target.object,
+                    target.version_id,
+                    target.scope,
+                    target.lease,
+                );
+            }
+        }
     }
 
     fn insert_mrf_repair_notice_target(
@@ -1279,7 +1294,20 @@ impl HealManager {
         }
         self.task_aliases.lock().await.clear();
         self.retrying_heals.lock().await.clear();
-        lock_mrf_repair_notice_targets(&self.mrf_repair_notice_targets).clear();
+        let mrf_targets = {
+            let mut registry = lock_mrf_repair_notice_targets(&self.mrf_repair_notice_targets);
+            registry.drain().flat_map(|(_, targets)| targets).collect::<Vec<_>>()
+        };
+        for target in mrf_targets {
+            rustfs_common::mrf_channel::release_mrf_identity(
+                target.kind,
+                &target.bucket,
+                &target.object,
+                target.version_id,
+                target.scope,
+                target.lease,
+            );
+        }
         crate::set_heal_queue_length(0);
 
         // update state
@@ -1311,12 +1339,32 @@ impl HealManager {
             .await
     }
 
+    #[cfg(test)]
     pub(crate) async fn submit_mrf_heal_request_with_receipt(
         &self,
         request: HealRequest,
         bucket: Arc<str>,
         object: Arc<str>,
         version_id: Option<[u8; 16]>,
+    ) -> Result<HealAdmissionReceipt> {
+        let kind = match &request.heal_type {
+            HealType::Metadata { .. } => rustfs_common::mrf_channel::MrfKind::MetadataCorruption,
+            HealType::ECDecode { .. } => rustfs_common::mrf_channel::MrfKind::DecodeFailure,
+            _ => rustfs_common::mrf_channel::MrfKind::PartialWrite,
+        };
+        self.submit_mrf_heal_request_with_receipt_and_identity(request, bucket, object, version_id, kind, None, None)
+            .await
+    }
+
+    pub(crate) async fn submit_mrf_heal_request_with_receipt_and_identity(
+        &self,
+        request: HealRequest,
+        bucket: Arc<str>,
+        object: Arc<str>,
+        version_id: Option<[u8; 16]>,
+        kind: rustfs_common::mrf_channel::MrfKind,
+        scope: Option<rustfs_common::mrf_channel::MrfScope>,
+        lease: Option<rustfs_common::mrf_channel::MrfIngressLease>,
     ) -> Result<HealAdmissionReceipt> {
         self.submit_heal_request_with_receipt_alias_and_mrf_notice(
             request,
@@ -1325,6 +1373,9 @@ impl HealManager {
                 bucket,
                 object,
                 version_id,
+                kind,
+                scope,
+                lease,
             }),
         )
         .await
@@ -1539,7 +1590,7 @@ impl HealManager {
             Self::insert_mrf_repair_notice_target(&mut targets, &task_id, target);
         }
         if let Some(displaced_task_id) = &displaced_task_id {
-            lock_mrf_repair_notice_targets(&self.mrf_repair_notice_targets).remove(displaced_task_id);
+            self.remove_mrf_repair_notice_targets_for_task(displaced_task_id);
         }
         drop(retrying_heals);
         drop(queue);

crates/heal/src/heal/manager/auto_scan.rs

@@ -506,7 +506,18 @@ impl HealManager {
                                     &displaced_terminal,
                                 )
                                 .await;
-                                lock_mrf_repair_notice_targets(&mrf_repair_notice_targets).remove(&displaced_task_id);
+                                if let Some(targets) = lock_mrf_repair_notice_targets(&mrf_repair_notice_targets).remove(&displaced_task_id) {
+                                    for target in targets {
+                                        rustfs_common::mrf_channel::release_mrf_identity(
+                                            target.kind,
+                                            &target.bucket,
+                                            &target.object,
+                                            target.version_id,
+                                            target.scope,
+                                            target.lease,
+                                        );
+                                    }
+                                }
                             }
                             if matches!(admission, HealAdmissionResult::Accepted) {
                                 if should_notify {

crates/heal/src/heal/manager/queue.rs

@@ -299,7 +299,11 @@ impl PriorityHealQueue {
 
     /// Create a deduplication key from a heal request
     pub(super) fn make_dedup_key(request: &HealRequest) -> String {
-        Self::make_dedup_key_for_type(&request.heal_type)
+        let base = Self::make_dedup_key_for_type(&request.heal_type);
+        match (&request.heal_type, request.options.set_key()) {
+            (HealType::Object { .. } | HealType::ECDecode { .. }, Some(scope)) => format!("{base}:scope:{scope}"),
+            _ => base,
+        }
     }
 
     pub(super) fn make_dedup_key_for_type(heal_type: &HealType) -> String {

crates/heal/src/heal/manager/scheduler.rs

@@ -349,6 +349,8 @@ impl HealManager {
                             let notice_targets = take_mrf_repair_notice_targets(&mrf_repair_notice_targets_clone, &task_id);
                             if successful_completion {
                                 emit_mrf_repaired_events(notice_targets);
+                            } else {
+                                release_mrf_repair_notice_targets(notice_targets);
                             }
                             task_aliases_clone
                                 .lock()
@@ -638,7 +640,19 @@ pub(super) fn running_heal_set_counts(active_heals: &HashMap<String, Arc<HealTas
 }
 
 fn remove_mrf_repair_notice_targets(registry: &Arc<StdMutex<HashMap<String, Vec<MrfRepairNoticeTarget>>>>, task_id: &str) {
-    lock_mrf_repair_notice_targets(registry).remove(task_id);
+    let targets = lock_mrf_repair_notice_targets(registry).remove(task_id);
+    if let Some(targets) = targets {
+        for target in targets {
+            rustfs_common::mrf_channel::release_mrf_identity(
+                target.kind,
+                &target.bucket,
+                &target.object,
+                target.version_id,
+                target.scope,
+                target.lease,
+            );
+        }
+    }
 }
 
 fn take_mrf_repair_notice_targets(
@@ -671,6 +685,27 @@ fn move_mrf_repair_notice_targets(
 fn emit_mrf_repaired_events(targets: Vec<MrfRepairNoticeTarget>) {
     for target in targets {
         rustfs_common::mrf_channel::note_mrf_repaired(&target.bucket, &target.object, target.version_id);
+        rustfs_common::mrf_channel::release_mrf_identity(
+            target.kind,
+            &target.bucket,
+            &target.object,
+            target.version_id,
+            target.scope,
+            target.lease,
+        );
+    }
+}
+
+fn release_mrf_repair_notice_targets(targets: Vec<MrfRepairNoticeTarget>) {
+    for target in targets {
+        rustfs_common::mrf_channel::release_mrf_identity(
+            target.kind,
+            &target.bucket,
+            &target.object,
+            target.version_id,
+            target.scope,
+            target.lease,
+        );
     }
 }
 

crates/heal/src/heal/mrf_queue.rs

@@ -37,7 +37,7 @@ use crate::heal::manager::HealManager;
 use metrics::{counter, gauge};
 use rustfs_common::heal_channel::{HealAdmissionDropReason, HealAdmissionResult};
 use rustfs_common::mrf_channel::{MRF_MAX_ATTEMPTS, MrfIntent};
-use std::collections::VecDeque;
+use std::collections::{HashSet, VecDeque};
 use std::sync::Arc;
 use std::time::Duration;
 use tokio::sync::mpsc;
@@ -48,15 +48,27 @@ use crate::heal::task::{HealOptions, HealPriority, HealRequest, HealType};
 /// Journal location inside the metadata bucket, following the resume-state
 /// layout.
 pub(crate) const MRF_JOURNAL_PATH: &str = "buckets/.heal/mrf/journal.bin";
+/// The scoped path is the authoritative snapshot for new readers and carries
+/// both v1 and v2 records. The legacy path is only a v1 compatibility mirror;
+/// older readers ignore the authoritative path, while new readers never merge
+/// the two files. This prevents a partial two-file flush from fabricating a
+/// mixed epoch.
+pub(crate) const MRF_SCOPED_JOURNAL_PATH: &str = "buckets/.heal/mrf/journal-scoped.bin";
 
