ci: add s3tests upstream head canary

.github/workflows/e2e-s3tests.yml

@@ -21,6 +21,9 @@
 #     suite and reports promotion candidates. Regressions, unclassified tests,
 #     incomplete execution, and infrastructure errors fail the job; classified
 #     failures for not-yet-implemented features remain informational.
+#   - Non-blocking upstream HEAD canary: collects current upstream node IDs and
+#     reports new, removed, duplicate, or overlapping classifications without
+#     making upstream drift a release gate.
 #   - Manual runs (workflow_dispatch): same, with configurable mode/scope.
 #
 # All test execution is delegated to scripts/s3-tests/run.sh (single source of
@@ -354,6 +357,85 @@ jobs:
           name: s3tests-${{ env.TEST_MODE }}-shard-${{ matrix.shard-index }}
           path: artifacts/**
 
+  upstream-head-canary:
+    name: Upstream HEAD classification canary
+    if: github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
+    continue-on-error: true
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    steps:
+      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
+        with:
+          persist-credentials: false
+
+      - name: Set up Python
+        uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6.3.0
+        with:
+          python-version: "3.12"
+
+      - name: Install collection tool
+        run: |
+          python3 -m pip install --user "tox==4.60.0"
+          python3 - <<'PY'
+          from importlib.metadata import version
+
+          assert version("tox") == "4.60.0"
+          PY
+          echo "$HOME/.local/bin" >> "$GITHUB_PATH"
+
+      - name: Compare upstream HEAD classifications
+        id: upstream-compare
+        run: |
+          ARTIFACT_DIR="artifacts/s3tests-upstream-head"
+          UPSTREAM_DIR="${RUNNER_TEMP}/s3-tests-upstream"
+          mkdir -p "${ARTIFACT_DIR}"
+          git clone --depth 1 https://github.com/ceph/s3-tests.git "${UPSTREAM_DIR}"
+          git -C "${UPSTREAM_DIR}" rev-parse HEAD > "${ARTIFACT_DIR}/upstream-sha.txt"
+          cp "${UPSTREAM_DIR}/s3tests.conf.SAMPLE" "${UPSTREAM_DIR}/s3tests.conf"
+          (
+            cd "${UPSTREAM_DIR}"
+            S3TEST_CONF="${UPSTREAM_DIR}/s3tests.conf" tox -- \
+              -q --collect-only s3tests/functional/test_s3.py \
+              -m "not rustfs_never_marker"
+          ) 2>&1 | tee "${ARTIFACT_DIR}/collect.log"
+          grep -E '^s3tests/functional/test_s3\.py::' \
+            "${ARTIFACT_DIR}/collect.log" > "${ARTIFACT_DIR}/collected-nodeids.txt"
+          python3 scripts/s3-tests/report_compat.py \
+            --lists-dir scripts/s3-tests \
+            --collected-nodeids "${ARTIFACT_DIR}/collected-nodeids.txt" \
+            --check-classifications-only 2>&1 | tee "${ARTIFACT_DIR}/classification-drift.txt"
+
+      - name: Publish canary report
+        if: always()
+        env:
+          CANARY_OUTCOME: ${{ steps.upstream-compare.outcome }}
+        run: |
+          {
+            echo "## ceph/s3-tests upstream HEAD canary"
+            echo
+            if [ -f artifacts/s3tests-upstream-head/upstream-sha.txt ]; then
+              echo "Upstream HEAD: $(cat artifacts/s3tests-upstream-head/upstream-sha.txt)"
+            fi
+            echo
+            echo '```text'
+            if [ -s artifacts/s3tests-upstream-head/classification-drift.txt ]; then
+              cat artifacts/s3tests-upstream-head/classification-drift.txt
+            elif [ "${CANARY_OUTCOME}" != "success" ]; then
+              echo "Canary did not complete; inspect the collection log artifact."
+            else
+              echo "No classification drift detected."
+            fi
+            echo '```'
+          } >> "$GITHUB_STEP_SUMMARY"
+
+      - name: Upload canary artifacts
+        if: always() && env.ACT != 'true'
+        uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
+        with:
+          name: s3tests-upstream-head
+          path: artifacts/s3tests-upstream-head/**
+          retention-days: 14
+
   alert-on-failure:
     name: Alert on scheduled failure
     needs: [s3tests]

Cargo.lock

@@ -12688,7 +12688,6 @@ dependencies = [
  "js-sys",
  "rand 0.10.2",
  "serde_core",
- "sha1_smol",
  "wasm-bindgen",
 ]
 

crates/ecstore/src/core/pools.rs

@@ -2026,7 +2026,7 @@ impl PoolMeta {
         self.load_no_lock(pool).await
     }
 
-    pub(crate) async fn load_no_lock<S>(&mut self, pool: Arc<S>) -> Result<()>
+    async fn load_no_lock<S>(&mut self, pool: Arc<S>) -> Result<()>
     where
         S: EcstoreObjectIO,
     {

crates/ecstore/src/core/sets.rs

@@ -988,11 +988,14 @@ impl crate::storage_api_contracts::multipart::MultipartOperations for Sets {
     }
 }
 
-impl Sets {
-    pub(crate) async fn heal_format_with_fence<F>(&self, dry_run: bool, fence_lost: F) -> Result<(HealResultItem, Option<Error>)>
-    where
-        F: Fn() -> bool + Send + Sync,
-    {
+#[async_trait::async_trait]
+impl crate::storage_api_contracts::heal::HealOperations for Sets {
+    type Error = Error;
+    type HealResultItem = HealResultItem;
+    type HealOptions = HealOpts;
+
+    #[tracing::instrument(skip(self))]
+    async fn heal_format(&self, dry_run: bool) -> Result<(HealResultItem, Option<Error>)> {
         let (disks, init_errs) = init_storage_disks_with_errors(
             &self.endpoints.endpoints,
             &DiskOption {
@@ -1065,9 +1068,6 @@ impl Sets {
             // Save new formats `format.json` on unformatted disks.
             for (index, (fm, disk)) in tmp_new_formats.iter_mut().zip(disks.iter()).enumerate() {
                 if fm.is_some() && disk.is_some() {
-                    if fence_lost() {
-                        return Ok((res, Some(StorageError::SlowDown)));
-                    }
                     if let Err(err) = save_format_file(disk, fm).await {
                         if let Some(disk) = disk.as_ref() {
                             let _ = disk.close().await;
@@ -1101,18 +1101,6 @@ impl Sets {
         }
         Ok((res, None))
     }
-}
-
-#[async_trait::async_trait]
-impl crate::storage_api_contracts::heal::HealOperations for Sets {
-    type Error = Error;
-    type HealResultItem = HealResultItem;
-    type HealOptions = HealOpts;
-
-    #[tracing::instrument(skip(self))]
-    async fn heal_format(&self, dry_run: bool) -> Result<(HealResultItem, Option<Error>)> {
-        self.heal_format_with_fence(dry_run, || false).await
-    }
     #[tracing::instrument(skip(self))]
     async fn heal_bucket(&self, bucket: &str, opts: &HealOpts) -> Result<HealResultItem> {
         let mut result = HealResultItem {

crates/ecstore/src/store/heal.rs

@@ -13,12 +13,7 @@
 // limitations under the License.
 
 use super::*;
-use crate::core::pools::POOL_META_NAME;
-use crate::services::rebalance::{REBAL_META_NAME, RebalStatus};
-use crate::set_disk::get_lock_acquire_timeout;
 use crate::storage_api_contracts::heal::HealOperations as _;
-use crate::storage_api_contracts::namespace::NamespaceLocking as _;
-use rustfs_lock::NamespaceLockGuard;
 use tracing::trace;
 
 const LOG_COMPONENT_ECSTORE: &str = "ecstore";
@@ -35,119 +30,7 @@ fn invalid_heal_pool_index(pool_idx: usize, pool_count: usize) -> Error {
     )
 }
 
-#[derive(Debug, Clone, Copy)]
-enum HealFormatPoolSkip {
-    Completed,
-    Retryable,
-}
-
-fn classify_heal_format_pool(
-    pool_idx: usize,
-    pool_cmd_line: &str,
-    pool_meta: &PoolMeta,
-    rebalance_meta: Option<&RebalanceMeta>,
-) -> Option<HealFormatPoolSkip> {
-    let Some(pool) = pool_meta.pools.get(pool_idx) else {
-        return Some(HealFormatPoolSkip::Retryable);
-    };
-
-    if pool.id != pool_idx || pool_cmd_line.is_empty() || pool.cmd_line.is_empty() || pool.cmd_line != pool_cmd_line {
-        return Some(HealFormatPoolSkip::Retryable);
-    }
-
-    if let Some(decommission) = pool.decommission.as_ref() {
-        if decommission.complete {
-            return Some(HealFormatPoolSkip::Completed);
-        }
-        if decommission.failed || decommission.canceled || decommission.queued || pool_meta.is_suspended(pool_idx) {
-            return Some(HealFormatPoolSkip::Retryable);
-        }
-    }
-
-    if let Some(meta) = rebalance_meta {
-        let Some(pool_stats) = meta.pool_stats.get(pool_idx) else {
-            return Some(HealFormatPoolSkip::Retryable);
-        };
-        if pool_stats.info.stopping || (pool_stats.participating && pool_stats.info.status == RebalStatus::Started) {
-            return Some(HealFormatPoolSkip::Retryable);
-        }
-    }
-
-    None
-}
-
-fn heal_format_pool_skip_error(skip: HealFormatPoolSkip) -> Error {
-    match skip {
-        HealFormatPoolSkip::Completed => StorageError::NoHealRequired,
-        HealFormatPoolSkip::Retryable => StorageError::SlowDown,
-    }
-}
-
-fn heal_format_fence_lost_error() -> Error {
-    StorageError::SlowDown
-}
-
 impl ECStore {
-    async fn acquire_heal_format_fence(
-        &self,
-    ) -> Result<(NamespaceLockGuard, NamespaceLockGuard, PoolMeta, Option<RebalanceMeta>)> {
-        let metadata_pool = self
-            .pools
-            .first()
-            .cloned()
-            .ok_or_else(|| Error::other("heal format requires at least one storage pool"))?;
-
-        // Metadata fence order is part of the decommission/rebalance protocol:
-        // pool.bin must always be acquired before rebalance.bin.
-        let pool_lock = metadata_pool.new_ns_lock(RUSTFS_META_BUCKET, POOL_META_NAME).await?;
-        let pool_guard = pool_lock.get_write_lock(get_lock_acquire_timeout()).await?;
-        let rebalance_lock = metadata_pool.new_ns_lock(RUSTFS_META_BUCKET, REBAL_META_NAME).await?;
-        let rebalance_guard = rebalance_lock.get_write_lock(get_lock_acquire_timeout()).await?;
-
-        if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
-            return Err(heal_format_fence_lost_error());
-        }
-
-        let mut pool_meta = PoolMeta::default();
-        pool_meta.load_no_lock(metadata_pool.clone()).await?;
-        if pool_meta.pools.len() != self.pools.len()
-            || pool_meta.pools.iter().enumerate().any(|(pool_idx, pool)| {
-                pool.id != pool_idx || pool.cmd_line.is_empty() || pool.cmd_line != self.pools[pool_idx].endpoints.cmd_line
-            })
-        {
-            return Err(heal_format_fence_lost_error());
-        }
-
-        let mut rebalance_meta = RebalanceMeta::new();
-        let rebalance_meta = match rebalance_meta
-            .load_with_opts(
-                metadata_pool,
-                ObjectOptions {
-                    no_lock: true,
-                    ..Default::default()
-                },
-            )
-            .await
-        {
-            Ok(()) => Some(rebalance_meta),
-            Err(Error::ConfigNotFound) => None,
-            Err(err) => return Err(err),
-        };
-
-        if rebalance_meta
-            .as_ref()
-            .is_some_and(|meta| meta.pool_stats.len() != self.pools.len())
-        {
-            return Err(heal_format_fence_lost_error());
-        }
-
-        if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
-            return Err(heal_format_fence_lost_error());
-        }
-
-        Ok((pool_guard, rebalance_guard, pool_meta, rebalance_meta))
-    }
-
     fn get_pools_for_heal_object(&self, opts: &HealOpts) -> Result<Vec<Arc<Sets>>> {
         match opts.pool {
             Some(pool_idx) => Ok(vec![
@@ -169,26 +52,9 @@ impl ECStore {
         };
 
