fix(ecstore): migrate tier free versions during decommission

crates/ecstore/src/core/pools.rs

@@ -1130,21 +1130,29 @@ fn should_cleanup_decommission_source_entry(decommissioned: usize, total_version
     decommissioned.saturating_add(expired) == total_versions
 }
 
-/// Disposition reason logged for tier free-version records that decommission
+const DECOMMISSION_FREE_VERSION_MIGRATED_REASON: &str = "tier_free_version_migrated";
-/// skips instead of migrating.
+const DECOMMISSION_FREE_VERSION_RETAINED_REASON: &str = "tier_free_version_migration_failed";
-const DECOMMISSION_FREE_VERSION_SKIP_REASON: &str = "tier_free_version_not_migrated";
+const DECOMMISSION_FREE_VERSION_SWEEP_REASON: &str = "tier_free_version_unresolved_after_decommission";
+const DECOMMISSION_FREE_VERSION_DISPOSITION_REASON: &str = "tier_free_version_disposition_recorded";
 
-/// Counts the tier free-version records present in a decommission entry
+#[derive(Debug, Default, Clone, Copy, PartialEq, Eq)]
-/// inventory. The exact loader (`load_file_info_versions_exact`) keeps these
+struct DecommissionFreeVersionDisposition {
-/// records inline in `versions` instead of separating them into
+    migrated: usize,
-/// `free_versions`, and the migration loop then routes them through the
+    retained: usize,
-/// generic delete-marker path: the free-version flag and its remote-tier
+}
-/// identity are never carried to the target pool, and a lone record is skipped
+
-/// by the empty-delete-marker rule. Accounting for them here keeps the final
+impl DecommissionFreeVersionDisposition {
-/// sweep from silently omitting records whose free-version disposition was
+    fn record_migrated(&mut self) {
-/// dropped (see docs/architecture/decommission-compatibility.md).
+        self.migrated += 1;
-fn decommission_free_versions_skipped(fivs: &FileInfoVersions) -> usize {
+    }
-    fivs.versions.iter().filter(|version| version.tier_free_version()).count()
+
+    fn record_retained(&mut self) {
+        self.retained += 1;
+    }
+
+    fn total(self) -> usize {
+        self.migrated.saturating_add(self.retained)
+    }
 }
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
@@ -2478,6 +2486,7 @@ fn decommission_remote_tiered_opts(
         user_defined: version.metadata.clone(),
         src_pool_idx,
         data_movement: true,
+        incl_free_versions: version.tier_free_version(),
         include_part_checksums: true,
         http_preconditions: Some(crate::data_movement::data_movement_target_precondition()),
         expected_bucket_incarnation_id,
@@ -3114,24 +3123,9 @@ impl ECStore {
         fivs.versions
             .sort_by_key(|v| (v.mod_time.is_none(), std::cmp::Reverse(v.mod_time)));
 
-        let skipped_free_versions = decommission_free_versions_skipped(&fivs);
-        if skipped_free_versions > 0 {
-            debug!(
-                event = EVENT_DECOMMISSION_ENTRY,
-                component = LOG_COMPONENT_ECSTORE,
-                subsystem = LOG_SUBSYSTEM_POOLS,
-                pool_index = idx,
-                bucket = %bucket,
-                object = %entry.name,
-                skipped_free_versions,
-                reason = DECOMMISSION_FREE_VERSION_SKIP_REASON,
-                state = "free_versions_skipped",
-                "Decommission skipped free-version migration"
-            );
-        }
-
         let mut decommissioned: usize = 0;
         let mut expired: usize = 0;
+        let mut free_version_disposition = DecommissionFreeVersionDisposition::default();
         let mut cleanup_preflight_allowed_missing = Vec::new();
 
         for version in fivs.versions.iter() {
@@ -3140,6 +3134,80 @@ impl ECStore {
             }
             decommission_cancel_signal_result(rx.is_cancelled())?;
 