 /// Record format tag.
 const MRF_JOURNAL_FORMAT: u8 = 1;
 /// Record layout version.
 const MRF_JOURNAL_VERSION: u8 = 1;
+const MRF_JOURNAL_VERSION_SCOPED: u8 = 2;
 
 /// Fixed header size: format, version, kind, attempts, enqueued_at_ms,
 /// has_version flag.
 const MRF_RECORD_FIXED_HEAD: usize = 1 + 1 + 1 + 1 + 8 + 1;
+const MRF_MAX_IDENTITY_COMPONENT: usize = 1024;
+
+fn metric_f64(value: usize) -> f64 {
+    f64::from(u32::try_from(value).unwrap_or(u32::MAX))
+}
 
 #[derive(Debug, Clone)]
 pub(crate) struct MrfConsumerConfig {
@@ -101,40 +113,90 @@ impl Default for MrfConsumerConfig {
 /// incoming intent (never a resident one) and counts the loss.
 pub(crate) struct MrfQueue {
     pending: VecDeque<MrfIntent>,
+    pending_keys: HashSet<MrfQueueKey>,
     bytes: usize,
     capacity: usize,
     byte_budget: usize,
 }
 
+#[derive(Debug, Clone, PartialEq, Eq, Hash)]
+struct MrfQueueKey {
+    kind: rustfs_common::mrf_channel::MrfKind,
+    bucket: Arc<str>,
+    object: Arc<str>,
+    version_id: Option<[u8; 16]>,
+    scope: Option<rustfs_common::mrf_channel::MrfScope>,
+}
+
+fn queue_key(intent: &MrfIntent) -> MrfQueueKey {
+    let version_id = intent.version_id.filter(|bytes| *bytes != [0; 16]);
+    let scope = (!matches!(intent.kind, rustfs_common::mrf_channel::MrfKind::MetadataCorruption))
+        .then_some(intent.scope)
+        .flatten();
+    MrfQueueKey {
+        kind: intent.kind,
+        bucket: intent.bucket.clone(),
+        object: intent.object.clone(),
+        version_id,
+        scope,
+    }
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub(crate) enum MrfQueuePushResult {
+    Enqueued,
+    Coalesced,
+    Rejected,
+}
+
 impl MrfQueue {
     pub(crate) fn new(capacity: usize, byte_budget: usize) -> Self {
         Self {
             pending: VecDeque::new(),
+            pending_keys: HashSet::new(),
             bytes: 0,
             capacity,
             byte_budget,
         }
     }
 
-    /// Returns `false` (after counting) when either ceiling would be crossed.
-    pub(crate) fn try_push(&mut self, intent: MrfIntent) -> bool {
+    pub(crate) fn try_push_typed(&mut self, intent: MrfIntent) -> MrfQueuePushResult {
+        if intent.bucket.len() > MRF_MAX_IDENTITY_COMPONENT || intent.object.len() > MRF_MAX_IDENTITY_COMPONENT {
+            counter!("rustfs_heal_mrf_dropped_total", "reason" => "identity_oversized").increment(1);
+            return MrfQueuePushResult::Rejected;
+        }
+        let key = queue_key(&intent);
+        if self.pending_keys.contains(&key) {
+            counter!("rustfs_heal_mrf_coalesced_total", "layer" => "queue").increment(1);
+            return MrfQueuePushResult::Coalesced;
+        }
         let cost = intent.estimated_bytes();
         if self.pending.len() >= self.capacity || self.bytes + cost > self.byte_budget {
             counter!("rustfs_heal_mrf_dropped_total", "reason" => "queue_overflow").increment(1);
-            return false;
+            return MrfQueuePushResult::Rejected;
         }
         self.bytes += cost;
+        self.pending_keys.insert(key);
         self.pending.push_back(intent);
-        true
+        MrfQueuePushResult::Enqueued
+    }
+
+    /// Bool compatibility adapter: only a newly executable queue item is
+    /// reported as accepted; a coalesced duplicate is not durable admission.
+    #[cfg(test)]
+    pub(crate) fn try_push(&mut self, intent: MrfIntent) -> bool {
+        matches!(self.try_push_typed(intent), MrfQueuePushResult::Enqueued)
     }
 
     pub(crate) fn pop_front(&mut self) -> Option<MrfIntent> {
         let intent = self.pending.pop_front()?;
+        self.pending_keys.remove(&queue_key(&intent));
         self.bytes = self.bytes.saturating_sub(intent.estimated_bytes());
         Some(intent)
     }
 
     pub(crate) fn push_back(&mut self, intent: MrfIntent) {
+        self.pending_keys.insert(queue_key(&intent));
         self.bytes += intent.estimated_bytes();
         self.pending.push_back(intent);
     }
@@ -157,10 +219,24 @@ impl MrfQueue {
 // ---------------------------------------------------------------------------
 
 /// Append one encoded record to `out`.
-pub(crate) fn encode_intent(intent: &MrfIntent, out: &mut Vec<u8>) {
+pub(crate) fn encode_intent(intent: &MrfIntent, out: &mut Vec<u8>) -> bool {
+    let Ok(bucket_len) = u32::try_from(intent.bucket.len()) else {
+        return false;
+    };
+    let Ok(object_len) = u32::try_from(intent.object.len()) else {
+        return false;
+    };
+    let scope = (!matches!(intent.kind, rustfs_common::mrf_channel::MrfKind::MetadataCorruption))
+        .then_some(intent.scope)
+        .flatten();
+    let version_id = intent.version_id.filter(|bytes| *bytes != [0; 16]);
     let start = out.len();
     out.push(MRF_JOURNAL_FORMAT);
-    out.push(MRF_JOURNAL_VERSION);
+    out.push(if scope.is_some() {
+        MRF_JOURNAL_VERSION_SCOPED
+    } else {
+        MRF_JOURNAL_VERSION
+    });
     out.push(match intent.kind {
         rustfs_common::mrf_channel::MrfKind::DecodeFailure => 1,
         rustfs_common::mrf_channel::MrfKind::MetadataCorruption => 2,
@@ -168,27 +244,36 @@ pub(crate) fn encode_intent(intent: &MrfIntent, out: &mut Vec<u8>) {
     });
     out.push(intent.attempts);
     out.extend_from_slice(&intent.enqueued_at_ms.to_le_bytes());
-    match intent.version_id {
+    match version_id {
         Some(bytes) => {
             out.push(1);
             out.extend_from_slice(&bytes);
         }
         None => out.push(0),
     }
-    out.extend_from_slice(&(intent.bucket.len() as u32).to_le_bytes());
-    out.extend_from_slice(&(intent.object.len() as u32).to_le_bytes());
+    if let Some(scope) = scope {
+        out.extend_from_slice(&scope.pool_index.to_le_bytes());
+        out.extend_from_slice(&scope.set_index.to_le_bytes());
+    }
+    out.extend_from_slice(&bucket_len.to_le_bytes());
+    out.extend_from_slice(&object_len.to_le_bytes());
     out.extend_from_slice(intent.bucket.as_bytes());
     out.extend_from_slice(intent.object.as_bytes());
     let mut hasher = crc_fast::Digest::new(crc_fast::CrcAlgorithm::Crc32IsoHdlc);
     hasher.update(&out[start..]);
-    out.extend_from_slice(&(hasher.finalize() as u32).to_le_bytes());
+    let Ok(checksum) = u32::try_from(hasher.finalize()) else {
+        out.truncate(start);
+        return false;
+    };
+    out.extend_from_slice(&checksum.to_le_bytes());
+    true
 }
 