         let mut count_no_heal = 0;
-        let mut count_completed = 0;
         let mut first_error = None;
-        for (pool_idx, pool) in self.pools.iter().enumerate() {
-            let (pool_guard, rebalance_guard, pool_meta, rebalance_meta) = self.acquire_heal_format_fence().await?;
-            if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
-                first_error.get_or_insert(heal_format_fence_lost_error());
-                break;
-            }
-            if let Some(skip) = classify_heal_format_pool(pool_idx, &pool.endpoints.cmd_line, &pool_meta, rebalance_meta.as_ref())
-            {
-                if matches!(skip, HealFormatPoolSkip::Completed) {
-                    count_completed += 1;
-                } else {
-                    first_error.get_or_insert(heal_format_pool_skip_error(skip));
-                }
-                continue;
-            }
-
-            let fence_lost = || pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost();
-            let (mut result, err) = pool.heal_format_with_fence(dry_run, fence_lost).await?;
+        for pool in self.pools.iter() {
+            let (mut result, err) = pool.heal_format(dry_run).await?;
             if let Some(err) = err {
                 match err {
                     StorageError::NoHealRequired => {
@@ -203,18 +69,11 @@ impl ECStore {
             r.set_count += result.set_count;
             r.before.drives.append(&mut result.before.drives);
             r.after.drives.append(&mut result.after.drives);
-
-            // A lease can be lost after the final write; fail closed before
-            // reporting the pool as successfully healed.
-            if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
-                first_error.get_or_insert(heal_format_fence_lost_error());
-                break;
-            }
         }
         if let Some(err) = first_error {
             return Ok((r, Some(err)));
         }
-        if count_no_heal + count_completed == self.pools.len() {
+        if count_no_heal == self.pools.len() {
             info!(
                 event = EVENT_HEAL_FORMAT_COMPLETED,
                 component = LOG_COMPONENT_ECSTORE,
@@ -443,7 +302,6 @@ mod tests {
     use crate::disk::{DeleteOptions, DiskOption, format::FormatV3, new_disk};
     use crate::layout::endpoints::{EndpointServerPools, Endpoints, PoolEndpoints};
     use crate::runtime::instance::InstanceContext;
-    use crate::services::rebalance::{RebalanceInfo, RebalanceStats};
     use crate::storage_api_contracts::bucket::{BucketOperations, MakeBucketOptions};
     use crate::storage_api_contracts::object::{ObjectIO as _, ObjectOperations};
     use crate::store::init_format::{load_format_erasure, save_format_file};
@@ -495,164 +353,6 @@ mod tests {
         }
     }
 
-    fn pool_meta_with_decommission(info: PoolDecommissionInfo) -> PoolMeta {
-        PoolMeta {
-            pools: vec![PoolStatus {
-                id: 0,
-                cmd_line: "pool-0".to_string(),
-                last_update: OffsetDateTime::UNIX_EPOCH,
-                decommission: Some(info),
-            }],
-            ..Default::default()
-        }
-    }
-
-    #[test]
-    fn heal_format_pool_state_barriers_are_classified() {
-        let active = pool_meta_with_decommission(PoolDecommissionInfo {
-            start_time: Some(OffsetDateTime::UNIX_EPOCH),
-            ..Default::default()
-        });
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &active, None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        for info in [
-            PoolDecommissionInfo {
-                failed: true,
-                ..Default::default()
-            },
-            PoolDecommissionInfo {
-                canceled: true,
-                ..Default::default()
-            },
-        ] {
-            assert!(matches!(
-                classify_heal_format_pool(0, "pool-0", &pool_meta_with_decommission(info), None),
-                Some(HealFormatPoolSkip::Retryable)
-            ));
-        }
-
-        let completed = pool_meta_with_decommission(PoolDecommissionInfo {
-            complete: true,
-            ..Default::default()
-        });
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &completed, None),
-            Some(HealFormatPoolSkip::Completed)
-        ));
-    }
-
-    #[test]
-    fn heal_format_pool_rebalance_barriers_and_identity_are_fail_closed() {
-        let identity_meta = pool_meta_with_decommission(PoolDecommissionInfo::default());
-        let rebalance = RebalanceMeta {
-            pool_stats: vec![RebalanceStats {
-                participating: true,
-                info: RebalanceInfo {
-                    status: RebalStatus::Started,
-                    ..Default::default()
-                },
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&rebalance)),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        let stopping = RebalanceMeta {
-            pool_stats: vec![RebalanceStats {
-                info: RebalanceInfo {
-                    stopping: true,
-                    ..Default::default()
-                },
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&stopping)),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        let identity = pool_meta_with_decommission(PoolDecommissionInfo::default());
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-new", &identity, None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        let identity_without_decommission = PoolMeta {
-            pools: vec![PoolStatus {
-                id: 0,
-                cmd_line: "pool-0".to_string(),
-                last_update: OffsetDateTime::UNIX_EPOCH,
-                decommission: None,
-            }],
-            ..Default::default()
-        };
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-new", &identity_without_decommission, None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        assert!(matches!(
-            classify_heal_format_pool(0, "", &identity_meta, None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &PoolMeta::default(), None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        let stopped = RebalanceMeta {
-            stopped_at: Some(OffsetDateTime::UNIX_EPOCH),
-            pool_stats: vec![RebalanceStats {
-                participating: true,
-                info: RebalanceInfo {
-                    status: RebalStatus::Stopped,
-                    ..Default::default()
-                },
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-        assert!(classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&stopped)).is_none());
-
-        let stopping_after_stop = RebalanceMeta {
-            stopped_at: Some(OffsetDateTime::UNIX_EPOCH),
-            pool_stats: vec![RebalanceStats {
-                participating: true,
-                info: RebalanceInfo {
-                    status: RebalStatus::Started,
-                    stopping: true,
-                    ..Default::default()
-                },
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&stopping_after_stop)),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-    }
-
-    #[test]
-    fn skipped_heal_format_pool_is_never_reported_as_success() {
-        assert!(matches!(
-            heal_format_pool_skip_error(HealFormatPoolSkip::Retryable),
-            StorageError::SlowDown
-        ));
-        assert!(matches!(
-            heal_format_pool_skip_error(HealFormatPoolSkip::Completed),
-            StorageError::NoHealRequired
-        ));
-    }
-
     async fn multi_pool_heal_store() -> (tempfile::TempDir, Arc<ECStore>, CancellationToken) {
         let temp_dir = tempfile::tempdir().expect("multi-pool heal test directory should be created");
         let mut pool_endpoints = Vec::new();
@@ -1189,18 +889,6 @@ mod tests {
             bucket_fence_registry: std::sync::Arc::default(),
         };
 
-        let err = store
-            .handle_heal_format(false)
-            .await
-            .expect_err("missing pool metadata must fail closed before format writes");
-        assert!(matches!(err, StorageError::SlowDown));
-
-        let pool_meta = PoolMeta::new(&store.pools, &PoolMeta::default());
-        pool_meta
-            .save(store.pools.clone())
-            .await
-            .expect("pool metadata should be persisted before format heal");
-
         let (result, err) = store
             .handle_heal_format(false)
             .await
@@ -1214,22 +902,5 @@ mod tests {
             .await
             .expect("the later pool should be healed despite the first pool error");
         assert_eq!(healed.erasure.this, recoverable_format.erasure.sets[0][2]);
-
-        let mut completed_meta = PoolMeta::new(&store.pools, &PoolMeta::default());
-        for status in &mut completed_meta.pools {
-            status.decommission = Some(PoolDecommissionInfo {
-                complete: true,
-                ..Default::default()
-            });
-        }
-        completed_meta
-            .save(store.pools.clone())
-            .await
-            .expect("completed pool metadata should be persisted");
-        let (_, err) = store
-            .handle_heal_format(false)
-            .await
-            .expect("completed pools should be reported as a no-op");
-        assert!(matches!(err, Some(StorageError::NoHealRequired)));
     }
 }

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/task/heal_erasure_set.rs

@@ -231,10 +231,6 @@ impl HealTask {
                             "Heal erasure set format repair skipped because no format heal was required"
                         );
                     } else {
-                        let error = e;
-                        if error.is_recoverable_heal() {
-                            return Err(error);
-                        }
                         error!(
                             target: "rustfs::heal::task",
                             event = EVENT_HEAL_ERASURE_SET_RESULT,
@@ -243,7 +239,7 @@ impl HealTask {
                             task_id = %self.id,
                             set_disk_id,
                             result = "format_failed",
-                            error = %error,
+                            error = %e,
                             "Heal erasure set failed"
                         );
                         {
@@ -251,7 +247,7 @@ impl HealTask {
                             progress.update_progress(4, 4, 0, 0);
                         }
                         return Err(Error::TaskExecutionFailed {
-                            message: format!("Failed to heal disk format for {set_disk_id}: {error}"),
+                            message: format!("Failed to heal disk format for {set_disk_id}: {e}"),
                         });
                     }
                 } else {
@@ -288,9 +284,6 @@ impl HealTask {
             Err(Error::TaskCancelled) => return Err(Error::TaskCancelled),
             Err(Error::TaskTimeout) => return Err(Error::TaskTimeout),
             Err(e) => {
-                if e.is_recoverable_heal() {
-                    return Err(e);
-                }
                 error!(
                     target: "rustfs::heal::task",
                     event = EVENT_HEAL_ERASURE_SET_RESULT,

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

@@ -547,7 +547,6 @@ struct MockStorage {
     heal_object_outcome: Mutex<Option<MockHealObjectOutcome>>,
     heal_object_outcomes: Mutex<HashMap<String, VecDeque<MockHealObjectOutcome>>>,
     format_no_heal_required: Mutex<bool>,
-    format_error: Mutex<Option<Error>>,
     global_format_calls: Mutex<u32>,
     replacement_format_calls: Mutex<Vec<(usize, usize, Vec<String>)>>,
     replacement_targets_ready: Mutex<bool>,
@@ -868,9 +867,6 @@ impl HealStorageAPI for MockStorage {
 
     async fn heal_format(&self, _dry_run: bool) -> Result<(HealResultItem, Option<Error>)> {
         *self.global_format_calls.lock().unwrap() += 1;
-        if let Some(error) = self.format_error.lock().unwrap().take() {
-            return Err(error);
-        }
         let no_heal_required = *self.format_no_heal_required.lock().unwrap();
         if no_heal_required {
             Ok((HealResultItem::default(), Some(Error::Storage(EcstoreError::NoHealRequired))))
@@ -2056,30 +2052,6 @@ async fn test_erasure_set_heal_continues_after_format_no_heal_required() {
     );
 }
 
-#[tokio::test]
-async fn erasure_set_format_slowdown_is_propagated() {
-    let storage = Arc::new(MockStorage {
-        format_error: Mutex::new(Some(Error::Storage(EcstoreError::SlowDown))),
-        ..Default::default()
-    });
-    let request = HealRequest::new(
-        HealType::ErasureSet {
-            buckets: Vec::new(),
-            set_disk_id: "pool_0_set_0".to_string(),
-        },
-        HealOptions::default(),
-        HealPriority::Normal,
-    );
-    let task = HealTask::from_request(request, storage);
-
-    let error = task
-        .execute()
-        .await
-        .expect_err("format SlowDown must remain recoverable for the task manager");
-
-    assert!(matches!(error, Error::Storage(EcstoreError::SlowDown)));
-}
-
 #[tokio::test]
 async fn erasure_set_bucket_prepass_failure_stops_before_object_heal() {
     let temp = TempDir::new().expect("temporary directory should be created");

crates/test-utils/src/lib.rs

@@ -245,18 +245,6 @@ impl TestECStoreEnvBuilder {
             .await
             .expect("build test ECStore");
 