+            if version.tier_free_version() {
+                let version_id = version.version_id.map(|v| v.to_string());
+                let mut migration_error = None;
+                let mut migrated = false;
+                for _ in 0..3 {
+                    match self
+                        .decommission_tiered_object(
+                            bucket.as_str(),
+                            &version.name,
+                            version,
+                            &decommission_remote_tiered_opts(version, version_id.clone(), idx, expected_bucket_incarnation_id),
+                        )
+                        .await
+                    {
+                        Ok(()) => {
+                            migrated = true;
+                            migration_error = None;
+                            break;
+                        }
+                        Err(err) if is_decommission_target_capacity_error(&err) => {
+                            return Err(with_decommission_entry_context(
+                                "decommission_tier_free_version",
+                                bucket.as_str(),
+                                version.name.as_str(),
+                                err,
+                            ));
+                        }
+                        Err(err) => migration_error = Some(err),
+                    }
+                }
+
+                {
+                    let mut pool_meta = self.pool_meta.write().await;
+                    if let Err(err) = count_decommission_item(&mut pool_meta, idx, 0, !migrated) {
+                        return Err(with_decommission_entry_context(
+                            "count_decommission_item",
+                            bucket.as_str(),
+                            entry.name.as_str(),
+                            err,
+                        ));
+                    }
+                }
+
+                if migrated {
+                    decommissioned += 1;
+                    free_version_disposition.record_migrated();
+                } else {
+                    free_version_disposition.record_retained();
+                }
+
+                debug!(
+                    event = EVENT_DECOMMISSION_ENTRY,
+                    component = LOG_COMPONENT_ECSTORE,
+                    subsystem = LOG_SUBSYSTEM_POOLS,
+                    pool_index = idx,
+                    bucket = %bucket,
+                    object = %version.name,
+                    version_id = ?version_id,
+                    result = ?migration_error,
+                    reason = if migrated {
+                        DECOMMISSION_FREE_VERSION_MIGRATED_REASON
+                    } else {
+                        DECOMMISSION_FREE_VERSION_RETAINED_REASON
+                    },
+                    state = if migrated { "free_version_migrated" } else { "free_version_retained" },
+                    "Decommission free-version disposition recorded"
+                );
+
+                if !migrated {
+                    break;
+                }
+                continue;
+            }
+
             if should_skip_lifecycle_for_data_movement(
                 self.clone(),
                 &bucket,
@@ -3427,6 +3495,23 @@ impl ECStore {
             }
         }
 
+        if free_version_disposition.total() > 0 {
+            debug!(
+                event = EVENT_DECOMMISSION_ENTRY,
+                component = LOG_COMPONENT_ECSTORE,
+                subsystem = LOG_SUBSYSTEM_POOLS,
+                pool_index = idx,
+                bucket = %bucket,
+                object = %entry.name,
+                free_versions_migrated = free_version_disposition.migrated,
+                free_versions_retained = free_version_disposition.retained,
+                free_versions_total = free_version_disposition.total(),
+                reason = DECOMMISSION_FREE_VERSION_DISPOSITION_REASON,
+                state = "free_version_disposition",
+                "Decommission free-version disposition summary"
+            );
+        }
+
         if should_cleanup_decommission_source_entry(decommissioned, fivs.versions.len(), expired) {
             if bucket_incarnation_fence.as_ref().is_some_and(|guard| guard.is_lock_lost()) {
                 return Err(Error::other("decommission bucket incarnation fence was lost before source cleanup"));
@@ -4393,6 +4478,7 @@ impl ECStore {
                 let lifecycle_config_cb = lifecycle_config.clone();
                 let object_lock_config_cb = object_lock_config.clone();
                 let store = Arc::clone(self);
+                let set_cb = Arc::clone(set);
                 let callback_rx_cb = callback_rx.clone();
 
                 let callback: ListCallback = Arc::new(move |entry: MetaCacheEntry| {
@@ -4402,6 +4488,7 @@ impl ECStore {
                     let lifecycle_config = lifecycle_config_cb.clone();
                     let object_lock_config = object_lock_config_cb.clone();
                     let store = Arc::clone(&store);
+                    let set = Arc::clone(&set_cb);
                     let callback_rx = callback_rx_cb.clone();
                     Box::pin(async move {
                         if callback_rx.is_cancelled() {
@@ -4416,11 +4503,14 @@ impl ECStore {
                             return;
                         }
 
-                        let fivs = match load_decommission_entry_versions(
+                        let fivs = match load_decommission_entry_exact_versions(
+                            &set,
                             &entry,
                             &bucket_name,
                             "check_after_decommission.file_info_versions",
-                        ) {
+                        )
+                        .await
+                        {
                             Ok(fivs) => fivs,
                             Err(err) => {
                                 let mut first_err = entry_error.lock().await;
@@ -4433,7 +4523,23 @@ impl ECStore {
                         };
 