 fn decode_one(data: &[u8]) -> Option<(MrfIntent, usize)> {
     if data.len() < MRF_RECORD_FIXED_HEAD + 8 {
         return None;
     }
-    if data[0] != MRF_JOURNAL_FORMAT || data[1] != MRF_JOURNAL_VERSION {
+    if data[0] != MRF_JOURNAL_FORMAT || !matches!(data[1], MRF_JOURNAL_VERSION | MRF_JOURNAL_VERSION_SCOPED) {
         return None;
     }
     let kind = match data[2] {
@@ -198,24 +283,38 @@ fn decode_one(data: &[u8]) -> Option<(MrfIntent, usize)> {
         _ => return None,
     };
     let attempts = data[3];
-    let enqueued_at_ms = u64::from_le_bytes(data[4..12].try_into().expect("slice length checked"));
+    let enqueued_at_ms = u64::from_le_bytes(data[4..12].try_into().ok()?);
     let has_version = data[12] != 0;
     let mut cursor = MRF_RECORD_FIXED_HEAD;
     let version_id = if has_version {
         if data.len() < cursor + 16 {
             return None;
         }
-        let bytes: [u8; 16] = data[cursor..cursor + 16].try_into().expect("slice length checked");
+        let bytes: [u8; 16] = data[cursor..cursor + 16].try_into().ok()?;
         cursor += 16;
         Some(bytes)
     } else {
         None
     };
+    let scope = if data[1] == MRF_JOURNAL_VERSION_SCOPED {
+        if data.len() < cursor + 8 {
+            return None;
+        }
+        let pool_index = u32::from_le_bytes(data[cursor..cursor + 4].try_into().ok()?);
+        let set_index = u32::from_le_bytes(data[cursor + 4..cursor + 8].try_into().ok()?);
+        cursor += 8;
+        Some(rustfs_common::mrf_channel::MrfScope { pool_index, set_index })
+    } else {
+        None
+    };
     if data.len() < cursor + 8 {
         return None;
     }
-    let bucket_len = u32::from_le_bytes(data[cursor..cursor + 4].try_into().expect("slice length checked")) as usize;
-    let object_len = u32::from_le_bytes(data[cursor + 4..cursor + 8].try_into().expect("slice length checked")) as usize;
+    let bucket_len = usize::try_from(u32::from_le_bytes(data[cursor..cursor + 4].try_into().ok()?)).ok()?;
+    let object_len = usize::try_from(u32::from_le_bytes(data[cursor + 4..cursor + 8].try_into().ok()?)).ok()?;
+    if bucket_len > MRF_MAX_IDENTITY_COMPONENT || object_len > MRF_MAX_IDENTITY_COMPONENT {
+        return None;
+    }
     cursor += 8;
     let body_end = cursor.checked_add(bucket_len)?.checked_add(object_len)?;
     let record_end = body_end.checked_add(4)?;
@@ -224,7 +323,7 @@ fn decode_one(data: &[u8]) -> Option<(MrfIntent, usize)> {
     }
     let mut hasher = crc_fast::Digest::new(crc_fast::CrcAlgorithm::Crc32IsoHdlc);
     hasher.update(&data[..body_end]);
-    if (hasher.finalize() as u32) != u32::from_le_bytes(data[body_end..record_end].try_into().expect("slice length checked")) {
+    if u32::try_from(hasher.finalize()).ok()? != u32::from_le_bytes(data[body_end..record_end].try_into().ok()?) {
         return None;
     }
     let bucket = std::sync::Arc::from(std::str::from_utf8(&data[cursor..cursor + bucket_len]).ok()?);
@@ -235,6 +334,12 @@ fn decode_one(data: &[u8]) -> Option<(MrfIntent, usize)> {
             object,
             version_id,
             kind,
+            scope: if matches!(kind, rustfs_common::mrf_channel::MrfKind::MetadataCorruption) {
+                None
+            } else {
+                scope
+            },
+            lease: None,
             enqueued_at_ms,
             attempts,
         },
@@ -269,9 +374,9 @@ async fn journal_disks() -> Vec<DiskStore> {
     map.values().flatten().cloned().collect()
 }
 
-async fn read_journal() -> Option<Vec<u8>> {
+async fn read_journal(path: &str) -> Option<Vec<u8>> {
     for disk in journal_disks().await {
-        match disk.read_all(super::RUSTFS_META_BUCKET, MRF_JOURNAL_PATH).await {
+        match disk.read_all(super::RUSTFS_META_BUCKET, path).await {
             Ok(bytes) => return Some(bytes.to_vec()),
             Err(_) => continue,
         }
@@ -282,35 +387,51 @@ async fn read_journal() -> Option<Vec<u8>> {
 /// Write the snapshot to every local disk; returns true when at least one
 /// disk accepted it, so a total write failure keeps the runtime dirty and
 /// the next tick retries the persist.
-async fn write_journal(data: &[u8]) -> bool {
+async fn write_journal(path: &str, data: &[u8]) -> bool {
     let payload = bytes::Bytes::copy_from_slice(data);
     let mut any_persisted = false;
     for disk in journal_disks().await {
-        match disk
-            .write_all(super::RUSTFS_META_BUCKET, MRF_JOURNAL_PATH, payload.clone())
-            .await
-        {
+        match disk.write_all(super::RUSTFS_META_BUCKET, path, payload.clone()).await {
             Ok(()) => any_persisted = true,
             Err(err) => warn_mrf_journal_write(&err),
         }
     }
-    if !data.is_empty() {
-        counter!("rustfs_heal_mrf_journal_fsync_total").increment(1);
-    }
-    gauge!("rustfs_heal_mrf_journal_bytes").set(data.len() as f64);
     any_persisted
 }
 
-async fn delete_journal() {
-    for disk in journal_disks().await {
-        let _ = disk
+async fn delete_journal(path: &str) -> bool {
+    let disks = journal_disks().await;
+    if disks.is_empty() {
+        counter!("rustfs_heal_mrf_journal_delete_failures_total").increment(1);
+        return false;
+    }
+    let mut all_deleted = true;
+    for disk in disks {
+        let result = disk
             .delete(
                 super::RUSTFS_META_BUCKET,
-                MRF_JOURNAL_PATH,
+                path,
                 crate::heal::storage_api::owner::EcstoreDeleteOptions::default(),
             )
             .await;
+        if let Err(err) = result {
+            // Delete is idempotent: a compatibility mirror that was never
+            // written (or was already removed) is clean, not a retry state.
+            if !matches!(err, super::DiskError::FileNotFound | super::DiskError::VolumeNotFound) {
+                all_deleted = false;
+            }
+        }
     }
+    if !all_deleted {
+        counter!("rustfs_heal_mrf_journal_delete_failures_total").increment(1);
+    }
+    all_deleted
+}
+
+async fn delete_journals() -> bool {
+    let authoritative_deleted = delete_journal(MRF_SCOPED_JOURNAL_PATH).await;
+    let legacy_deleted = delete_journal(MRF_JOURNAL_PATH).await;
+    authoritative_deleted && legacy_deleted
 }
 