-        // The production bootstrap only persists pool.bin from the elected
-        // first cluster node.  Test stores intentionally have no cluster
-        // election, but heal-format still requires that durable fence before
-        // it can write any disk format.  Materialize the validated topology
-        // here so the shared fixture models a ready single-node store.
-        let mut pool_meta = ecstore.pool_meta.read().await.clone();
-        pool_meta.dont_save = false;
-        pool_meta
-            .save(ecstore.pools.clone())
-            .await
-            .expect("persist test pool metadata");
-
         if self.init_bucket_metadata {
             let buckets_list = ecstore
                 .list_bucket(&BucketOptions {

rustfs/Cargo.toml

@@ -322,7 +322,7 @@ thiserror = { workspace = true }
 tracing.workspace = true
 url = { workspace = true }
 urlencoding = { workspace = true }
-uuid = { workspace = true, features = ["v4", "v5", "fast-rng", "macro-diagnostics"] }
+uuid = { workspace = true, features = ["v4", "fast-rng", "macro-diagnostics"] }
 zip = { workspace = true }
 libc = { workspace = true }
 rand = { workspace = true, features = ["serde"] }
@@ -345,7 +345,7 @@ libsystemd.workspace = true
 libmimalloc-sys.workspace = true
 
 [dev-dependencies]
-uuid = { workspace = true, features = ["v4", "v5", "fast-rng", "macro-diagnostics"] }
+uuid = { workspace = true, features = ["v4", "fast-rng", "macro-diagnostics"] }
 serial_test = { workspace = true }
 tempfile = { workspace = true }
 aws-config = { workspace = true }

rustfs/src/admin/handlers/site_replication.rs

@@ -41,7 +41,7 @@ use crate::admin::storage_api::config::save_admin_config;
 use crate::admin::storage_api::contract::bucket::{
     BucketOperations, BucketOptions, DeleteBucketOptions, MakeBucketOptions, SRBucketDeleteOp,
 };
-use crate::admin::storage_api::error::{Error as StorageError, is_err_bucket_not_found};
+use crate::admin::storage_api::error::Error as StorageError;
 use crate::admin::storage_api::runtime::ECStore;
 use crate::admin::utils::{encode_compatible_admin_payload, read_compatible_admin_body};
 use crate::auth::constant_time_eq;
@@ -55,7 +55,6 @@ use crate::storage::storage_api::{
 use base64::Engine;
 use base64::engine::general_purpose::STANDARD as BASE64_STANDARD;
 use base64::engine::general_purpose::URL_SAFE_NO_PAD;
-use futures::StreamExt;
 use hmac::{Hmac, Mac};
 use http::header::{CONTENT_TYPE, HOST};
 use http::{HeaderMap, HeaderValue, Uri};
@@ -2097,18 +2096,6 @@ async fn remote_add_preflight_info(site: &PeerSite) -> S3Result<SiteReplicationA
             format!("invalid site replication metainfo from `{}`: {e}", site.endpoint),
         )
     })?;
-    if info.deployment_id.is_empty() {
-        // The peer will be tracked under a locally derived fallback ID
-        // (deployment_id_for_endpoint) instead of its real deployment ID.
-        warn!(
-            event = EVENT_ADMIN_SITE_REPLICATION_STATE,
-            component = LOG_COMPONENT_ADMIN,
-            subsystem = LOG_SUBSYSTEM_SITE_REPLICATION,
-            result = "peer_deployment_id_missing",
-            peer_endpoint = %site.endpoint,
-            "admin site replication state"
-        );
-    }
 
     let idp_body = send_peer_admin_get_request_with_client(
         &client,
@@ -2219,30 +2206,20 @@ fn site_replication_bootstrap_token(uri: &Uri) -> Option<String> {
     query_pairs(uri).get("bootstrapToken").cloned()
 }
 
-/// Query for a peer `make-with-versioning` bucket op. `versioningEnabled`
-/// always travels so the outbound query matches MinIO's site-replication
-/// make-bucket wire contract: MinIO's own create-bucket hook sends
-/// `versioningEnabled=true` on this op. RustFS's inbound handler
-/// force-enables versioning either way.
-fn make_with_versioning_bucket_op_path(bucket: &str, created_at: Option<&str>, lock_enabled: bool) -> String {
+fn bootstrap_bucket_make_op_path(bucket: &SRBucketInfo) -> String {
     let mut query = form_urlencoded::Serializer::new(String::new());
-    query.append_pair("bucket", bucket);
-    query.append_pair("operation", SITE_REPLICATION_BUCKET_OP_MAKE_WITH_VERSIONING);
-    query.append_pair("versioningEnabled", "true");
-    if let Some(created_at) = created_at {
-        query.append_pair("createdAt", created_at);
+    query.append_pair("bucket", &bucket.bucket);
+    query.append_pair("operation", "make-with-versioning");
+    if let Some(created_at) = bucket
+        .created_at
+        .and_then(|value| value.format(&time::format_description::well_known::Rfc3339).ok())
+    {
+        query.append_pair("createdAt", &created_at);
     }
-    if lock_enabled {
+    if bucket.object_lock_config.is_some() {
         query.append_pair("lockEnabled", "true");
     }
-    format!("{SITE_REPLICATION_PEER_BUCKET_OPS_PATH}?{}", query.finish())
-}
-
-fn bootstrap_bucket_make_op_path(bucket: &SRBucketInfo) -> String {
-    let created_at = bucket
-        .created_at
-        .and_then(|value| value.format(&time::format_description::well_known::Rfc3339).ok());
-    make_with_versioning_bucket_op_path(&bucket.bucket, created_at.as_deref(), bucket.object_lock_config.is_some())
+    format!("/rustfs/admin/v3/site-replication/peer/bucket-ops?{}", query.finish())
 }
 
 fn bootstrap_bucket_meta_item(bucket: &SRBucketInfo, item_type: &str, updated_at: Option<OffsetDateTime>) -> SRBucketMeta {
@@ -4269,7 +4246,16 @@ async fn broadcast_site_replication_make_bucket(
         .format(&time::format_description::well_known::Rfc3339)
         .unwrap_or_default();
 
-    let path = make_with_versioning_bucket_op_path(bucket, Some(&created_at), lock_enabled);
+    let path = {
+        let mut query = form_urlencoded::Serializer::new(String::new());
+        query.append_pair("bucket", bucket);
+        query.append_pair("operation", "make-with-versioning");
+        query.append_pair("createdAt", &created_at);
+        if lock_enabled {
+            query.append_pair("lockEnabled", "true");
+        }
+        format!("/rustfs/admin/v3/site-replication/peer/bucket-ops?{}", query.finish())
+    };
     let path = if let Some(token) = bootstrap_token {
         with_site_replication_bootstrap_token(&path, token)
     } else {
@@ -10220,25 +10206,13 @@ impl Operation for SiteReplicationStatusHandler {
     }
 }
 
-/// `POST /v3/site-replication/devnull` — peer link-check upload drain.
-/// MinIO streams multi-megabyte probe bodies here during site netperf link
-/// checks and expects an unbounded discard (its handler copies to io.Discard);
-/// buffering through the 1MB admin body cap turned any larger probe into a
-/// 400 and a false link failure. Stream and discard instead — no size cap.
-async fn drain_site_replication_devnull(mut input: Body) -> S3Result<()> {
-    while let Some(chunk) = input.next().await {
-        chunk.map_err(|e| s3_error!(InvalidRequest, "failed to read devnull stream: {}", e))?;
-    }
-    Ok(())
-}
-
 pub struct SiteReplicationDevNullHandler {}
 
 #[async_trait::async_trait]
 impl Operation for SiteReplicationDevNullHandler {
     async fn call(&self, req: S3Request<Body>, _params: Params<'_, '_>) -> S3Result<S3Response<(StatusCode, Body)>> {
         validate_site_replication_admin_request(&req, AdminAction::SiteReplicationOperationAction).await?;
-        drain_site_replication_devnull(req.input).await?;
+        let _ = read_plain_admin_body(req.input).await?;
         Ok(empty_response(StatusCode::NO_CONTENT))
     }
 }
@@ -10497,19 +10471,6 @@ impl Operation for SRPeerJoinHandler {
     }
 }
 
-/// Outcome of a peer-driven `purge-deleted-bucket` replay. A bucket that is
-/// already gone means the purge raced an earlier replay or a local delete —
-/// that is success — but any other failure must reach the sender like the
-/// sibling delete branches do: swallowing it answered 200 while the bucket
-/// survived on this site.
-fn purge_deleted_bucket_result(result: Result<(), StorageError>) -> S3Result<()> {
-    match result {
-        Ok(()) => Ok(()),
-        Err(err) if is_err_bucket_not_found(&err) => Ok(()),
-        Err(err) => Err(ApiError::from(err).into()),
-    }
-}
-
 pub struct SRPeerBucketOpsHandler {}
 
 #[async_trait::async_trait]
@@ -10609,18 +10570,16 @@ impl Operation for SRPeerBucketOpsHandler {
                     .map_err(ApiError::from)?;
             }
             "purge-deleted-bucket" => {
-                purge_deleted_bucket_result(
-                    store
-                        .delete_bucket(
-                            &bucket,
-                            &DeleteBucketOptions {
-                                force: true,
-                                srdelete_op: SRBucketDeleteOp::Purge,
-                                ..Default::default()
-                            },
-                        )
-                        .await,
-                )?;
+                let _ = store
+                    .delete_bucket(
+                        &bucket,
+                        &DeleteBucketOptions {
+                            force: true,
+                            srdelete_op: SRBucketDeleteOp::Purge,
+                            ..Default::default()
+                        },
+                    )
+                    .await;
             }
             _ => return Err(s3_error!(InvalidRequest, "unsupported site replication bucket operation")),
         }
@@ -13966,54 +13925,6 @@ mod tests {
         assert!(!query_flag(&uri, "missing"));
     }
 
-    /// A5 red-light: a `purge-deleted-bucket` replay must report success when
-    /// the bucket is already gone, and must propagate every other failure —
-    /// the swallowed error answered 200 while the bucket survived.
-    #[test]
-    fn test_purge_deleted_bucket_result_tolerates_only_missing_bucket() {
-        assert!(purge_deleted_bucket_result(Ok(())).is_ok());
-        assert!(purge_deleted_bucket_result(Err(StorageError::BucketNotFound("photos".to_string()))).is_ok());
-        assert!(purge_deleted_bucket_result(Err(StorageError::VolumeNotFound)).is_ok());
-        let err = purge_deleted_bucket_result(Err(StorageError::StorageFull))
-            .expect_err("non-not-found delete failures must propagate");
-        assert_ne!(*err.code(), S3ErrorCode::NoSuchBucket);
-    }
-
-    /// C5 red-light: the site-replication devnull drain must accept bodies
-    /// beyond the 1MB admin body cap — MinIO's link check streams large
-    /// probe bodies and treats a 400 as a broken link.
-    #[tokio::test]
-    async fn test_site_replication_devnull_drains_body_beyond_admin_cap() {
-        let body = Body::from(vec![0u8; MAX_ADMIN_REQUEST_BODY_SIZE + 1]);
-        drain_site_replication_devnull(body)
-            .await
-            .expect("devnull must drain bodies larger than the admin body cap");
-    }
-
-    /// A3 red-light: `versioningEnabled` must travel on every outbound
-    /// make-with-versioning bucket op so the query matches MinIO's
-    /// site-replication make-bucket wire contract (MinIO's own hook sends
-    /// `versioningEnabled=true` on this op).
-    #[test]
-    fn test_make_with_versioning_op_paths_send_versioning_enabled() {
-        let bucket = SRBucketInfo {
-            bucket: "photos".to_string(),
-            created_at: Some(OffsetDateTime::UNIX_EPOCH),
-            object_lock_config: Some(BASE64_STANDARD.encode("<ObjectLockConfiguration/>")),
-            ..Default::default()
-        };
-        let bootstrap = bootstrap_bucket_make_op_path(&bucket);
-        assert!(bootstrap.contains("operation=make-with-versioning"), "{bootstrap}");
-        assert!(bootstrap.contains("versioningEnabled=true"), "{bootstrap}");
-        assert!(bootstrap.contains("createdAt="), "{bootstrap}");
-        assert!(bootstrap.contains("lockEnabled=true"), "{bootstrap}");
-
-        // The broadcast path (create-bucket hook) shares the same builder.
-        let broadcast = make_with_versioning_bucket_op_path("photos", Some("1970-01-01T00:00:00Z"), false);
-        assert!(broadcast.contains("versioningEnabled=true"), "{broadcast}");
-        assert!(!broadcast.contains("lockEnabled"), "{broadcast}");
-    }
-
     #[tokio::test]
     #[serial]
     async fn test_add_bootstrap_scope_only_allows_expected_bucket_setup_until_guard_drops() {

rustfs/src/admin/site_replication_identity.rs

@@ -13,9 +13,9 @@
 // limitations under the License.
 