                         let mut remaining = 0;
-                        for version in &fivs.versions {
+                        for version in fivs.versions.iter().chain(fivs.free_versions.iter()) {
+                            if version.tier_free_version() {
+                                remaining += 1;
+                                debug!(
+                                    event = EVENT_DECOMMISSION_ENTRY,
+                                    component = LOG_COMPONENT_ECSTORE,
+                                    subsystem = LOG_SUBSYSTEM_POOLS,
+                                    pool_index = idx,
+                                    bucket = %bucket_name,
+                                    object = %entry.name,
+                                    version_id = ?version.version_id,
+                                    reason = DECOMMISSION_FREE_VERSION_SWEEP_REASON,
+                                    state = "free_version_retained",
+                                    "Decommission final sweep retained a free version"
+                                );
+                                continue;
+                            }
                             if version.deleted {
                                 continue;
                             }
@@ -4787,6 +4893,12 @@ mod tests {
         assert!(opts.include_part_checksums);
         assert!(opts.http_preconditions.is_some());
         assert_eq!(opts.expected_bucket_incarnation_id, Some(incarnation));
+        assert!(!opts.incl_free_versions);
+
+        let mut free_version = version;
+        free_version.set_tier_free_version();
+        let free_opts = decommission_remote_tiered_opts(&free_version, Some("free-version-id".to_string()), 9, Some(incarnation));
+        assert!(free_opts.incl_free_versions);
     }
 
     #[test]
@@ -5491,13 +5603,13 @@ pub(crate) fn fallback_free_capacity_dedup(disks: &[rustfs_madmin::Disk]) -> usi
 #[cfg(test)]
 mod pools_tests {
     use super::{
-        DECOMMISSION_FREE_VERSION_SKIP_REASON, DECOMMISSION_PROGRESS_SAVE_INTERVAL, DECOMMISSION_PROGRESS_SAVE_ITEM_THRESHOLD,
+        DECOMMISSION_FREE_VERSION_MIGRATED_REASON, DECOMMISSION_FREE_VERSION_RETAINED_REASON,
-        DECOMMISSION_PROGRESS_SAVE_RETRY_BACKOFF, DecomBucketInfo, DecommissionStartPoolState, DecommissionTerminalState,
+        DECOMMISSION_PROGRESS_SAVE_INTERVAL, DECOMMISSION_PROGRESS_SAVE_ITEM_THRESHOLD, DECOMMISSION_PROGRESS_SAVE_RETRY_BACKOFF,
-        ListCallback, PoolDecommissionInfo, PoolMeta, PoolSpaceInfo, PoolStatus, apply_decommission_status_space_info,
+        DecomBucketInfo, DecommissionFreeVersionDisposition, DecommissionStartPoolState, DecommissionTerminalState, ListCallback,
+        PoolDecommissionInfo, PoolMeta, PoolSpaceInfo, PoolStatus, apply_decommission_status_space_info,
         bind_decommission_cancelers, bind_missing_decommission_cancelers, cancel_decommission_canceler,
-        classify_decommission_terminal_state, count_decommission_item, decommission_cancel_signal_result,
+        classify_decommission_terminal_state, count_decommission_item, decommission_cancel_signal_result, decommission_item_size,
-        decommission_free_versions_skipped, decommission_item_size, decommission_meta_bucket_options,
+        decommission_meta_bucket_options, decommission_start_pool_state, dedup_indices, default_decommission_bucket_concurrency,
-        decommission_start_pool_state, dedup_indices, default_decommission_bucket_concurrency,
         ensure_decommission_cancel_allowed, ensure_decommission_clear_allowed, ensure_decommission_listing_disks_available,
         ensure_decommission_not_rebalancing, ensure_decommission_start_allowed, ensure_decommission_start_keeps_active_pool,
         ensure_decommission_start_local_leader, ensure_decommission_start_pool_states,
@@ -6797,24 +6909,16 @@ mod pools_tests {
     }
 
     #[test]
-    fn decommission_free_version_accounting_reports_skipped_records() {
+    fn decommission_free_version_accounting_records_migration_and_retention() {
-        let mut fivs = FileInfoVersions::default();
+        let mut disposition = DecommissionFreeVersionDisposition::default();
-        assert_eq!(decommission_free_versions_skipped(&fivs), 0);
+        disposition.record_migrated();
+        disposition.record_retained();
 