 fn warn_mrf_journal_write(err: &super::DiskError) {
@@ -331,7 +452,10 @@ fn warn_mrf_journal_write(err: &super::DiskError) {
 pub(crate) fn build_heal_request(intent: &MrfIntent) -> HealRequest {
     let bucket = intent.bucket.to_string();
     let object = intent.object.to_string();
-    let version_id = intent.version_id.map(|bytes| Uuid::from_bytes(bytes).to_string());
+    let version_id = intent
+        .version_id
+        .filter(|bytes| *bytes != [0; 16])
+        .map(|bytes| Uuid::from_bytes(bytes).to_string());
     let (heal_type, priority) = match intent.kind {
         rustfs_common::mrf_channel::MrfKind::DecodeFailure => (
             HealType::ECDecode {
@@ -351,18 +475,28 @@ pub(crate) fn build_heal_request(intent: &MrfIntent) -> HealRequest {
             HealPriority::Normal,
         ),
     };
-    let mut request = HealRequest::new(heal_type, HealOptions::default(), priority);
+    let mut options = HealOptions::default();
+    if !matches!(intent.kind, rustfs_common::mrf_channel::MrfKind::MetadataCorruption)
+        && let Some(scope) = intent.scope
+    {
+        options.pool_index = usize::try_from(scope.pool_index).ok();
+        options.set_index = usize::try_from(scope.set_index).ok();
+    }
+    let mut request = HealRequest::new(heal_type, options, priority);
     request.source = rustfs_common::heal_channel::HealRequestSource::Mrf;
     request
 }
 
 async fn submit_mrf_heal_request(manager: &HealManager, intent: &MrfIntent) -> crate::Result<HealAdmissionResult> {
     let receipt = manager
-        .submit_mrf_heal_request_with_receipt(
+        .submit_mrf_heal_request_with_receipt_and_identity(
             build_heal_request(intent),
             intent.bucket.clone(),
             intent.object.clone(),
             intent.version_id,
+            intent.kind,
+            intent.scope,
+            intent.lease,
         )
         .await?;
     Ok(receipt.result)
@@ -387,16 +521,38 @@ struct MrfRuntime {
 }
 
 impl MrfRuntime {
-    fn snapshot(&self) -> Vec<u8> {
-        let mut buf = Vec::new();
+    fn snapshot(&self) -> (Vec<u8>, Vec<u8>) {
+        let mut authoritative = Vec::new();
+        let mut legacy = Vec::new();
         for intent in self.queue.intents() {
-            encode_intent(intent, &mut buf);
+            let scoped_identity =
+                !matches!(intent.kind, rustfs_common::mrf_channel::MrfKind::MetadataCorruption) && intent.scope.is_some();
+            if !encode_intent(intent, &mut authoritative) {
+                counter!("rustfs_heal_mrf_dropped_total", "reason" => "journal_identity_oversized").increment(1);
+            }
+            if !scoped_identity && !encode_intent(intent, &mut legacy) {
+                counter!("rustfs_heal_mrf_dropped_total", "reason" => "journal_identity_oversized").increment(1);
+            }
         }
-        buf
+        (authoritative, legacy)
     }
 
     async fn flush(&mut self) {
-        let persisted = write_journal(&self.snapshot()).await;
+        let (authoritative, legacy) = self.snapshot();
+        let authoritative_persisted = write_journal(MRF_SCOPED_JOURNAL_PATH, &authoritative).await;
+        if !authoritative.is_empty() {
+            counter!("rustfs_heal_mrf_journal_fsync_total").increment(1);
+        }
+        gauge!("rustfs_heal_mrf_journal_bytes").set(metric_f64(authoritative.len()));
+        // Publish the compatibility mirror only after the authoritative
+        // snapshot has reached at least one disk. This ordering prevents an
+        // old reader from observing a newer epoch that a new reader cannot
+        // see when the canonical write is unavailable.
+        let legacy_persisted = authoritative_persisted && write_journal(MRF_JOURNAL_PATH, &legacy).await;
+        // Keep dirty until both the authoritative snapshot and its
+        // compatibility mirror have been accepted; otherwise a one-sided
+        // failure would never retry the missing file.
+        let persisted = authoritative_persisted && legacy_persisted;
         self.new_since_flush = 0;
         // Keep the dirty flag when every disk write failed: a clean backlog
         // would otherwise never rewrite, losing the periodic persist retry a
@@ -404,7 +560,7 @@ impl MrfRuntime {
         if persisted {
             self.dirty = false;
         }
-        self.journal_on_disk = true;
+        self.journal_on_disk |= authoritative_persisted || legacy_persisted;
     }
 
     /// Drain pending intents into the heal manager until it is full, the
@@ -430,6 +586,7 @@ impl MrfRuntime {
                     intent.attempts = intent.attempts.saturating_add(1);
                     if intent.attempts >= MRF_MAX_ATTEMPTS {
                         counter!("rustfs_heal_mrf_dropped_total", "reason" => "attempts_exhausted").increment(1);
+                        rustfs_common::mrf_channel::release_mrf_intent(&intent);
                         continue;
                     }
                     self.queue.push_back(intent);
@@ -438,11 +595,13 @@ impl MrfRuntime {
                 }
                 Ok(HealAdmissionResult::Dropped(_)) => {
                     counter!("rustfs_heal_mrf_dropped_total", "reason" => "admission_policy").increment(1);
+                    rustfs_common::mrf_channel::release_mrf_intent(&intent);
                 }
                 Err(_) => {
                     intent.attempts = intent.attempts.saturating_add(1);
                     if intent.attempts >= MRF_MAX_ATTEMPTS {
                         counter!("rustfs_heal_mrf_dropped_total", "reason" => "attempts_exhausted").increment(1);
+                        rustfs_common::mrf_channel::release_mrf_intent(&intent);
                         continue;
                     }
                     self.queue.push_back(intent);
@@ -451,8 +610,8 @@ impl MrfRuntime {
                 }
             }
         }
-        gauge!("rustfs_heal_mrf_queue_depth").set(self.queue.depth() as f64);
-        gauge!("rustfs_heal_mrf_queue_bytes").set(self.queue.bytes() as f64);
+        gauge!("rustfs_heal_mrf_queue_depth").set(metric_f64(self.queue.depth()));
+        gauge!("rustfs_heal_mrf_queue_bytes").set(metric_f64(self.queue.bytes()));
     }
 }
 
@@ -496,7 +655,12 @@ pub async fn replay_journal_once(manager: &Arc<HealManager>) -> usize {
     let config = MrfConsumerConfig::default();
     let mut queue = MrfQueue::new(config.queue_capacity, config.journal_max_bytes);
     let mut backoff_until: Option<tokio::time::Instant> = None;
-    replay_into(manager, &mut queue, &mut backoff_until).await
+    replay_into(manager, &mut queue, &mut backoff_until).await.replayed
+}
+
+struct ReplayOutcome {
+    replayed: usize,
+    journal_on_disk: bool,
 }
 
 /// Shared replay core: read + decode + re-arm + delete, then drain what fits.
@@ -504,11 +668,25 @@ async fn replay_into(
     manager: &Arc<HealManager>,
     queue: &mut MrfQueue,
     backoff_until: &mut Option<tokio::time::Instant>,
-) -> usize {
-    let Some(data) = read_journal().await else {
-        return 0;
+) -> ReplayOutcome {
+    // The scoped file is a complete authoritative snapshot. Fall back to the
+    // legacy mirror only when the authoritative path is unavailable; merging
+    // both files could combine records from different flush epochs.
+    let data = match read_journal(MRF_SCOPED_JOURNAL_PATH).await {
+        Some(data) => data,
+        None => match read_journal(MRF_JOURNAL_PATH).await {
+            Some(data) => data,
+            None => {
+                return ReplayOutcome {
+                    replayed: 0,
+                    journal_on_disk: false,
+                };
+            }
+        },
     };
-    let (intents, truncated) = decode_journal(&data);
+    let mut intents = Vec::new();
+    let (decoded, truncated) = decode_journal(&data);
+    intents.extend(decoded);
     if truncated > 0 {
         tracing::warn!(
             target: "rustfs::heal::mrf",
@@ -516,12 +694,15 @@ async fn replay_into(
             "MRF journal had a torn tail; truncated records were discarded"
         );
     }
-    counter!("rustfs_heal_mrf_replayed_total").increment(intents.len() as u64);
+    counter!("rustfs_heal_mrf_replayed_total").increment(u64::try_from(intents.len()).unwrap_or(u64::MAX));
     let replayed = intents.len();
     for intent in intents {
-        queue.try_push(intent);
+        let result = queue.try_push_typed(intent.clone());
+        if !matches!(result, MrfQueuePushResult::Enqueued) {
+            rustfs_common::mrf_channel::release_mrf_intent(&intent);
+        }
     }
-    delete_journal().await;
+    let journal_on_disk = !delete_journals().await;
 