 use rustfs_madmin::{PeerInfo, SyncStatus};
-use std::collections::BTreeMap;
+use std::collections::{BTreeMap, hash_map::DefaultHasher};
+use std::hash::{Hash, Hasher};
 use url::Url;
-use uuid::Uuid;
 
 fn has_http_scheme(endpoint: &str) -> bool {
     endpoint.get(..7).is_some_and(|prefix| prefix.eq_ignore_ascii_case("http://"))
@@ -66,12 +66,10 @@ pub fn site_identity_key(endpoint: &str) -> String {
         .unwrap_or_else(|| trimmed.to_ascii_lowercase())
 }
 
-/// Fallback deployment ID for a peer that reported none. UUIDv5 over the
-/// canonical endpoint: the ID is persisted in site-replication state and
-/// broadcast to peers, so it must be identical across Rust toolchains
-/// (`DefaultHasher` is not) and across spellings of the same endpoint.
 pub fn deployment_id_for_endpoint(endpoint: &str) -> String {
-    Uuid::new_v5(&Uuid::NAMESPACE_URL, canonical_endpoint(endpoint).as_bytes()).to_string()
+    let mut hasher = DefaultHasher::new();
+    endpoint.hash(&mut hasher);
+    format!("{:016x}", hasher.finish())
 }
 
 pub fn same_identity_endpoint(left: &str, right: &str) -> bool {
@@ -176,23 +174,6 @@ mod tests {
         }
     }
 
-    /// B8 red-light: the fallback deployment ID must be a toolchain-stable
-    /// UUIDv5 over the canonical endpoint — `DefaultHasher` output is not
-    /// guaranteed stable across Rust releases, yet the ID is persisted in
-    /// site-replication state and broadcast to peers.
-    #[test]
-    fn deployment_id_for_endpoint_is_stable_uuid_v5_over_canonical_endpoint() {
-        let endpoint = "https://node-a.example.com:9000";
-        let id = deployment_id_for_endpoint(endpoint);
-        let parsed = uuid::Uuid::parse_str(&id).expect("fallback deployment ID must be a UUID");
-        assert_eq!(parsed.get_version_num(), 5, "fallback deployment ID must be UUIDv5");
-        // Deterministic for the same endpoint and for spelling variants that
-        // share a canonical form; distinct endpoints stay distinct.
-        assert_eq!(id, deployment_id_for_endpoint(endpoint));
-        assert_eq!(id, deployment_id_for_endpoint(" HTTPS://Node-A.Example.Com:9000/ "));
-        assert_ne!(id, deployment_id_for_endpoint("https://node-b.example.com:9000"));
-    }
-
     #[test]
     fn canonical_endpoint_accepts_case_insensitive_scheme() {
         assert_eq!(

rustfs/src/admin/storage_api.rs

@@ -51,7 +51,7 @@ mod ecstore_disk {
 }
 
 mod ecstore_error {
-    pub(crate) use crate::storage::storage_api::ecstore_error::{StorageError, is_err_bucket_not_found};
+    pub(crate) use crate::storage::storage_api::ecstore_error::StorageError;
 }
 
 #[allow(unused_imports)]
@@ -919,7 +919,6 @@ pub(crate) mod contract {
 }
 
 pub(crate) mod error {
-    pub(crate) use super::ecstore_error::is_err_bucket_not_found;
     pub(crate) use super::{Error, StorageError};
 }
 

rustfs/src/connect/heartbeat.rs

@@ -19,19 +19,23 @@ use std::sync::atomic::{AtomicU64, Ordering};
 use std::time::Duration;
 
 use chrono::{DateTime, SecondsFormat, Utc};
+use reqwest::{Client, StatusCode, Url, header};
+use rustls::RootCertStore;
+use rustls::pki_types::{CertificateDer, pem::PemObject as _};
 use serde::{Deserialize, Serialize};
 use uuid::Uuid;
+use zeroize::Zeroizing;
 
 use super::config::HeartbeatConfig;
-use super::credential_store::CredentialStoreError;
+use super::credential_store::{CredentialStoreError, DeviceCredential};
 use super::identity::IdentityError;
 use super::identity_store::StoreError;
-use super::registration::CredentialValidationError;
-use super::telemetry::{TelemetryDelivery, TelemetryError, TelemetryTransport, is_exact_utc_seconds};
+use super::registration::{CredentialValidationError, validate_stored_credential};
 
 const PROTOCOL_VERSION: &str = "v1";
 const AGENT_VERSION: &str = concat!("rustfs-agent/", env!("CARGO_PKG_VERSION"));
 const MAX_SEQUENCE: u64 = 9_007_199_254_740_991;
+const MAX_RESPONSE_BYTES: usize = 64 * 1024;
 #[cfg(unix)]
 const FILE_MODE: u32 = 0o600;
 static STAGING_SEQUENCE: AtomicU64 = AtomicU64::new(0);
@@ -118,39 +122,133 @@ pub(crate) enum Delivery {
 }
 
 pub(crate) struct HeartbeatSender {
-    transport: TelemetryTransport,
+    endpoint: Url,
+    root_store: RootCertStore,
+    roots: Vec<CertificateDer<'static>>,
+    config: HeartbeatConfig,
 }
 
 impl HeartbeatSender {
     pub(crate) fn new(config: HeartbeatConfig) -> Result<Self, HeartbeatError> {
+        let mut endpoint = Url::parse(&config.endpoint).map_err(|_| HeartbeatError::Endpoint)?;
+        if endpoint.scheme() != "https"
+            || endpoint.cannot_be_a_base()
+            || !endpoint.username().is_empty()
+            || endpoint.password().is_some()
+            || endpoint.query().is_some()
+            || endpoint.fragment().is_some()
+        {
+            return Err(HeartbeatError::Endpoint);
+        }
+        if !endpoint.path().ends_with('/') {
+            endpoint.set_path(&format!("{}/", endpoint.path()));
+        }
+        let roots = CertificateDer::pem_slice_iter(&config.root_ca_pem)
+            .collect::<Result<Vec<_>, _>>()
+            .map_err(|_| HeartbeatError::RootCertificate)?;
+        if roots.is_empty() {
+            return Err(HeartbeatError::RootCertificate);
+        }
+        let mut root_store = RootCertStore::empty();
+        let (accepted, rejected) = root_store.add_parsable_certificates(roots.clone());
+        if accepted != roots.len() || rejected != 0 {
+            return Err(HeartbeatError::RootCertificate);
+        }
         let schedule = config.schedule;
-        if schedule.cadence.is_zero() || schedule.jitter > schedule.cadence {
+        if schedule.cadence.is_zero()
+            || schedule.timeout.is_zero()
+            || schedule.timeout > Duration::from_secs(5)
+            || schedule.initial_backoff.is_zero()
+            || schedule.max_backoff < schedule.initial_backoff
+            || schedule.max_backoff > Duration::from_secs(5 * 60)
+            || schedule.jitter > schedule.cadence
+        {
             return Err(HeartbeatError::Schedule);
         }
         Ok(Self {
-            transport: TelemetryTransport::new(config)?,
+            endpoint,
+            root_store,
+            roots,
+            config,
         })
     }
 
     pub(crate) async fn send(&self, heartbeat: &PendingHeartbeat) -> Result<Delivery, HeartbeatError> {
-        match self.transport.post("heartbeats", heartbeat).await? {
-            TelemetryDelivery::Accepted { body, .. } => {
-                let accepted: HeartbeatResponse = serde_json::from_slice(&body).map_err(|_| HeartbeatError::Response)?;
-                if accepted.accepted_version != PROTOCOL_VERSION
-                    || accepted.capability_hints.len() > 32
-                    || accepted.capability_hints.iter().any(|hint| hint.len() > 32)
-                    || !is_exact_utc_seconds(&accepted.server_time)
-                {
-                    return Err(HeartbeatError::Response);
-                }
-                Ok(Delivery::Accepted {
-                    server_time: accepted.server_time,
-                })
+        let (cluster_uid, client) = {
+            let _lock = self.config.credential_store.lock().await?;
+            let credential = self.config.credential_store.load()?.ok_or(HeartbeatError::NotRegistered)?;
+            let identity = self.config.identity_store.load()?.ok_or(HeartbeatError::IdentityMissing)?;
+            validate_stored_credential(&credential, &identity, &self.root_store, &self.roots)?;
+            let now = Utc::now().timestamp();
+            if now < credential.not_before_unix || now >= credential.not_after_unix {
+                return Err(HeartbeatError::CredentialExpired);
             }
-            TelemetryDelivery::Retry { retry_after } => Ok(Delivery::Retry { retry_after }),
-            TelemetryDelivery::AuthenticationStopped { status, reason } => Ok(Delivery::AuthenticationStopped { status, reason }),
-            TelemetryDelivery::Rejected { status, reason } => Ok(Delivery::Rejected { status, reason }),
+            let cluster_uid = cluster_uid(&credential)?.to_owned();
+            let client = self.client(&credential, &identity.to_pkcs8_pem()?)?;
+            (cluster_uid, client)
+        };
+        let url = self.endpoint.join(&format!("clusters/{cluster_uid}/heartbeats"))?;
+        let response = match client.post(url).json(heartbeat).send().await {
+            Ok(response) => response,
+            Err(error) if error.is_timeout() || error.is_connect() || error.is_request() => {
+                return Ok(Delivery::Retry { retry_after: None });
+            }
+            Err(error) => return Err(error.into()),
+        };
+        let status = response.status();
+        if status == StatusCode::TOO_MANY_REQUESTS {
+            return Ok(Delivery::Retry {
+                retry_after: retry_after(response.headers(), Utc::now(), self.config.schedule.max_backoff),
+            });
         }
+        if status == StatusCode::REQUEST_TIMEOUT || status.is_server_error() {
+            return Ok(Delivery::Retry { retry_after: None });
+        }
+        if matches!(status, StatusCode::UNAUTHORIZED | StatusCode::FORBIDDEN) {
+            return Ok(Delivery::AuthenticationStopped {
+                status: status.as_u16(),
+                reason: response_reason(response).await,
+            });
+        }
+        if status != StatusCode::OK {
+            return Ok(Delivery::Rejected {
+                status: status.as_u16(),
+                reason: response_reason(response).await,
+            });
+        }
+        let accepted: HeartbeatResponse =
+            serde_json::from_slice(&bounded_body(response).await?).map_err(|_| HeartbeatError::Response)?;
+        if accepted.accepted_version != PROTOCOL_VERSION
+            || accepted.capability_hints.len() > 32
+            || accepted.capability_hints.iter().any(|hint| hint.len() > 32)
+            || !is_exact_utc_seconds(&accepted.server_time)
+        {
+            return Err(HeartbeatError::Response);
+        }
+        Ok(Delivery::Accepted {
+            server_time: accepted.server_time,
+        })
+    }
+
+    fn client(&self, credential: &DeviceCredential, key: &Zeroizing<String>) -> Result<Client, HeartbeatError> {
+        let mut pem = Zeroizing::new(Vec::with_capacity(credential.certificate_chain.len() + key.len() + 1));
+        pem.extend_from_slice(credential.certificate_chain.as_bytes());
+        pem.push(b'\n');
+        pem.extend_from_slice(key.as_bytes());
+        let identity = reqwest::Identity::from_pem(&pem).map_err(|_| HeartbeatError::IdentityCertificate)?;
+        let roots = self
+            .roots
+            .iter()
+            .map(|root| reqwest::Certificate::from_der(root.as_ref()))
+            .collect::<Result<Vec<_>, _>>()?;
+        Client::builder()
+            .https_only(true)
+            .redirect(reqwest::redirect::Policy::none())
+            .timeout(self.config.schedule.timeout)
+            .tls_certs_only(roots)
+            .identity(identity)
+            .build()
+            .map_err(Into::into)
     }
 }
 