-        fivs.versions.push(FileInfo {
+        assert_eq!(disposition.migrated, 1);
-            name: "object.txt".to_string(),
+        assert_eq!(disposition.retained, 1);
-            ..Default::default()
+        assert_eq!(disposition.total(), 2);
-        });
+        assert_eq!(DECOMMISSION_FREE_VERSION_MIGRATED_REASON, "tier_free_version_migrated");
-        let mut free_one = FileInfo::default();
+        assert_eq!(DECOMMISSION_FREE_VERSION_RETAINED_REASON, "tier_free_version_migration_failed");
-        free_one.set_tier_free_version();
-        fivs.versions.push(free_one);
-        let mut free_two = FileInfo::default();
-        free_two.set_tier_free_version();
-        free_two.transition_tier = "WARM".to_string();
-        fivs.versions.push(free_two);
-
-        assert_eq!(decommission_free_versions_skipped(&fivs), 2);
-        assert_eq!(DECOMMISSION_FREE_VERSION_SKIP_REASON, "tier_free_version_not_migrated");
     }
 
     #[test]

crates/ecstore/src/set_disk/mod.rs

@@ -4642,6 +4642,63 @@ fn resolve_delete_version_state(opts: &ObjectOptions, goi: &ObjectInfo, version_
 }
 
 impl SetDisks {
+    /// Publish an internal tier free-version record without changing its
+    /// delete-marker shape or remote-tier identity. The caller holds the
+    /// source and target object locks; a write quorum is required before the
+    /// source cleanup may remove the original record.
+    #[tracing::instrument(skip(self, fi, opts))]
+    pub(crate) async fn decommission_tier_free_version(
+        &self,
+        bucket: &str,
+        object: &str,
+        fi: &FileInfo,
+        opts: &ObjectOptions,
+    ) -> Result<()> {
+        if !fi.deleted || !fi.tier_free_version() {
+            return Err(Error::other("decommission tier free-version write requires a free version record"));
+        }
+        if opts
+            .namespace_lock_fence
+            .as_ref()
+            .is_some_and(NamespaceLockFence::is_lock_lost)
+            || opts
+                .bucket_lifecycle_lock_fence
+                .as_ref()
+                .is_some_and(NamespaceLockFence::is_lock_lost)
+        {
+            return Err(StorageError::NamespaceLockQuorumUnavailable {
+                mode: "decommission_tier_free_version_commit",
+                bucket: bucket.to_string(),
+                object: object.to_string(),
+                required: 1,
+                achieved: 0,
+            });
+        }
+
+        let disks = self.disks.read().await.clone();
+        let write_quorum = self.default_write_quorum();
+        let futures = disks.into_iter().map(|disk| {
+            let file_info = fi.clone();
+            async move {
+                if let Some(disk) = disk {
+                    disk.write_metadata("", bucket, object, file_info).await
+                } else {
+                    Err(DiskError::DiskNotFound)
+                }
+            }
+        });
+
+        let mut errs = Vec::new();
+        for result in join_all(futures).await {
+            match result {
+                Ok(_) => errs.push(None),
+                Err(err) => errs.push(Some(err)),
+            }
+        }
+
+        resolve_tiered_decommission_write_quorum_result(&errs, write_quorum, bucket, object)
+    }
+
     #[tracing::instrument(skip(self, fi, opts))]
     pub(crate) async fn decommission_tiered_object(
         &self,
@@ -9864,6 +9921,57 @@ mod tests {
         assert_ne!(updated.erasure.distribution, original.erasure.distribution);
     }
 