     // Drain the replayed intents immediately; whatever the manager refuses
     // stays armed in `queue` for the consumer's retry loop.
@@ -541,7 +722,10 @@ async fn replay_into(
             }
         }
     }
-    replayed
+    ReplayOutcome {
+        replayed,
+        journal_on_disk,
+    }
 }
 
 /// Replay the journal, then keep draining the channel into the heal manager
@@ -559,7 +743,8 @@ async fn run_mrf_consumer(manager: Arc<HealManager>, mut receiver: mpsc::Receive
 
     // Replay: read the journal, re-arm intents (duplicates are merged by the
     // manager's dedup key), then drop the file so the next flush starts clean.
-    replay_into(&manager, &mut runtime.queue, &mut runtime.backoff_until).await;
+    let replay = replay_into(&manager, &mut runtime.queue, &mut runtime.backoff_until).await;
+    runtime.journal_on_disk = replay.journal_on_disk;
     // The replay deleted the journal file; anything still pending (e.g. the
     // manager was full and backoff armed) must be re-persisted by the next
     // flush or a crash before it would lose those intents.
@@ -587,9 +772,14 @@ async fn run_mrf_consumer(manager: Arc<HealManager>, mut receiver: mpsc::Receive
                     return;
                 }
                 for intent in batch.drain(..) {
-                    if runtime.queue.try_push(intent) {
-                        runtime.new_since_flush += 1;
-                        runtime.dirty = true;
+                    match runtime.queue.try_push_typed(intent.clone()) {
+                        MrfQueuePushResult::Enqueued => {
+                            runtime.new_since_flush += 1;
+                            runtime.dirty = true;
+                        }
+                        MrfQueuePushResult::Coalesced | MrfQueuePushResult::Rejected => {
+                            rustfs_common::mrf_channel::release_mrf_intent(&intent);
+                        }
                     }
                 }
                 runtime.dispatch(manager.as_ref()).await;
@@ -613,13 +803,14 @@ async fn run_mrf_consumer(manager: Arc<HealManager>, mut receiver: mpsc::Receive
                     TickAction::DeleteJournal => {
                         // All intents consumed: remove the journal so a restart
                         // replays nothing (mirrors MinIO's post-replay unlink).
-                        delete_journal().await;
-                        runtime.journal_on_disk = false;
-                        gauge!("rustfs_heal_mrf_journal_bytes").set(0.0);
+                        if delete_journals().await {
+                            runtime.journal_on_disk = false;
+                            gauge!("rustfs_heal_mrf_journal_bytes").set(0.0);
+                        }
                     }
                     TickAction::Idle => {}
                 }
-                gauge!("rustfs_heal_mrf_queue_depth").set(runtime.queue.depth() as f64);
+                gauge!("rustfs_heal_mrf_queue_depth").set(metric_f64(runtime.queue.depth()));
             }
         }
     }
@@ -664,6 +855,8 @@ mod tests {
             object: StdArc::from(object),
             version_id: Some([7u8; 16]),
             kind: MrfKind::DecodeFailure,
+            scope: None,
+            lease: None,
             enqueued_at_ms: 1_700_000_000_000,
             attempts,
         }
@@ -694,17 +887,101 @@ mod tests {
     fn queue_enforces_count_and_byte_ceilings() {
         let mut queue = MrfQueue::new(2, usize::MAX);
         assert!(queue.try_push(intent("b", "o", 0)));
-        assert!(queue.try_push(intent("b", "o", 0)));
-        assert!(!queue.try_push(intent("b", "o", 0)), "count ceiling must drop");
+        assert!(queue.try_push(intent("b", "o2", 0)));
+        assert!(!queue.try_push(intent("b", "o3", 0)), "count ceiling must drop");
 
         let mut tiny = MrfQueue::new(usize::MAX, intent("bucket", "object", 0).estimated_bytes());
         assert!(tiny.try_push(intent("bucket", "object", 0)));
         assert!(
-            !tiny.try_push(intent("bucket", "object", 0)),
+            !tiny.try_push(intent("bucket", "object2", 0)),
             "byte budget must drop before the second intent fits"
         );
     }
 
+    #[test]
+    fn duplicate_mrf_intents_coalesce_to_one_execution() {
+        let mut queue = MrfQueue::new(1000, usize::MAX);
+        let mut enqueued = 0;
+        let mut coalesced = 0;
+        assert_eq!(queue.try_push_typed(intent("bucket", "object", 0)), MrfQueuePushResult::Enqueued);
+        enqueued += 1;
+        for _ in 0..999 {
+            match queue.try_push_typed(intent("bucket", "object", 0)) {
+                MrfQueuePushResult::Coalesced => coalesced += 1,
+                other => panic!("duplicate intent was not coalesced: {other:?}"),
+            }
+        }
+        assert_eq!(enqueued, 1);
+        assert_eq!(coalesced, 999);
+        assert_eq!(queue.depth(), 1);
+    }
+
+    #[test]
+    fn mrf_dedupe_does_not_merge_adjacent_version_pool_or_kind() {
+        let mut queue = MrfQueue::new(8, usize::MAX);
+        let mut first = intent("bucket", "object", 0);
+        first.kind = MrfKind::PartialWrite;
+        first.scope = Some(rustfs_common::mrf_channel::MrfScope {
+            pool_index: 1,
+            set_index: 1,
+        });
+        assert!(queue.try_push(first.clone()));
+        first.version_id = Some([8u8; 16]);
+        assert!(queue.try_push(first));
+        let mut other_scope = intent("bucket", "object", 0);
+        other_scope.kind = MrfKind::PartialWrite;
+        other_scope.scope = Some(rustfs_common::mrf_channel::MrfScope {
+            pool_index: 2,
+            set_index: 1,
+        });
+        assert!(queue.try_push(other_scope));
+        let mut other_kind = intent("bucket", "object", 0);
+        other_kind.kind = MrfKind::DecodeFailure;
+        other_kind.scope = None;
+        assert!(queue.try_push(other_kind));
+        assert_eq!(queue.depth(), 4);
+    }
+
+    #[test]
+    fn mrf_dedupe_full_returns_rejected_with_durable_pending() {
+        let mut queue = MrfQueue::new(1, usize::MAX);
+        assert_eq!(queue.try_push_typed(intent("bucket", "object", 0)), MrfQueuePushResult::Enqueued);
+        assert_eq!(queue.try_push_typed(intent("bucket", "other", 0)), MrfQueuePushResult::Rejected);
+        assert_eq!(queue.depth(), 1);
+        let mut snapshot = Vec::new();
+        assert!(encode_intent(queue.intents().next().expect("resident intent"), &mut snapshot));
+        assert!(!snapshot.is_empty(), "the resident intent remains journalable after rejection");
+    }
+
+    #[test]
+    fn mrf_dedupe_failure_releases_key_for_retry() {
+        let mut queue = MrfQueue::new(1, usize::MAX);
+        assert_eq!(queue.try_push_typed(intent("bucket", "object", 0)), MrfQueuePushResult::Enqueued);
+        let _failed = queue.pop_front().expect("queued intent");
+        assert_eq!(queue.try_push_typed(intent("bucket", "object", 1)), MrfQueuePushResult::Enqueued);
+        assert_eq!(queue.depth(), 1);
+    }
+
+    #[test]
+    fn mrf_dedupe_key_and_map_are_bounded() {
+        let mut queue = MrfQueue::new(2, usize::MAX);
+        assert_eq!(queue.try_push_typed(intent("bucket", "object", 0)), MrfQueuePushResult::Enqueued);
+        assert_eq!(queue.try_push_typed(intent("bucket", "other", 0)), MrfQueuePushResult::Enqueued);
+        assert_eq!(queue.pending_keys.len(), 2);
+        assert_eq!(queue.try_push_typed(intent("bucket", "third", 0)), MrfQueuePushResult::Rejected);
+        assert_eq!(queue.depth(), 2);
+    }
+
+    #[test]
+    fn cross_node_duplicate_execution_remains_idempotent() {
+        // Node-local ingress maps intentionally do not merge across nodes;
+        // the manager's existing identity key absorbs the duplicate later.
+        let mut node_a = MrfQueue::new(8, usize::MAX);
+        let mut node_b = MrfQueue::new(8, usize::MAX);
+        assert_eq!(node_a.try_push_typed(intent("bucket", "object", 0)), MrfQueuePushResult::Enqueued);
+        assert_eq!(node_b.try_push_typed(intent("bucket", "object", 0)), MrfQueuePushResult::Enqueued);
+    }
+
     #[test]
     fn journal_roundtrip_preserves_intents() {
         let intents = vec![
@@ -715,6 +992,8 @@ mod tests {
                 object: StdArc::from("object/c"),
                 version_id: None,
                 kind: MrfKind::MetadataCorruption,
+                scope: None,
+                lease: None,
                 enqueued_at_ms: 5,
                 attempts: 1,
             },
@@ -766,6 +1045,8 @@ mod tests {
             object: StdArc::from("o"),
             version_id: None,
             kind: MrfKind::MetadataCorruption,
+            scope: None,
+            lease: None,
             enqueued_at_ms: 0,
             attempts: 0,
         });
@@ -777,6 +1058,8 @@ mod tests {
             object: StdArc::from("o"),
             version_id: None,
             kind: MrfKind::PartialWrite,
+            scope: None,
+            lease: None,
             enqueued_at_ms: 0,
             attempts: 0,
         });