@@ -280,6 +378,73 @@ impl HeartbeatStateStore {
     }
 }
 
+fn cluster_uid(credential: &DeviceCredential) -> Result<&str, HeartbeatError> {
+    let mut parts = credential.name.split('/');
+    let valid = parts.next() == Some("organizations");
+    let organization_uid = parts.next();
+    let valid = valid && parts.next() == Some("clusters");
+    let cluster_uid = parts.next();
+    let valid = valid && parts.next() == Some("clusterDevices");
+    let device_uid = parts.next();
+    if !valid
+        || organization_uid.is_none_or(str::is_empty)
+        || cluster_uid.is_none_or(str::is_empty)
+        || device_uid != Some(credential.uid.as_str())
+        || parts.next().is_some()
+    {
+        return Err(HeartbeatError::CredentialName);
+    }
+    cluster_uid.ok_or(HeartbeatError::CredentialName)
+}
+
+fn retry_after(headers: &header::HeaderMap, now: DateTime<Utc>, maximum: Duration) -> Option<Duration> {
+    let value = headers.get(header::RETRY_AFTER)?.to_str().ok()?;
+    let delay = value.parse::<u64>().ok().map(Duration::from_secs).or_else(|| {
+        DateTime::parse_from_rfc2822(value)
+            .ok()
+            .and_then(|at| (at.with_timezone(&Utc) - now).to_std().ok())
+    })?;
+    Some(delay.min(maximum))
+}
+
+fn is_exact_utc_seconds(value: &str) -> bool {
+    DateTime::parse_from_rfc3339(value).is_ok_and(|time| {
+        time.offset().local_minus_utc() == 0
+            && value.ends_with('Z')
+            && time.with_timezone(&Utc).to_rfc3339_opts(SecondsFormat::Secs, true) == value
+    })
+}
+
+async fn response_reason(response: reqwest::Response) -> Option<String> {
+    #[derive(Deserialize)]
+    struct Envelope {
+        #[serde(default)]
+        details: Vec<Detail>,
+    }
+    #[derive(Deserialize)]
+    struct Detail {
+        #[serde(default)]
+        reason: String,
+    }
+
+    serde_json::from_slice::<Envelope>(&bounded_body(response).await.ok()?)
+        .ok()?
+        .details
+        .into_iter()
+        .find_map(|detail| (!detail.reason.is_empty()).then_some(detail.reason))
+}
+
+async fn bounded_body(mut response: reqwest::Response) -> Result<Vec<u8>, HeartbeatError> {
+    let mut body = Vec::new();
+    while let Some(chunk) = response.chunk().await? {
+        if body.len().saturating_add(chunk.len()) > MAX_RESPONSE_BYTES {
+            return Err(HeartbeatError::ResponseTooLarge);
+        }
+        body.extend_from_slice(&chunk);
+    }
+    Ok(body)
+}
+
 fn parent(path: &Path) -> Result<&Path, HeartbeatError> {
     path.parent()
         .ok_or_else(|| state_io(path, io::Error::new(io::ErrorKind::InvalidInput, "state path has no parent")))
@@ -418,25 +583,3 @@ pub enum HeartbeatError {
     #[error(transparent)]
     CredentialValidation(#[from] CredentialValidationError),
 }
-
-impl From<TelemetryError> for HeartbeatError {
-    fn from(error: TelemetryError) -> Self {
-        match error {
-            TelemetryError::Endpoint => Self::Endpoint,
-            TelemetryError::RootCertificate => Self::RootCertificate,
-            TelemetryError::Schedule => Self::Schedule,
-            TelemetryError::NotRegistered => Self::NotRegistered,
-            TelemetryError::IdentityMissing => Self::IdentityMissing,
-            TelemetryError::IdentityCertificate => Self::IdentityCertificate,
-            TelemetryError::CredentialName => Self::CredentialName,
-            TelemetryError::CredentialExpired => Self::CredentialExpired,
-            TelemetryError::ResponseTooLarge => Self::ResponseTooLarge,
-            TelemetryError::Url(error) => Self::Url(error),
-            TelemetryError::Transport(error) => Self::Transport(error),
-            TelemetryError::Identity(error) => Self::Identity(error),
-            TelemetryError::IdentityStore(error) => Self::IdentityStore(error),
-            TelemetryError::CredentialStore(error) => Self::CredentialStore(error),
-            TelemetryError::CredentialValidation(error) => Self::CredentialValidation(error),
-        }
-    }
-}