+    #[tokio::test]
+    async fn decommission_tier_free_version_preserves_remote_identity() {
+        let set_disks = make_local_bucket_test_set_disks().await;
+        let bucket = "free-version-decommission";
+        let object = "object.txt";
+        let version_id = Uuid::new_v4();
+        let mut free_version = FileInfo {
+            name: object.to_string(),
+            volume: bucket.to_string(),
+            version_id: Some(version_id),
+            mod_time: Some(time::OffsetDateTime::now_utc()),
+            deleted: true,
+            transition_tier: "WARM-TIER".to_string(),
+            transitioned_objname: "remote/object".to_string(),
+            ..Default::default()
+        };
+        free_version.set_tier_free_version();
+
+        set_disks
+            .decommission_tier_free_version(bucket, object, &free_version, &ObjectOptions::default())
+            .await
+            .expect("free-version metadata should reach the target quorum");
+        set_disks
+            .decommission_tier_free_version(bucket, object, &free_version, &ObjectOptions::default())
+            .await
+            .expect("replaying the same free-version metadata should be idempotent");
+
+        let versions = set_disks
+            .load_file_info_versions_exact(bucket, object)
+            .await
+            .expect("migrated free-version metadata should decode")
+            .expect("migrated free-version metadata should exist");
+        let migrated = versions
+            .versions
+            .iter()
+            .find(|version| version.version_id == Some(version_id))
+            .expect("free version should be present on the target");
+
+        assert_eq!(
+            versions
+                .versions
+                .iter()
+                .filter(|version| version.version_id == Some(version_id))
+                .count(),
+            1
+        );
+        assert!(migrated.tier_free_version());
+        assert_eq!(migrated.transition_tier, "WARM-TIER");
+        assert_eq!(migrated.transitioned_objname, "remote/object");
+    }
+
     #[test]
     fn test_resolve_tiered_decommission_write_quorum_result_allows_successful_quorum() {
         let errs = vec![None, None, Some(DiskError::DiskNotFound), None];

crates/ecstore/src/store/object.rs

@@ -1152,6 +1152,15 @@ fn tiered_data_movement_source_matches(
         && expected_backend == current_backend)
 }
 
+fn decommission_free_version_overwrite_error(bucket: &str, object: &str, version_id: Option<Uuid>) -> Error {
+    StorageError::DataMovementOverwriteErr(
+        bucket.to_owned(),
+        object.to_owned(),
+        version_id.map(|id| id.to_string()).unwrap_or_default(),
+    )
+    .into()
+}
+
 fn should_check_data_movement_resume_target(src_pool_idx: usize, target_pool_idx: usize) -> bool {
     target_pool_idx != src_pool_idx
 }
@@ -1776,6 +1785,43 @@ impl ECStore {
         )
     }
 
+    async fn has_equivalent_data_movement_tier_free_version(
+        &self,
+        bucket: &str,
+        object: &str,
+        source: &rustfs_filemeta::FileInfo,
+        target_pool_idx: usize,
+    ) -> Result<bool> {
+        let pool = self
+            .pools
+            .get(target_pool_idx)
+            .ok_or_else(|| Error::other(format!("invalid tiered data movement target pool {target_pool_idx}")))?;
+        let logical_object = decode_dir_object(object);
+        let Some(versions) = pool
+            .get_disks_by_key(object)
+            .load_file_info_versions_exact(bucket, &logical_object)
+            .await?
+        else {
+            return Ok(false);
+        };
+
+        let Some(target) = versions
+            .versions
+            .iter()
+            .find(|version| version.version_id == source.version_id)
+        else {
+            return Ok(false);
+        };
+        if !target.tier_free_version() {
+            return Err(decommission_free_version_overwrite_error(bucket, object, source.version_id));
+        }
+        if tiered_data_movement_source_matches(source, target)? {
+            Ok(true)
+        } else {
+            Err(decommission_free_version_overwrite_error(bucket, object, source.version_id))
+        }
+    }
+
     fn resolve_decommission_target_pool_idx_result(result: Result<usize>, bucket: &str, object: &str) -> Result<usize> {
         result.map_err(|err| Error::other(format!("failed to select decommission target pool for {bucket}/{object}: {err}")))
     }
@@ -1795,6 +1841,10 @@ impl ECStore {
         check_put_object_args(bucket, object)?;
 