crates/heal/tests/mrf_pipeline_test.rs

@@ -35,6 +35,7 @@ use storage_api::endpoint_index::{Endpoint, EndpointServerPools, Endpoints, Pool
 
 const META_BUCKET: &str = ".rustfs.sys";
 const JOURNAL_REL: &str = "buckets/.heal/mrf/journal.bin";
+const SCOPED_JOURNAL_REL: &str = "buckets/.heal/mrf/journal-scoped.bin";
 
 async fn heal_env() -> (Vec<std::path::PathBuf>, Arc<dyn HealStorageAPI>) {
     let env = rustfs_test_utils::TestECStoreEnv::builder()
@@ -79,14 +80,18 @@ fn journal_record(kind: u8, bucket: &str, object: &str, version: Option<[u8; 16]
     body
 }
 
-fn write_journal_to_disks(disk_paths: &[std::path::PathBuf], data: &[u8]) {
+fn write_journal_path_to_disks(disk_paths: &[std::path::PathBuf], relative_path: &str, data: &[u8]) {
     for path in disk_paths {
-        let journal = path.join(META_BUCKET).join(JOURNAL_REL);
+        let journal = path.join(META_BUCKET).join(relative_path);
         std::fs::create_dir_all(journal.parent().expect("journal parent")).expect("create journal dir");
         std::fs::write(&journal, data).expect("write journal fixture");
     }
 }
 
+fn write_journal_to_disks(disk_paths: &[std::path::PathBuf], data: &[u8]) {
+    write_journal_path_to_disks(disk_paths, JOURNAL_REL, data);
+}
+
 async fn wait_until<F, Fut>(deadline: Duration, mut probe: F) -> bool
 where
     F: FnMut() -> Fut,
@@ -187,8 +192,73 @@ async fn journal_replay_arms_intents_and_deletes_the_file() {
             .all(|path| !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()),
         "the journal file must be removed after a successful replay"
     );
+    assert!(
+        disk_paths
+            .iter()
+            .all(|path| !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()),
+        "the authoritative journal file must also be removed after replay"
+    );
 
     let snapshot = manager.operations_snapshot().await;
     assert_eq!(snapshot.queued_by_priority.urgent, 1, "the decode-failure record must replay as Urgent");
     assert!(snapshot.queued_by_priority.normal >= 1, "the partial-write record must replay as Normal");
 }
+
+/// A canonical snapshot and its compatibility mirror may differ after a
+/// partial flush. Replay must choose the complete canonical epoch instead of
+/// combining records that never coexisted in memory.
+#[tokio::test(flavor = "multi_thread", worker_threads = 4)]
+#[serial]
+async fn authoritative_journal_is_not_merged_with_legacy_mirror() {
+    let (disk_paths, storage) = heal_env().await;
+    let mut endpoints: Vec<Endpoint> = disk_paths
+        .iter()
+        .map(|p| Endpoint::try_from(p.to_string_lossy().as_ref()).expect("endpoint from disk path"))
+        .collect();
+    for (i, endpoint) in endpoints.iter_mut().enumerate() {
+        endpoint.set_pool_index(0);
+        endpoint.set_set_index(0);
+        endpoint.set_disk_index(i);
+    }
+    let pool = PoolEndpoints {
+        legacy: false,
+        set_count: 1,
+        drives_per_set: endpoints.len(),
+        endpoints: Endpoints::from(endpoints),
+        cmd_line: "mrf-authoritative-test".to_string(),
+        platform: String::new(),
+    };
+    init_local_disks(EndpointServerPools::from(vec![pool]))
+        .await
+        .expect("local disks should register");
+
+    let authoritative = journal_record(1, "authoritative-bucket", "authoritative-object", None, 0);
+    let legacy = journal_record(1, "legacy-bucket", "legacy-object", None, 0);
+    write_journal_path_to_disks(&disk_paths, SCOPED_JOURNAL_REL, &authoritative);
+    write_journal_path_to_disks(&disk_paths, JOURNAL_REL, &legacy);
+
+    let manager = make_manager(storage);
+    let replayed = mrf_queue::replay_journal_once(&manager).await;
+    assert_eq!(replayed, 1, "only the authoritative snapshot epoch may replay");
+
+    let snapshot = manager.operations_snapshot().await;
+    assert_eq!(snapshot.queued_by_source.mrf, 1);
+    assert!(
+        disk_paths.iter().all(|path| {
+            !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
+                && !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
+        }),
+        "replay cleanup must remove both journal paths"
+    );
+
+    // A scoped-only snapshot is valid during a rollout where no legacy
+    // compatibility mirror was written. Missing legacy files must not leave
+    // the runtime in a permanent cleanup-retry state.
+    let scoped_only = journal_record(1, "scoped-only-bucket", "scoped-only-object", None, 0);
+    write_journal_path_to_disks(&disk_paths, SCOPED_JOURNAL_REL, &scoped_only);
+    assert_eq!(mrf_queue::replay_journal_once(&manager).await, 1);
+    assert!(disk_paths.iter().all(|path| {
+        !Path::new(path).join(META_BUCKET).join(JOURNAL_REL).exists()
+            && !Path::new(path).join(META_BUCKET).join(SCOPED_JOURNAL_REL).exists()
+    }));
+}

crates/scanner/src/scanner_folder.rs

@@ -1375,13 +1375,14 @@ impl FolderScanner {
                             // Single-flight (backlog#1894 axis A) — the
                             // recording mode and its guarantees are pinned by
                             // corrupt_metadata_recording below.
-                            let mrf_accepted = rustfs_common::mrf_channel::try_send_mrf_intent(
+                            let mrf_result = rustfs_common::mrf_channel::try_send_mrf_intent_typed(
                                 rustfs_common::mrf_channel::MrfKind::MetadataCorruption,
                                 &item.bucket,
                                 &object,
                                 None,
+                                None,
                             );
-                            match corrupt_metadata_recording(mrf_accepted) {
+                            match corrupt_metadata_recording(mrf_result) {
                                 CorruptMetadataRecording::LedgerOnly => {
                                     // Recorded as Full (retry-later): admission
                                     // for this target happens in the MRF

crates/scanner/src/scanner_folder/item_actions.rs

@@ -50,11 +50,12 @@ pub(super) enum CorruptMetadataRecording {
     ImmediateAndLedger,
 }
 