rustfs/src/connect/inventory.rs

@@ -1,594 +0,0 @@
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-use std::collections::BTreeSet;
-use std::fs;
-use std::io::{self, Write as _};
-use std::path::{Path, PathBuf};
-use std::sync::atomic::{AtomicU64, Ordering};
-use std::time::Duration;
-
-use serde::{Deserialize, Serialize};
-use sha2::{Digest as _, Sha256};
-use uuid::Uuid;
-
-use super::config::HeartbeatConfig;
-use super::telemetry::{TelemetryDelivery, TelemetryError, TelemetryTransport, is_exact_utc_seconds};
-
-const PROTOCOL_VERSION: &str = "v1";
-const RUSTFS_VERSION: &str = concat!(
-    env!("CARGO_PKG_VERSION_MAJOR"),
-    ".",
-    env!("CARGO_PKG_VERSION_MINOR"),
-    ".",
-    env!("CARGO_PKG_VERSION_PATCH")
-);
-const HASH_PREFIX: &[u8] = b"rustfs-connect/agent/v1/inventory-snapshot\n";
-const MAX_SEQUENCE: u64 = 9_007_199_254_740_991;
-const MAX_SAFE_INTEGER: u64 = 9_007_199_254_740_991;
-#[cfg(unix)]
-const FILE_MODE: u32 = 0o600;
-static STAGING_SEQUENCE: AtomicU64 = AtomicU64::new(0);
-
-#[derive(Clone, Copy, Debug, PartialEq, Eq)]
-pub struct InventorySchedule {
-    pub cadence: Duration,
-    pub jitter: Duration,
-}
-
-impl Default for InventorySchedule {
-    fn default() -> Self {
-        Self {
-            cadence: Duration::from_secs(6 * 60 * 60),
-            jitter: Duration::from_secs(30 * 60),
-        }
-    }
-}
-
-#[derive(Clone, Debug, PartialEq, Eq)]
-pub enum InventoryStatus {
-    Starting,
-    Unchanged { content_hash: String },
-    Online { content_hash: String, received_at: String },
-    BackingOff { delay: Duration },
-    AuthenticationStopped { status: u16, reason: Option<String> },
-    Failed { reason: String },
-    Stopped,
-}
-
-#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord, Serialize, Deserialize)]
-pub enum InventoryFlag {
-    #[serde(rename = "capacity.critical")]
-    CapacityCritical,
-    #[serde(rename = "capacity.warning")]
-    CapacityWarning,
-    #[serde(rename = "clock.skew")]
-    ClockSkew,
-    #[serde(rename = "cluster.degraded")]
-    ClusterDegraded,
-    #[serde(rename = "cluster.healing")]
-    ClusterHealing,
-    #[serde(rename = "cluster.readonly")]
-    ClusterReadonly,
-    #[serde(rename = "drive.offline")]
-    DriveOffline,
-    #[serde(rename = "node.offline")]
-    NodeOffline,
-}
-
-#[derive(Clone, Copy, Debug, PartialEq, Eq, Serialize, Deserialize)]
-#[serde(rename_all = "lowercase")]
-pub enum OperatingSystemFamily {
-    Linux,
-    Darwin,
-    Windows,
-    Freebsd,
-    Other,
-}
-
-#[derive(Clone, Copy, Debug, PartialEq, Eq, Serialize, Deserialize)]
-#[serde(deny_unknown_fields, rename_all = "camelCase")]
-pub struct InventoryOsVersion {
-    family: OperatingSystemFamily,
-    major: u16,
-    minor: u16,
-}
-
-impl InventoryOsVersion {
-    pub fn new(family: OperatingSystemFamily, major: u16, minor: u16) -> Result<Self, InventoryError> {
-        if major > 9999 || minor > 9999 {
-            return Err(InventoryError::OsVersion);
-        }
-        Ok(Self { family, major, minor })
-    }
-}
-
-#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
-#[serde(deny_unknown_fields, rename_all = "camelCase")]
-pub struct InventorySnapshot {
-    rustfs_version: String,
-    os_version: Option<InventoryOsVersion>,
-    node_count: u16,
-    drive_count: u32,
-    capacity_total_bytes: u64,
-    capacity_used_bytes: u64,
-    coarse_flags: Vec<InventoryFlag>,
-}
-
-impl InventorySnapshot {
-    pub fn current(
-        node_count: usize,
-        drive_count: usize,
-        capacity_total_bytes: u64,
-        capacity_free_bytes: u64,
-        coarse_flags: impl IntoIterator<Item = InventoryFlag>,
-    ) -> Result<Self, InventoryError> {
-        let capacity_used_bytes = capacity_total_bytes
-            .checked_sub(capacity_free_bytes)
-            .ok_or(InventoryError::Capacity)?;
-        Self::new(
-            RUSTFS_VERSION,
-            None,
-            node_count,
-            drive_count,
-            capacity_total_bytes,
-            capacity_used_bytes,
-            coarse_flags,
-        )
-    }
-
-    pub fn new(
-        rustfs_version: impl Into<String>,
-        os_version: Option<InventoryOsVersion>,
-        node_count: usize,
-        drive_count: usize,
-        capacity_total_bytes: u64,
-        capacity_used_bytes: u64,
-        coarse_flags: impl IntoIterator<Item = InventoryFlag>,
-    ) -> Result<Self, InventoryError> {
-        let snapshot = Self {
-            rustfs_version: rustfs_version.into(),
-            os_version,
-            node_count: u16::try_from(node_count).map_err(|_| InventoryError::NodeCount)?,
-            drive_count: u32::try_from(drive_count).map_err(|_| InventoryError::DriveCount)?,
-            capacity_total_bytes,
-            capacity_used_bytes,
-            coarse_flags: coarse_flags.into_iter().collect::<BTreeSet<_>>().into_iter().collect(),
-        };
-        snapshot.validate()?;
-        Ok(snapshot)
-    }
-
-    pub fn content_hash(&self) -> Result<String, InventoryError> {
-        #[derive(Serialize)]
-        #[serde(rename_all = "camelCase")]
-        struct Canonical<'a> {
-            capacity_total_bytes: u64,
-            capacity_used_bytes: u64,
-            coarse_flags: &'a [InventoryFlag],
-            drive_count: u32,
-            node_count: u16,
-            os_version: Option<InventoryOsVersion>,
-            rustfs_version: &'a str,
-        }
-
-        let canonical = serde_json::to_vec(&Canonical {
-            capacity_total_bytes: self.capacity_total_bytes,
-            capacity_used_bytes: self.capacity_used_bytes,
-            coarse_flags: &self.coarse_flags,
-            drive_count: self.drive_count,
-            node_count: self.node_count,
-            os_version: self.os_version,
-            rustfs_version: &self.rustfs_version,
-        })?;
-        let mut digest = Sha256::new();
-        digest.update(HASH_PREFIX);
-        digest.update(canonical);
-        Ok(hex_simd::encode_to_string(digest.finalize(), hex_simd::AsciiCase::Lower))
-    }
-
-    fn validate(&self) -> Result<(), InventoryError> {
-        if !valid_version(&self.rustfs_version) {
-            return Err(InventoryError::RustfsVersion);
-        }
-        if self.node_count == 0 || self.node_count > 4096 {
-            return Err(InventoryError::NodeCount);
-        }
-        if self.drive_count > 1_048_576 {
-            return Err(InventoryError::DriveCount);
-        }
-        if self.capacity_total_bytes > MAX_SAFE_INTEGER || self.capacity_used_bytes > self.capacity_total_bytes {
-            return Err(InventoryError::Capacity);
-        }
-        Ok(())
-    }
-}
-
-fn valid_version(version: &str) -> bool {
-    let components = version.split('.').collect::<Vec<_>>();
-    components.len() == 3
-        && components.iter().all(|component| {
-            !component.is_empty()
-                && component.len() <= 4
-                && (component == &"0" || !component.starts_with('0'))
-                && component.parse::<u16>().is_ok_and(|value| value <= 9999)
-        })
-}
-
-#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
-#[serde(deny_unknown_fields, rename_all = "camelCase")]
-pub(crate) struct PendingInventory {
-    protocol_version: String,
-    request_id: String,
-    sequence: u64,
-    #[serde(flatten)]
-    snapshot: InventorySnapshot,
-}
-
-impl PendingInventory {
-    fn new(snapshot: InventorySnapshot, sequence: u64) -> Self {
-        Self {
-            protocol_version: PROTOCOL_VERSION.to_owned(),
-            request_id: Uuid::new_v4().to_string(),
-            sequence,
-            snapshot,
-        }
-    }
-
-    fn is_valid(&self) -> bool {
-        self.protocol_version == PROTOCOL_VERSION
-            && self.sequence <= MAX_SEQUENCE
-            && self.snapshot.validate().is_ok()
-            && Uuid::parse_str(&self.request_id)
-                .is_ok_and(|request_id| request_id.get_version_num() == 4 && request_id.to_string() == self.request_id)
-    }
-
-    fn content_hash(&self) -> Result<String, InventoryError> {
-        self.snapshot.content_hash()
-    }
-}
-
-pub(crate) enum InventoryDelivery {
-    Accepted { content_hash: String, received_at: String },
-    Retry { retry_after: Option<Duration> },
-    AuthenticationStopped { status: u16, reason: Option<String> },
-    Rejected { status: u16, reason: Option<String> },
-}
-
-pub(crate) struct InventorySender {
-    transport: TelemetryTransport,
-}
-
-impl InventorySender {
-    pub(crate) fn new(config: HeartbeatConfig) -> Result<Self, InventoryError> {
-        Ok(Self {
-            transport: TelemetryTransport::new(config)?,
-        })
-    }
-
-    pub(crate) async fn send(&self, inventory: &PendingInventory) -> Result<InventoryDelivery, InventoryError> {
-        match self.transport.post("inventorySnapshots", inventory).await? {
-            TelemetryDelivery::Accepted { cluster_name, body } => {
-                #[derive(Deserialize)]
-                #[serde(rename_all = "camelCase")]
-                struct InventoryResponse {
-                    name: String,
-                    uid: String,
-                    content_hash: String,
-                    received_at: String,
-                }
-
-                let accepted: InventoryResponse = serde_json::from_slice(&body).map_err(|_| InventoryError::Response)?;
-                let uid = Uuid::parse_str(&accepted.uid).map_err(|_| InventoryError::Response)?;
-                let content_hash = inventory.content_hash()?;
-                if uid.get_version_num() != 7
-                    || uid.to_string() != accepted.uid
-                    || accepted.name != format!("{cluster_name}/inventorySnapshots/{}", accepted.uid)
-                    || accepted.content_hash != content_hash
-                    || !is_exact_utc_seconds(&accepted.received_at)
-                {
-                    return Err(InventoryError::Response);
-                }
-                Ok(InventoryDelivery::Accepted {
-                    content_hash,
-                    received_at: accepted.received_at,
-                })
-            }
-            TelemetryDelivery::Retry { retry_after } => Ok(InventoryDelivery::Retry { retry_after }),
-            TelemetryDelivery::AuthenticationStopped { status, reason } => {
-                Ok(InventoryDelivery::AuthenticationStopped { status, reason })
-            }
-            TelemetryDelivery::Rejected { status, reason } => Ok(InventoryDelivery::Rejected { status, reason }),
-        }
-    }
-}
-
-#[derive(Clone)]
-pub(crate) struct InventoryStateStore {
-    path: PathBuf,
-}
-
-#[derive(Default, Serialize, Deserialize)]
-#[serde(deny_unknown_fields, rename_all = "camelCase")]
-struct InventoryState {
-    next_sequence: u64,
-    pending: Option<PendingInventory>,
-    last_accepted_content_hash: Option<String>,
-}
-
-impl InventoryStateStore {
-    pub(crate) fn from_heartbeat_path(path: &Path) -> Result<Self, InventoryError> {
-        let root = path.parent().and_then(Path::parent).ok_or(InventoryError::StatePath)?;
-        Ok(Self {
-            path: root.join("inventory/state.json"),
-        })
-    }
-
-    pub(crate) fn try_runtime_lock(&self) -> Result<fs::File, InventoryError> {
-        let directory = parent(&self.path)?;
-        fs::create_dir_all(directory).map_err(|source| state_io(directory, source))?;
-        let name = filename(&self.path)?;
-        let path = directory.join(format!(".{name}.lock"));
-        let mut options = fs::OpenOptions::new();
-        options.create(true).truncate(false).read(true).write(true);
-        #[cfg(unix)]
-        {
-            use std::os::unix::fs::OpenOptionsExt as _;
-            options.mode(FILE_MODE);
-        }
-        let lock = options.open(&path).map_err(|source| state_io(&path, source))?;
-        check_mode(&path)?;
-        lock.try_lock().map_err(|_| InventoryError::AlreadyRunning)?;
-        Ok(lock)
-    }
-
-    pub(crate) async fn pending(&self) -> Result<Option<PendingInventory>, InventoryError> {
-        let store = self.clone();
-        tokio::task::spawn_blocking(move || {
-            let state = store.read()?;
-            if state.pending.is_none() && state.next_sequence > MAX_SEQUENCE {
-                return Err(InventoryError::SequenceExhausted);
-            }
-            Ok(state.pending)
-        })
-        .await
-        .map_err(|source| state_io(&self.path, io::Error::other(source)))?
-    }
-
-    pub(crate) async fn prepare(&self, snapshot: InventorySnapshot) -> Result<Option<PendingInventory>, InventoryError> {
-        let store = self.clone();
-        tokio::task::spawn_blocking(move || store.prepare_sync(snapshot))
-            .await
-            .map_err(|source| state_io(&self.path, io::Error::other(source)))?
-    }
-
-    pub(crate) async fn mark_accepted(&self, accepted: &PendingInventory) -> Result<(), InventoryError> {
-        let store = self.clone();
-        let accepted = accepted.clone();
-        tokio::task::spawn_blocking(move || store.mark_accepted_sync(&accepted))
-            .await
-            .map_err(|source| state_io(&self.path, io::Error::other(source)))?
-    }
-
-    fn prepare_sync(&self, snapshot: InventorySnapshot) -> Result<Option<PendingInventory>, InventoryError> {
-        let mut state = self.read()?;
-        if state.pending.is_some() {
-            return Ok(state.pending);
-        }
-        let content_hash = snapshot.content_hash()?;
-        if state.last_accepted_content_hash.as_deref() == Some(&content_hash) {
-            return Ok(None);
-        }
-        if state.next_sequence > MAX_SEQUENCE {
-            return Err(InventoryError::SequenceExhausted);
-        }
-        let pending = PendingInventory::new(snapshot, state.next_sequence);
-        state.pending = Some(pending.clone());
-        self.write(&state)?;
-        Ok(Some(pending))
-    }
-
-    fn mark_accepted_sync(&self, accepted: &PendingInventory) -> Result<(), InventoryError> {
-        let mut state = self.read()?;
-        if state.pending.as_ref() != Some(accepted) {
-            return Err(InventoryError::StateConflict);
-        }
-        state.next_sequence = accepted.sequence.checked_add(1).ok_or(InventoryError::SequenceExhausted)?;
-        state.last_accepted_content_hash = Some(accepted.content_hash()?);
-        state.pending = None;
-        self.write(&state)
-    }
-
-    fn read(&self) -> Result<InventoryState, InventoryError> {
-        let bytes = match fs::read(&self.path) {
-            Ok(bytes) => bytes,
-            Err(source) if source.kind() == io::ErrorKind::NotFound => return Ok(InventoryState::default()),
-            Err(source) => return Err(state_io(&self.path, source)),
-        };
-        check_mode(&self.path)?;
-        let state: InventoryState = serde_json::from_slice(&bytes).map_err(|source| InventoryError::StateInvalid {
-            path: self.path.clone(),
-            source,
-        })?;
-        let last_hash_valid = state.last_accepted_content_hash.as_deref().is_none_or(valid_content_hash);
-        let pending_valid = state.pending.as_ref().is_none_or(|pending| {
-            pending.sequence == state.next_sequence
-                && pending.is_valid()
-                && pending
-                    .content_hash()
-                    .is_ok_and(|hash| state.last_accepted_content_hash.as_deref() != Some(&hash))
-        });
-        if state.next_sequence > MAX_SEQUENCE + 1 || !last_hash_valid || !pending_valid {
-            return Err(InventoryError::StateCorrupt { path: self.path.clone() });
-        }
-        Ok(state)
-    }
-
-    fn write(&self, state: &InventoryState) -> Result<(), InventoryError> {
-        let bytes = serde_json::to_vec(state).map_err(|source| InventoryError::StateInvalid {
-            path: self.path.clone(),
-            source,
-        })?;
-        let directory = parent(&self.path)?;
-        fs::create_dir_all(directory).map_err(|source| state_io(directory, source))?;
-        let temp = stage(directory, &self.path, &bytes)?;
-        let result = fs::rename(&temp, &self.path)
-            .map_err(|source| state_io(&self.path, source))
-            .and_then(|()| fsync_dir(directory).map_err(|source| state_io(directory, source)));
-        if result.is_err() {
-            let _ = fs::remove_file(temp);
-        }
-        result
-    }
-}
-
-fn valid_content_hash(value: &str) -> bool {
-    value.len() == 64
-        && value
-            .bytes()
-            .all(|byte| byte.is_ascii_digit() || (b'a'..=b'f').contains(&byte))
-}
-
-fn parent(path: &Path) -> Result<&Path, InventoryError> {
-    path.parent()
-        .ok_or_else(|| state_io(path, io::Error::new(io::ErrorKind::InvalidInput, "state path has no parent")))
-}
-
-fn filename(path: &Path) -> Result<&str, InventoryError> {
-    path.file_name()
-        .and_then(|name| name.to_str())
-        .ok_or_else(|| state_io(path, io::Error::new(io::ErrorKind::InvalidInput, "state filename is invalid")))
-}
-
-fn stage(directory: &Path, destination: &Path, bytes: &[u8]) -> Result<PathBuf, InventoryError> {
-    let name = filename(destination)?;
-    loop {
-        let path = directory.join(format!(
-            ".{name}.{}.{}.tmp",
-            std::process::id(),
-            STAGING_SEQUENCE.fetch_add(1, Ordering::Relaxed)
-        ));
-        let mut options = fs::OpenOptions::new();
-        options.write(true).create_new(true);
-        #[cfg(unix)]
-        {
-            use std::os::unix::fs::OpenOptionsExt as _;
-            options.mode(FILE_MODE);
-        }
-        let mut file = match options.open(&path) {
-            Ok(file) => file,
-            Err(source) if source.kind() == io::ErrorKind::AlreadyExists => continue,
-            Err(source) => return Err(state_io(&path, source)),
-        };
-        if let Err(source) = file.write_all(bytes).and_then(|()| file.sync_all()) {
-            let _ = fs::remove_file(&path);
-            return Err(state_io(&path, source));
-        }
-        return Ok(path);
-    }
-}
-
-fn state_io(path: &Path, source: io::Error) -> InventoryError {
-    InventoryError::StateIo {
-        path: path.to_path_buf(),
-        source,
-    }
-}
-
-#[cfg(unix)]
-fn check_mode(path: &Path) -> Result<(), InventoryError> {
-    use std::os::unix::fs::PermissionsExt as _;
-
-    let mode = fs::metadata(path)
-        .map_err(|source| state_io(path, source))?
-        .permissions()
-        .mode()
-        & 0o7777;
-    if mode != FILE_MODE {
-        return Err(InventoryError::StatePermissions {
-            path: path.to_path_buf(),
-            mode,
-            expected: FILE_MODE,
-        });
-    }
-    Ok(())
-}
-
-#[cfg(not(unix))]
-fn check_mode(_path: &Path) -> Result<(), InventoryError> {
-    Ok(())
-}
-
-fn fsync_dir(directory: &Path) -> io::Result<()> {
-    #[cfg(unix)]
-    fs::File::open(directory)?.sync_all()?;
-    #[cfg(not(unix))]
-    let _ = directory;
-    Ok(())
-}
-
-#[derive(Debug, thiserror::Error)]
-pub enum InventoryError {
-    #[error("the RustFS inventory version is outside protocol bounds")]
-    RustfsVersion,
-    #[error("the RustFS inventory operating-system version is outside protocol bounds")]
-    OsVersion,
-    #[error("the RustFS inventory node count is outside protocol bounds")]
-    NodeCount,
-    #[error("the RustFS inventory drive count is outside protocol bounds")]
-    DriveCount,
-    #[error("the RustFS inventory capacity is outside protocol bounds")]
-    Capacity,
-    #[error("the RustFS inventory snapshot is incomplete: observed {observed} of {expected} configured drives")]
-    SnapshotIncomplete { expected: usize, observed: usize },
-    #[error("the Connect inventory schedule is invalid")]
-    Schedule,
-    #[error("the Connect inventory sequence is exhausted")]
-    SequenceExhausted,
-    #[error("a Connect inventory runtime already owns this state")]
-    AlreadyRunning,
-    #[error("the persisted Connect inventory changed while delivery was in flight")]
-    StateConflict,
-    #[error("the Connect inventory state path is invalid")]
-    StatePath,
-    #[error("Connect inventory state I/O failed at {path}: {source}")]
-    StateIo {
-        path: PathBuf,
-        #[source]
-        source: io::Error,
-    },
-    #[error("Connect inventory state at {path} is invalid: {source}")]
-    StateInvalid {
-        path: PathBuf,
-        #[source]
-        source: serde_json::Error,
-    },
-    #[error("Connect inventory state at {path} violates the protocol invariants")]
-    StateCorrupt { path: PathBuf },
-    #[cfg(unix)]
-    #[error("Connect inventory state at {path} has mode {mode:o}, expected {expected:o}")]
-    StatePermissions { path: PathBuf, mode: u32, expected: u32 },
-    #[error("Connect returned an invalid inventory response")]
-    Response,
-    #[error(transparent)]
-    Json(#[from] serde_json::Error),
-    #[error("Connect inventory delivery failed: {0}")]
-    Telemetry(String),
-}
-
-impl From<TelemetryError> for InventoryError {
-    fn from(error: TelemetryError) -> Self {
-        Self::Telemetry(error.to_string())
-    }
-}