         let mut opts = opts.clone();
+        let is_free_version = fi.tier_free_version();
+        if is_free_version {
+            opts.incl_free_versions = true;
+        }
         let bucket_incarnation_fence = if is_meta_bucketname(bucket) {
             None
         } else {
@@ -1849,7 +1899,7 @@ impl ECStore {
                 versions
                     .versions
                     .iter()
-                    .find(|current| current.version_id == fi.version_id && !current.tier_free_version())
+                    .find(|current| current.version_id == fi.version_id && current.tier_free_version() == is_free_version)
             })
             .ok_or_else(|| to_object_err(StorageError::FileNotFound, vec![bucket, object.as_str()]))?;
         if !tiered_data_movement_source_matches(fi, current_source)? {
@@ -1864,24 +1914,40 @@ impl ECStore {
                 .get_available_pool_idx_excluding(bucket, &object, fi.size, opts.src_pool_idx)
                 .await;
             let target_pool_idx = resolve_data_movement_resume_target_pool(idx, resume_target_pool_idx, opts.src_pool_idx);
+            if is_free_version && target_pool_idx == opts.src_pool_idx {
+                return Err(Error::DiskFull);
+            }
+            let equivalent = if is_free_version {
+                self.has_equivalent_data_movement_tier_free_version(bucket, &object, &fi, target_pool_idx)
+                    .await?
+            } else {
+                self.has_equivalent_data_movement_tiered_object(bucket, &object, &fi, &opts, target_pool_idx)
+                    .await?
+            };
+            if equivalent {
+                return Ok(());
+            }
+
+            return Err(decommission_free_version_overwrite_error(bucket, &object, fi.version_id));
+        }
+
+        let result = if is_free_version {
             if self
-                .has_equivalent_data_movement_tiered_object(bucket, &object, &fi, &opts, target_pool_idx)
+                .has_equivalent_data_movement_tier_free_version(bucket, &object, &fi, idx)
                 .await?
             {
                 return Ok(());
             }
-
+            self.pools[idx]
-            return Err(StorageError::DataMovementOverwriteErr(
+                .get_disks_by_key(&object)
-                bucket.to_owned(),
+                .decommission_tier_free_version(bucket, &object, &fi, &opts)
-                object.to_owned(),
+                .await
-                opts.version_id.clone().unwrap_or_default(),
+        } else {
-            ));
+            self.pools[idx]
-        }
+                .get_disks_by_key(&object)
-
+                .decommission_tiered_object(bucket, &object, &fi, &opts)
-        let result = self.pools[idx]
+                .await
-            .get_disks_by_key(&object)
+        };
-            .decommission_tiered_object(bucket, &object, &fi, &opts)
-            .await;
         if matches!(result, Err(Error::PreconditionFailed)) {
             if self
                 .has_equivalent_data_movement_tiered_object(bucket, &object, &fi, &opts, idx)

crates/filemeta/src/filemeta.rs

@@ -102,10 +102,9 @@ pub const TRANSITION_PENDING: &str = "pending";
 /// delete paths the same obligation is also carried by a committed tier-journal
 /// entry; deletes without such an entry (for example a removed version whose
 /// transition state decodes as unknown) rely on this record alone until the
-/// worker removes it after a successful remote delete. Decommission does not
+/// worker removes it after a successful remote delete. Decommission preserves
-/// preserve these semantics: its exact inventory keeps the records inline in
+/// the record and its remote identity on the target pool before source cleanup
-/// `versions` and the migration loop treats them as ordinary delete markers —
+/// — see docs/architecture/decommission-compatibility.md.
-/// see docs/architecture/decommission-compatibility.md.
 pub const FREE_VERSION: &str = "free-version";
 
 pub const TRANSITION_STATUS: &str = "transition-status";

crates/filemeta/src/filemeta/version.rs

@@ -2730,10 +2730,10 @@ impl MetaObject {
     /// identity so the lifecycle worker can issue the idempotent remote delete
     /// and only then remove the record; until then the recovery scan and the
     /// usage scanner keep re-enqueueing it. S3 and lifecycle deletes also
-    /// persist a committed tier-journal entry for the same remote delete, so a
+    /// persist a committed tier-journal entry for the same remote delete. The
-    /// record destroyed without its remote delete (as decommission does when it
+    /// decommission path copies this record unchanged before source cleanup,
-    /// treats these records as ordinary delete markers) strands only the
+    /// including when the transition state is unknown — see
-    /// journal-less cases — see docs/architecture/decommission-compatibility.md.
+    /// docs/architecture/decommission-compatibility.md.
     pub fn init_free_version(&self, fi: &FileInfo) -> Result<(FileMetaVersion, bool)> {
         if fi.skip_tier_free_version() {
             return Ok((FileMetaVersion::default(), false));

docs/architecture/decommission-compatibility.md

@@ -188,55 +188,54 @@ them.
 ### Decommission Handling
 
 The exact decommission inventory loader (`load_file_info_versions_exact` via
-`get_all_file_info_versions`) keeps free-version records inline in `versions`; it
+`get_all_file_info_versions`) keeps free-version records inline in `versions`.
-never populates `free_versions`, so the source-cleanup preflight comparison of
+The migration loop handles them before lifecycle expiry and delete-marker
-`free_versions` is vacuous for decommission. The migration loop then routes every
+shortcuts. It selects a target pool using the free-version-aware lookup, then
-record through the generic delete-marker handling:
+writes the original free record to every target disk with the normal metadata
+write quorum. The free-version marker, local version id, transition identity,
+transition state, and destination id are preserved at the FileInfo/metadata
+boundary.
 