-pub(super) fn corrupt_metadata_recording(mrf_accepted: bool) -> CorruptMetadataRecording {
-    if mrf_accepted {
-        CorruptMetadataRecording::LedgerOnly
-    } else {
-        CorruptMetadataRecording::ImmediateAndLedger
+pub(super) fn corrupt_metadata_recording(result: rustfs_common::mrf_channel::MrfIngressResult) -> CorruptMetadataRecording {
+    match result {
+        rustfs_common::mrf_channel::MrfIngressResult::Enqueued | rustfs_common::mrf_channel::MrfIngressResult::Coalesced => {
+            CorruptMetadataRecording::LedgerOnly
+        }
+        rustfs_common::mrf_channel::MrfIngressResult::Dropped(_) => CorruptMetadataRecording::ImmediateAndLedger,
     }
 }
 

crates/scanner/src/scanner_folder/tests.rs

@@ -48,8 +48,20 @@ fn scanner_alert_wire_names_match_canonical_event_names() {
 /// the backstop survives regardless of delivery.
 #[test]
 fn corrupt_metadata_recording_maps_delivery_to_backstop() {
-    assert_eq!(corrupt_metadata_recording(true), CorruptMetadataRecording::LedgerOnly);
-    assert_eq!(corrupt_metadata_recording(false), CorruptMetadataRecording::ImmediateAndLedger);
+    assert_eq!(
+        corrupt_metadata_recording(rustfs_common::mrf_channel::MrfIngressResult::Enqueued),
+        CorruptMetadataRecording::LedgerOnly
+    );
+    assert_eq!(
+        corrupt_metadata_recording(rustfs_common::mrf_channel::MrfIngressResult::Coalesced),
+        CorruptMetadataRecording::LedgerOnly
+    );
+    assert_eq!(
+        corrupt_metadata_recording(rustfs_common::mrf_channel::MrfIngressResult::Dropped(
+            rustfs_common::mrf_channel::MrfDropReason::Full
+        )),
+        CorruptMetadataRecording::ImmediateAndLedger
+    );
 }
 
 fn cooldown_map_len() -> usize {

deny.toml

@@ -43,6 +43,9 @@ allow-git = [
     # RustFS fork carrying presigned expiry and constant-time authentication fixes.
     # owner: rustfs-maintainers  review: 2026-10
     "https://github.com/rustfs/s3s.git",
+    # MiMalloc fork pinned for hotpath allocation counting support.
+    # owner: houseme  review: 2026-10
+    "https://github.com/xonatius/mimalloc_rust.git",
 ]
 
 [bans]

docs/testing/e2e-suite-inventory.md

@@ -58,7 +58,7 @@
 | heal_erasure_disk_rebuild_test | 4 | 🌙 |
 | inline_fast_path_cluster_test | 16 |  |
 | internode_rpc_signature_e2e_test | 5 |  |
-| kms | 47 |  |
+| kms | 46 |  |
 | leading_slash_key_test | 2 | ✅ |
 | lifecycle_regression_test | 4 |  |
 | list_buckets_auth_test | 1 | ✅ |
@@ -99,4 +99,4 @@
 | tls_hot_reload_test | 1 | ✅ |
 | version_id_regression_test | 10 | ✅ |
 
-**Total listed: 576 tests across 82 modules · PR smoke: 163 tests / 36 modules · merge/main full: 454 tests / 73 modules · nightly replication: 55 tests · nightly cluster faults: 28 tests / 7 modules · nightly protocols: 16 tests** · updated 2026-08-23.
+**Total listed: 575 tests across 82 modules · PR smoke: 163 tests / 36 modules · merge/main full: 453 tests / 73 modules · nightly replication: 55 tests · nightly cluster faults: 28 tests / 7 modules · nightly protocols: 16 tests** · updated 2026-08-23.

rustfs/Cargo.toml

@@ -336,13 +336,13 @@ opentelemetry = { workspace = true }
 tracing-opentelemetry = { workspace = true }
 # Data structures
 hashbrown = { workspace = true, features = ["serde", "rayon"] }
-rustfs-mimalloc = { workspace = true }
+mimalloc = { workspace = true }
 
 [target.'cfg(target_os = "linux")'.dependencies]
 libsystemd.workspace = true
 
 [target.'cfg(not(target_os = "windows"))'.dependencies]
-rustfs-mimalloc-sys.workspace = true
+libmimalloc-sys.workspace = true
 
 [dev-dependencies]
 uuid = { workspace = true, features = ["v4", "v5", "fast-rng", "macro-diagnostics"] }

rustfs/src/allocator_reclaim.rs

@@ -369,8 +369,14 @@ pub fn allocator_reclaim_controller_snapshot(ctx: &CancellationToken) -> Allocat
 }
 
 #[cfg(not(target_os = "windows"))]
+#[allow(unsafe_code)]
 fn collect_allocator_memory(force: bool) -> Result<(), String> {
-    rustfs_mimalloc::MiMalloc::collect(force);
+    // SAFETY: `mi_collect` is provided by the active global allocator backend
+    // on this target family. It is explicitly intended to reclaim retained
+    // pages/segments and does not require additional invariants from the caller.
+    unsafe {
+        libmimalloc_sys::mi_collect(force);
+    }
     Ok(())
 }
 

rustfs/src/main.rs

@@ -26,22 +26,22 @@ struct MiMallocAllocator;
 unsafe impl GlobalAlloc for MiMallocAllocator {
     unsafe fn alloc(&self, layout: Layout) -> *mut u8 {
         // SAFETY: the caller upholds GlobalAlloc's contract for layout.
-        unsafe { rustfs_mimalloc::MiMalloc.alloc(layout) }
+        unsafe { mimalloc::MiMalloc.alloc(layout) }
     }
 
     unsafe fn alloc_zeroed(&self, layout: Layout) -> *mut u8 {
         // SAFETY: the caller upholds GlobalAlloc's contract for layout.
-        unsafe { rustfs_mimalloc::MiMalloc.alloc_zeroed(layout) }
+        unsafe { mimalloc::MiMalloc.alloc_zeroed(layout) }
     }
 
     unsafe fn dealloc(&self, ptr: *mut u8, layout: Layout) {
         // SAFETY: ptr and layout came from this allocator and are forwarded unchanged.
-        unsafe { rustfs_mimalloc::MiMalloc.dealloc(ptr, layout) }
+        unsafe { mimalloc::MiMalloc.dealloc(ptr, layout) }
     }
 
     unsafe fn realloc(&self, ptr: *mut u8, layout: Layout, new_size: usize) -> *mut u8 {
         // SAFETY: ptr and layout came from this allocator and are forwarded unchanged.
-        unsafe { rustfs_mimalloc::MiMalloc.realloc(ptr, layout, new_size) }
+        unsafe { mimalloc::MiMalloc.realloc(ptr, layout, new_size) }
     }
 }
 