rustfs/src/connect/mod.rs

@@ -31,11 +31,9 @@ pub mod credential_store;
 pub mod heartbeat;
 pub mod identity;
 pub mod identity_store;
-pub mod inventory;
 pub mod offline;
 pub mod registration;
 pub mod runtime;
-mod telemetry;
 
 pub use client::{ClientError, ConnectClient, ConnectConfig};
 pub use config::{HeartbeatConfig, HeartbeatConfigError, HeartbeatSchedule};
@@ -43,10 +41,6 @@ pub use credential_store::{CredentialStore, DeviceCredential};
 pub use heartbeat::{CoarseNodeSummary, HeartbeatError, HeartbeatStatus};
 pub use identity::{DeviceIdentity, IdentityError, RegistrationProof, RegistrationTranscript};
 pub use identity_store::{IdentityStore, StoreError};
-pub use inventory::{
-    InventoryError, InventoryFlag, InventoryOsVersion, InventorySchedule, InventorySnapshot, InventoryStatus,
-    OperatingSystemFamily,
-};
 pub use offline::{EnrollmentError, OfflineEnrollment, OfflineKeyStore, VerifiedChallenge};
 pub use registration::{RegistrationToken, TokenError};
-pub use runtime::{HeartbeatRuntime, InventoryRuntime, spawn_heartbeat_runtime, spawn_inventory_runtime};
+pub use runtime::{HeartbeatRuntime, spawn_heartbeat_runtime};

rustfs/src/connect/runtime.rs

@@ -23,16 +23,11 @@ use tokio_util::sync::CancellationToken;
 
 use super::config::HeartbeatConfig;
 use super::heartbeat::{CoarseNodeSummary, Delivery, HeartbeatError, HeartbeatSender, HeartbeatStateStore, HeartbeatStatus};
-use super::inventory::{
-    InventoryDelivery, InventoryError, InventorySchedule, InventorySender, InventorySnapshot, InventoryStateStore,
-    InventoryStatus,
-};
 
 pub struct HeartbeatRuntime {
     shutdown: CancellationToken,
     status: watch::Receiver<HeartbeatStatus>,
     task: Option<JoinHandle<()>>,
-    inventory: Option<InventoryRuntime>,
 }
 
 impl HeartbeatRuntime {
@@ -40,19 +35,11 @@ impl HeartbeatRuntime {
         self.status.clone()
     }
 
-    pub(crate) fn with_inventory(mut self, inventory: Option<InventoryRuntime>) -> Self {
-        self.inventory = inventory;
-        self
-    }
-
     pub async fn shutdown(mut self) {
         self.shutdown.cancel();
         if let Some(task) = self.task.take() {
             let _ = task.await;
         }
-        if let Some(inventory) = self.inventory.take() {
-            inventory.shutdown().await;
-        }
     }
 }
 
@@ -62,31 +49,6 @@ impl Drop for HeartbeatRuntime {
     }
 }
 
-pub struct InventoryRuntime {
-    shutdown: CancellationToken,
-    status: watch::Receiver<InventoryStatus>,
-    task: Option<JoinHandle<()>>,
-}
-
-impl InventoryRuntime {
-    pub fn status(&self) -> watch::Receiver<InventoryStatus> {
-        self.status.clone()
-    }
-
-    pub async fn shutdown(mut self) {
-        self.shutdown.cancel();
-        if let Some(task) = self.task.take() {
-            let _ = task.await;
-        }
-    }
-}
-
-impl Drop for InventoryRuntime {
-    fn drop(&mut self) {
-        self.shutdown.cancel();
-    }
-}
-
 pub fn spawn_heartbeat_runtime<F>(
     config: Option<HeartbeatConfig>,
     parent_shutdown: &CancellationToken,
@@ -160,126 +122,6 @@ where
         shutdown,
         status: status_rx,
         task: Some(task),
-        inventory: None,
-    }))
-}
-
-pub fn spawn_inventory_runtime<F, Fut>(
-    config: Option<HeartbeatConfig>,
-    schedule: InventorySchedule,
-    parent_shutdown: &CancellationToken,
-    sample: F,
-) -> Result<Option<InventoryRuntime>, InventoryError>
-where
-    F: Fn() -> Fut + Send + Sync + 'static,
-    Fut: Future<Output = Result<InventorySnapshot, InventoryError>> + Send + 'static,
-{
-    let Some(config) = config else {
-        return Ok(None);
-    };
-    if schedule.cadence.is_zero() || schedule.jitter > schedule.cadence {
-        return Err(InventoryError::Schedule);
-    }
-    let retry_schedule = config.schedule;
-    let store = InventoryStateStore::from_heartbeat_path(&config.state_path)?;
-    let lock = store.try_runtime_lock()?;
-    let sender = InventorySender::new(config)?;
-    let shutdown = parent_shutdown.child_token();
-    let task_shutdown = shutdown.clone();
-    let (status_tx, status_rx) = watch::channel(InventoryStatus::Starting);
-    let task = tokio::spawn(async move {
-        let _lock = lock;
-        let mut backoff = retry_schedule.initial_backoff;
-        loop {
-            if task_shutdown.is_cancelled() {
-                break;
-            }
-            let pending = match store.pending().await {
-                Ok(Some(pending)) => pending,
-                Ok(None) => {
-                    let snapshot = match cancellable(&task_shutdown, sample()).await {
-                        Some(Ok(snapshot)) => snapshot,
-                        Some(Err(InventoryError::SnapshotIncomplete { .. })) => {
-                            let delay = backoff;
-                            backoff = backoff.saturating_mul(2).min(retry_schedule.max_backoff);
-                            let _ = status_tx.send(InventoryStatus::BackingOff { delay });
-                            if sleep_or_cancel(&task_shutdown, delay).await {
-                                break;
-                            }
-                            continue;
-                        }
-                        Some(Err(error)) => return failed_inventory(&status_tx, error),
-                        None => break,
-                    };
-                    let content_hash = match snapshot.content_hash() {
-                        Ok(content_hash) => content_hash,
-                        Err(error) => return failed_inventory(&status_tx, error),
-                    };
-                    match store.prepare(snapshot).await {
-                        Ok(Some(pending)) => pending,
-                        Ok(None) => {
-                            backoff = retry_schedule.initial_backoff;
-                            let _ = status_tx.send(InventoryStatus::Unchanged { content_hash });
-                            if sleep_or_cancel(&task_shutdown, schedule.cadence.saturating_add(jitter(schedule.jitter))).await {
-                                break;
-                            }
-                            continue;
-                        }
-                        Err(error) => return failed_inventory(&status_tx, error),
-                    }
-                }
-                Err(error) => return failed_inventory(&status_tx, error),
-            };
-            let delivery = match cancellable(&task_shutdown, sender.send(&pending)).await {
-                Some(Ok(delivery)) => delivery,
-                Some(Err(error)) => return failed_inventory(&status_tx, error),
-                None => break,
-            };
-            let delay = match delivery {
-                InventoryDelivery::Accepted {
-                    content_hash,
-                    received_at,
-                } => {
-                    if let Err(error) = store.mark_accepted(&pending).await {
-                        return failed_inventory(&status_tx, error);
-                    }
-                    backoff = retry_schedule.initial_backoff;
-                    let _ = status_tx.send(InventoryStatus::Online {
-                        content_hash,
-                        received_at,
-                    });
-                    schedule.cadence.saturating_add(jitter(schedule.jitter))
-                }
-                InventoryDelivery::Retry { retry_after } => {
-                    let delay = retry_after
-                        .unwrap_or(backoff)
-                        .clamp(retry_schedule.initial_backoff, retry_schedule.max_backoff);
-                    backoff = backoff.saturating_mul(2).min(retry_schedule.max_backoff);
-                    let _ = status_tx.send(InventoryStatus::BackingOff { delay });
-                    delay
-                }
-                InventoryDelivery::AuthenticationStopped { status, reason } => {
-                    let _ = status_tx.send(InventoryStatus::AuthenticationStopped { status, reason });
-                    return;
-                }
-                InventoryDelivery::Rejected { status, reason } => {
-                    let suffix = reason.map_or_else(String::new, |reason| format!("; reason={reason}"));
-                    let _ = status_tx.send(InventoryStatus::Failed {
-                        reason: format!("Connect rejected inventory with HTTP {status}{suffix}"),
-                    });
-                    return;
-                }
-            };
-            if sleep_or_cancel(&task_shutdown, delay).await {
-                break;
-            }
-        }
-        let _ = status_tx.send(InventoryStatus::Stopped);
-    });
-    Ok(Some(InventoryRuntime {
-        shutdown,
-        status: status_rx,
-        task: Some(task),
     }))
 }
 
@@ -289,12 +131,6 @@ fn failed(status: &watch::Sender<HeartbeatStatus>, error: HeartbeatError) {
     });
 }
 