-- a record that is the only remaining version without replication is skipped by the
+The source record is physically removed only after the target write quorum has
-  empty-delete-marker rule and counted as done;
+committed and the source cleanup preflight still matches the exact inventory.
-- any other record is copied to the target pool as an ordinary delete marker with the
+If target capacity, metadata validation, lock fencing, or quorum fails, the
-  same version id and mod time.
+source record remains and the entry records `state = "free_version_retained"`
-
+with reason `tier_free_version_migration_failed`; the worker retries the
-In both cases the free-version flag and its remote-tier identity are dropped:
+operation on a later pass. A target record with the same version id is accepted
-decommission neither preserves free-version semantics nor performs or reschedules the
+only when its free-version identity matches; a conflicting ordinary version or
-pending remote-tier delete. Source cleanup then removes the original records together
+different free record is an overwrite error. This makes retries idempotent and
-with the source xl.meta.
+prevents a free record from replacing a user-visible version.
-
-Allowed physical-delete timing: the source record may be removed once the migration
-loop has dispositioned it (copied as a plain marker or skipped as lone), which
-happens regardless of whether its remote-tier delete was ever performed.
 
 ### Reference-Audit Result
 
-No cluster-local consumer resolves a free version after decommission finishes: GET,
+After migration, user-facing GET/list/transition/replication/restore paths still
-listing, transition planning, replication, restore, and heal operate either on
+exclude the record. Recovery, usage scanning, lifecycle tier cleanup, and heal
-user-visible versions or while the record still exists. The remote exposure is
+continue to see it when they request free versions, so an unresolved remote
-bounded:
+delete remains actionable on the target pool. The committed tier journal remains
+an independent retry source where one exists; it is not used as a reason to drop
+the xl.meta record. In particular, `Unknown` transition state records are
+migrated unchanged rather than discarded: the lifecycle worker retains them if
+remote identity validation cannot make a delete request.
 
-- On every user-facing delete path the remote-delete obligation is durably carried
+Each migrated record emits `state = "free_version_migrated"` with reason
-  by the committed tier-journal entry, which the tier sweeper processes
+`tier_free_version_migrated`. Each failed record emits the retained state and
-  independently of xl.meta; the free-version record is an idempotent second
+failure reason above. The entry also emits a disposition summary with migrated,
-  pointer, not the only one. Dropping it during decommission therefore does not
+retained, and total counts. The final decommission sweep uses the exact loader,
-  orphan the remote object.
+counts free records still present, and emits one retained record/reason for each
-- Residual exposure: for records whose version state decoded as `Unknown` no
+unresolved free version before failing the sweep. This makes both successful
-  journal entry exists, so dropping the unconsumed record loses that cleanup hint
+migration and retained cleanup obligations visible instead of silently omitting
-  and the remote-tier object is orphaned. The same applies to any future internal
+free records.
-  delete path that removes transitioned versions without a journal entry.
 
-Copying a pending record as an ordinary delete marker also adds a user-visible
+No new S3-visible version or admin response field is needed: free versions remain
-tombstone to the target pool's version history that the source never exposed.
+internal and are never counted as user-visible versions. The structured
-
+`decommission_entry` events are the operational status surface for the
-Because of the residual journal-less case, decommission must account for every
+free-version disposition; the existing decommission item/failed counters still
-free-version record instead of omitting it silently:
+report the enclosing object migration result.
-
-- `decommission_free_versions_skipped` counts the records per decommission entry;
-- entries with a non-zero count log `state = "free_versions_skipped"` with reason
-  `tier_free_version_not_migrated`.
 
 Regression guard:
 
-- `decommission_free_version_accounting_reports_skipped_records`
+- `decommission_free_version_accounting_records_migration_and_retention`
+- `decommission_tier_free_version_preserves_remote_identity`
 
 ## Regression Guard