@@ -51,7 +51,7 @@ static GLOBAL: hotpath::CountingAllocator<MiMallocAllocator> = hotpath::Counting
 
 #[cfg(not(all(feature = "hotpath", feature = "hotpath-alloc")))]
 #[global_allocator]
-static GLOBAL: rustfs_mimalloc::MiMalloc = rustfs_mimalloc::MiMalloc;
+static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;
 
 fn main() {
     let _hotpath_guard = hotpath::HotpathGuardBuilder::new("main").build();
@@ -71,9 +71,8 @@ mod tests {
         allocation.extend_from_slice(&[7_u8; 64]);
 
         assert_eq!(allocation.len(), 64);
-        let heap = rustfs_mimalloc::heap::Heap::main();
         // SAFETY: the live Vec pointer is valid to inspect for heap ownership.
-        assert!(unsafe { heap.contains(allocation.as_ptr()) });
+        assert!(unsafe { libmimalloc_sys::mi_is_in_heap_region(allocation.as_ptr().cast()) });
     }
 
     #[test]
@@ -86,13 +85,12 @@ mod tests {
         let layout = Layout::from_size_align(32, 8).expect("valid test allocation layout");
         let grown_layout = Layout::from_size_align(64, 8).expect("valid grown test allocation layout");
         let allocator = super::MiMallocAllocator;
-        let heap = rustfs_mimalloc::heap::Heap::main();
 
         // SAFETY: The pointer is checked for null before use and later released
         // through the same allocator with the corresponding layout.
         let ptr = unsafe { allocator.alloc_zeroed(layout) };
         assert!(!ptr.is_null());
-        assert!(unsafe { heap.contains(ptr) });
+        assert!(unsafe { libmimalloc_sys::mi_is_in_heap_region(ptr.cast()) });
         assert!(unsafe { std::slice::from_raw_parts(ptr, 32).iter().all(|byte| *byte == 0) });
 
         // SAFETY: `ptr` was allocated by `allocator` with `layout`; on failure
@@ -104,7 +102,7 @@ mod tests {
             panic!("mimalloc realloc failed in allocator smoke test");
         }
 
-        assert!(unsafe { heap.contains(grown_ptr) });
+        assert!(unsafe { libmimalloc_sys::mi_is_in_heap_region(grown_ptr.cast()) });
         // SAFETY: `grown_ptr` was reallocated by `allocator` and is released
         // with the matching grown layout.
         unsafe { allocator.dealloc(grown_ptr, grown_layout) };

rustfs/src/memory_observability.rs

@@ -17,7 +17,10 @@ use rustfs_io_metrics::{
     record_cpu_usage, record_memory_usage, record_process_memory_split,
 };
 use serde::Serialize;
+#[cfg(any(test, not(target_os = "windows")))]
 use serde_json::Value;
+#[cfg(not(target_os = "windows"))]
+use std::ffi::CStr;
 use std::path::Path;
 use std::sync::{Arc, Mutex, OnceLock};
 use std::time::Duration;
@@ -228,18 +231,7 @@ fn read_cgroup_memory_snapshot() -> Option<CgroupMemorySnapshot> {
     read_cgroup_v2().or_else(read_cgroup_v1)
 }
 
-fn read_allocator_memory_snapshot() -> Option<AllocatorMemorySnapshot> {
-    let json = rustfs_mimalloc::MiMalloc::stats_json();
-    if json.is_empty() {
-        return None;
-    }
-    let observation = parse_mimalloc_stats_json(&json)?;
-    Some(AllocatorMemorySnapshot {
-        backend: crate::allocator_reclaim::allocator_backend(),
-        observation,
-    })
-}
-
+#[cfg(any(test, not(target_os = "windows")))]
 fn numeric_json_value(value: &Value) -> Option<u64> {
     match value {
         Value::Number(number) => number
@@ -250,6 +242,7 @@ fn numeric_json_value(value: &Value) -> Option<u64> {
     }
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn numeric_json_field(value: &Value, field: &str) -> Option<u64> {
     match value {
         Value::Object(fields) => fields
@@ -261,6 +254,7 @@ fn numeric_json_field(value: &Value, field: &str) -> Option<u64> {
     }
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn mimalloc_stat_field(value: &Value, metric: &str, field: &str) -> Option<u64> {
     match value {
         Value::Object(fields) => {
@@ -277,10 +271,12 @@ fn mimalloc_stat_field(value: &Value, metric: &str, field: &str) -> Option<u64>
     }
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn mimalloc_stat_current(value: &Value, metric: &str) -> Option<u64> {
     mimalloc_stat_field(value, metric, "current")
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn mimalloc_stat_sum(value: &Value, metrics: &[&str], field: &str) -> Option<u64> {
     metrics
         .iter()
@@ -289,6 +285,7 @@ fn mimalloc_stat_sum(value: &Value, metrics: &[&str], field: &str) -> Option<u64
         .filter(|value| *value > 0)
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn parse_mimalloc_stats_json(stats_json: &str) -> Option<AllocatorMemoryObservation> {
     let value = serde_json::from_str::<Value>(stats_json).ok()?;
     let malloc_metrics = ["malloc_normal", "malloc_huge"];
@@ -315,6 +312,33 @@ fn parse_mimalloc_stats_json(stats_json: &str) -> Option<AllocatorMemoryObservat
     }
 }
 
+#[cfg(not(target_os = "windows"))]
+#[allow(unsafe_code)]
+fn read_allocator_memory_snapshot() -> Option<AllocatorMemorySnapshot> {
+    // SAFETY: `mi_stats_get_json` returns a null-terminated JSON buffer owned by
+    // mimalloc when called with a null input buffer. The mimalloc API requires
+    // freeing that buffer with `mi_free`; parsing finishes before the buffer is freed.
+    let observation = unsafe {
+        let stats_ptr = libmimalloc_sys::mi_stats_get_json(0, std::ptr::null_mut());
+        if stats_ptr.is_null() {
+            return None;
+        }
+
+        let observation = CStr::from_ptr(stats_ptr).to_str().ok().and_then(parse_mimalloc_stats_json);
+        libmimalloc_sys::mi_free(stats_ptr.cast());
+        observation?
+    };
+    Some(AllocatorMemorySnapshot {
+        backend: crate::allocator_reclaim::allocator_backend(),
+        observation,
+    })
+}
+
+#[cfg(target_os = "windows")]
+fn read_allocator_memory_snapshot() -> Option<AllocatorMemorySnapshot> {
+    None
+}
+
 fn configured_memory_observability_interval_secs() -> u64 {
     rustfs_utils::get_env_u64(ENV_MEMORY_OBSERVABILITY_INTERVAL_SECS, DEFAULT_MEMORY_OBSERVABILITY_INTERVAL_SECS).max(1)
 }
@@ -542,13 +566,6 @@ mod tests {
         assert_eq!(parse_mimalloc_stats_json(r#"{ "allocator": "unknown" }"#), None);
     }
 
-    #[test]
-    fn read_allocator_memory_snapshot_uses_mimalloc_stats_json() {
-        let snapshot = super::read_allocator_memory_snapshot();
-        #[cfg(not(target_os = "windows"))]
-        assert!(snapshot.is_some(), "allocator snapshot should be available on non-Windows");
-    }
-
     #[test]
     fn memory_observability_snapshot_reports_disabled_when_metrics_are_disabled() {
         let snapshot = build_memory_observability_status_snapshot(false, 15, false);