-fn failed_inventory(status: &watch::Sender<InventoryStatus>, error: InventoryError) {
-    let _ = status.send(InventoryStatus::Failed {
-        reason: error.to_string(),
-    });
-}
-
 fn jitter(maximum: Duration) -> Duration {
     if maximum.is_zero() {
         Duration::ZERO

rustfs/src/connect/telemetry.rs

@@ -1,265 +0,0 @@
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-use std::time::Duration;
-
-use chrono::{DateTime, SecondsFormat, Utc};
-use reqwest::{Client, StatusCode, Url, header};
-use rustls::RootCertStore;
-use rustls::pki_types::{CertificateDer, pem::PemObject as _};
-use serde::{Deserialize, Serialize};
-use zeroize::Zeroizing;
-
-use super::config::HeartbeatConfig;
-use super::credential_store::{CredentialStoreError, DeviceCredential};
-use super::identity::IdentityError;
-use super::identity_store::StoreError;
-use super::registration::{CredentialValidationError, validate_stored_credential};
-
-const MAX_RESPONSE_BYTES: usize = 64 * 1024;
-
-pub(crate) enum TelemetryDelivery {
-    Accepted { cluster_name: String, body: Vec<u8> },
-    Retry { retry_after: Option<Duration> },
-    AuthenticationStopped { status: u16, reason: Option<String> },
-    Rejected { status: u16, reason: Option<String> },
-}
-
-pub(crate) struct TelemetryTransport {
-    endpoint: Url,
-    root_store: RootCertStore,
-    roots: Vec<CertificateDer<'static>>,
-    config: HeartbeatConfig,
-}
-
-impl TelemetryTransport {
-    pub(crate) fn new(config: HeartbeatConfig) -> Result<Self, TelemetryError> {
-        let mut endpoint = Url::parse(&config.endpoint).map_err(|_| TelemetryError::Endpoint)?;
-        if endpoint.scheme() != "https"
-            || endpoint.cannot_be_a_base()
-            || !endpoint.username().is_empty()
-            || endpoint.password().is_some()
-            || endpoint.query().is_some()
-            || endpoint.fragment().is_some()
-        {
-            return Err(TelemetryError::Endpoint);
-        }
-        if !endpoint.path().ends_with('/') {
-            endpoint.set_path(&format!("{}/", endpoint.path()));
-        }
-        let roots = CertificateDer::pem_slice_iter(&config.root_ca_pem)
-            .collect::<Result<Vec<_>, _>>()
-            .map_err(|_| TelemetryError::RootCertificate)?;
-        if roots.is_empty() {
-            return Err(TelemetryError::RootCertificate);
-        }
-        let mut root_store = RootCertStore::empty();
-        let (accepted, rejected) = root_store.add_parsable_certificates(roots.clone());
-        if accepted != roots.len() || rejected != 0 {
-            return Err(TelemetryError::RootCertificate);
-        }
-        let schedule = config.schedule;
-        if schedule.timeout.is_zero()
-            || schedule.timeout > Duration::from_secs(5)
-            || schedule.initial_backoff.is_zero()
-            || schedule.max_backoff < schedule.initial_backoff
-            || schedule.max_backoff > Duration::from_secs(5 * 60)
-        {
-            return Err(TelemetryError::Schedule);
-        }
-        Ok(Self {
-            endpoint,
-            root_store,
-            roots,
-            config,
-        })
-    }
-
-    pub(crate) async fn post<T: Serialize>(&self, collection: &str, value: &T) -> Result<TelemetryDelivery, TelemetryError> {
-        let (cluster_name, cluster_uid, client) = self.authenticated_client().await?;
-        let url = self.endpoint.join(&format!("clusters/{cluster_uid}/{collection}"))?;
-        let response = match client.post(url).json(value).send().await {
-            Ok(response) => response,
-            Err(error) if error.is_timeout() || error.is_connect() || error.is_request() => {
-                return Ok(TelemetryDelivery::Retry { retry_after: None });
-            }
-            Err(error) => return Err(error.into()),
-        };
-        let status = response.status();
-        if status == StatusCode::TOO_MANY_REQUESTS {
-            return Ok(TelemetryDelivery::Retry {
-                retry_after: retry_after(response.headers(), Utc::now(), self.config.schedule.max_backoff),
-            });
-        }
-        if status == StatusCode::REQUEST_TIMEOUT || status.is_server_error() {
-            return Ok(TelemetryDelivery::Retry { retry_after: None });
-        }
-        if matches!(status, StatusCode::UNAUTHORIZED | StatusCode::FORBIDDEN) {
-            return Ok(TelemetryDelivery::AuthenticationStopped {
-                status: status.as_u16(),
-                reason: response_reason(response).await,
-            });
-        }
-        if status != StatusCode::OK {
-            return Ok(TelemetryDelivery::Rejected {
-                status: status.as_u16(),
-                reason: response_reason(response).await,
-            });
-        }
-        Ok(TelemetryDelivery::Accepted {
-            cluster_name,
-            body: bounded_body(response).await?,
-        })
-    }
-
-    async fn authenticated_client(&self) -> Result<(String, String, Client), TelemetryError> {
-        let _lock = self.config.credential_store.lock().await?;
-        let credential = self.config.credential_store.load()?.ok_or(TelemetryError::NotRegistered)?;
-        let identity = self.config.identity_store.load()?.ok_or(TelemetryError::IdentityMissing)?;
-        validate_stored_credential(&credential, &identity, &self.root_store, &self.roots)?;
-        let now = Utc::now().timestamp();
-        if now < credential.not_before_unix || now >= credential.not_after_unix {
-            return Err(TelemetryError::CredentialExpired);
-        }
-        let (organization_uid, cluster_uid) = credential_parent(&credential)?;
-        let cluster_name = format!("organizations/{organization_uid}/clusters/{cluster_uid}");
-        let client = self.client(&credential, &identity.to_pkcs8_pem()?)?;
-        Ok((cluster_name, cluster_uid.to_owned(), client))
-    }
-
-    fn client(&self, credential: &DeviceCredential, key: &Zeroizing<String>) -> Result<Client, TelemetryError> {
-        let mut pem = Zeroizing::new(Vec::with_capacity(credential.certificate_chain.len() + key.len() + 1));
-        pem.extend_from_slice(credential.certificate_chain.as_bytes());
-        pem.push(b'\n');
-        pem.extend_from_slice(key.as_bytes());
-        let identity = reqwest::Identity::from_pem(&pem).map_err(|_| TelemetryError::IdentityCertificate)?;
-        let roots = self
-            .roots
-            .iter()
-            .map(|root| reqwest::Certificate::from_der(root.as_ref()))
-            .collect::<Result<Vec<_>, _>>()?;
-        Client::builder()
-            .https_only(true)
-            .redirect(reqwest::redirect::Policy::none())
-            .timeout(self.config.schedule.timeout)
-            .tls_certs_only(roots)
-            .identity(identity)
-            .build()
-            .map_err(Into::into)
-    }
-}
-
-fn credential_parent(credential: &DeviceCredential) -> Result<(&str, &str), TelemetryError> {
-    let mut parts = credential.name.split('/');
-    let valid = parts.next() == Some("organizations");
-    let organization_uid = parts.next();
-    let valid = valid && parts.next() == Some("clusters");
-    let cluster_uid = parts.next();
-    let valid = valid && parts.next() == Some("clusterDevices");
-    let device_uid = parts.next();
-    if !valid
-        || organization_uid.is_none_or(str::is_empty)
-        || cluster_uid.is_none_or(str::is_empty)
-        || device_uid != Some(credential.uid.as_str())
-        || parts.next().is_some()
-    {
-        return Err(TelemetryError::CredentialName);
-    }
-    Ok((
-        organization_uid.ok_or(TelemetryError::CredentialName)?,
-        cluster_uid.ok_or(TelemetryError::CredentialName)?,
-    ))
-}
-
-fn retry_after(headers: &header::HeaderMap, now: DateTime<Utc>, maximum: Duration) -> Option<Duration> {
-    let value = headers.get(header::RETRY_AFTER)?.to_str().ok()?;
-    let delay = value.parse::<u64>().ok().map(Duration::from_secs).or_else(|| {
-        DateTime::parse_from_rfc2822(value)
-            .ok()
-            .and_then(|at| (at.with_timezone(&Utc) - now).to_std().ok())
-    })?;
-    Some(delay.min(maximum))
-}
-
-pub(crate) fn is_exact_utc_seconds(value: &str) -> bool {
-    DateTime::parse_from_rfc3339(value).is_ok_and(|time| {
-        time.offset().local_minus_utc() == 0
-            && value.ends_with('Z')
-            && time.with_timezone(&Utc).to_rfc3339_opts(SecondsFormat::Secs, true) == value
-    })
-}
-
-async fn response_reason(response: reqwest::Response) -> Option<String> {
-    #[derive(Deserialize)]
-    struct Envelope {
-        #[serde(default)]
-        details: Vec<Detail>,
-    }
-    #[derive(Deserialize)]
-    struct Detail {
-        #[serde(default)]
-        reason: String,
-    }
-
-    serde_json::from_slice::<Envelope>(&bounded_body(response).await.ok()?)
-        .ok()?
-        .details
-        .into_iter()
-        .find_map(|detail| (!detail.reason.is_empty()).then_some(detail.reason))
-}
-
-async fn bounded_body(mut response: reqwest::Response) -> Result<Vec<u8>, TelemetryError> {
-    let mut body = Vec::new();
-    while let Some(chunk) = response.chunk().await? {
-        if body.len().saturating_add(chunk.len()) > MAX_RESPONSE_BYTES {
-            return Err(TelemetryError::ResponseTooLarge);
-        }
-        body.extend_from_slice(&chunk);
-    }
-    Ok(body)
-}
-
-#[derive(Debug, thiserror::Error)]
-pub(crate) enum TelemetryError {
-    #[error("Connect telemetry endpoint must be an HTTPS base URL without credentials, query, or fragment")]
-    Endpoint,
-    #[error("Connect telemetry root CA configuration is invalid")]
-    RootCertificate,
-    #[error("Connect telemetry retry schedule is invalid")]
-    Schedule,
-    #[error("RustFS is not registered with Connect")]
-    NotRegistered,
-    #[error("the Connect device private key is missing")]
-    IdentityMissing,
-    #[error("the stored Connect certificate and device private key cannot form a TLS identity")]
-    IdentityCertificate,
-    #[error("the stored Connect credential name is invalid")]
-    CredentialName,
-    #[error("the stored Connect device certificate is not currently valid")]
-    CredentialExpired,
-    #[error("Connect telemetry response exceeded 64 KiB")]
-    ResponseTooLarge,
-    #[error(transparent)]
-    Url(#[from] url::ParseError),
-    #[error(transparent)]
-    Transport(#[from] reqwest::Error),
-    #[error(transparent)]
-    Identity(#[from] IdentityError),
-    #[error(transparent)]
-    IdentityStore(#[from] StoreError),
-    #[error(transparent)]
-    CredentialStore(#[from] CredentialStoreError),
-    #[error(transparent)]
-    CredentialValidation(#[from] CredentialValidationError),
-}

rustfs/src/startup_services.rs

@@ -13,13 +13,10 @@
 // limitations under the License.
 
 use crate::site_replication_reconcile::spawn_site_replication_reconcile_task;
-use crate::storage_api::startup::services::{ECStore, EndpointServerPools, ServerContextSlot, StorageAdminApi};
+use crate::storage_api::startup::services::{ECStore, EndpointServerPools, ServerContextSlot};
 use crate::{
     config::Config,
-    connect::{
-        CoarseNodeSummary, HeartbeatConfig, HeartbeatRuntime, InventoryError, InventoryFlag, InventoryRuntime, InventorySchedule,
-        InventorySnapshot, spawn_heartbeat_runtime, spawn_inventory_runtime,
-    },
+    connect::{CoarseNodeSummary, HeartbeatConfig, HeartbeatRuntime, spawn_heartbeat_runtime},
     init::{init_buffer_profile_system, init_kms_system},
     server::ServiceStateManager,
     startup_audit::init_audit_runtime,
@@ -80,9 +77,6 @@ pub(crate) async fn init_startup_runtime_services(
     let optional_runtimes = init_optional_runtime_services().await?;
     let heartbeat_config = HeartbeatConfig::from_env().map_err(std::io::Error::other)?;
     let heartbeat_nodes = heartbeat_config.as_ref().map(|_| endpoint_pools.get_nodes().len());
-    let inventory_drives = heartbeat_config
-        .as_ref()
-        .map(|_| endpoint_pools.as_ref().iter().map(|pool| pool.endpoints.as_ref().len()).sum());
 
     init_buffer_profile_system(config);
     init_deadlock_detector_runtime();
@@ -102,9 +96,7 @@ pub(crate) async fn init_startup_runtime_services(
     init_notification_runtime(endpoint_pools, buckets).await?;
     let enable_scanner = init_background_service_runtime(store.clone()).await?;
     init_observability_runtime(store.clone(), ctx.clone()).await;
-    let heartbeat = start_heartbeat_runtime(heartbeat_config.clone(), heartbeat_nodes, &ctx)?;
-    let inventory = start_inventory_runtime(heartbeat_config, heartbeat_nodes, inventory_drives, store, &ctx)?;
-    let heartbeat = heartbeat.map(|heartbeat| heartbeat.with_inventory(inventory));
+    let heartbeat = start_heartbeat_runtime(heartbeat_config, heartbeat_nodes, &ctx)?;
 
     Ok(StartupServiceRuntime {
         optional_runtimes,
@@ -128,69 +120,3 @@ fn start_heartbeat_runtime(
         .ok_or_else(|| std::io::Error::other("Connect heartbeat node count is outside protocol bounds"))?;
     spawn_heartbeat_runtime(Some(config), shutdown, move || summary).map_err(std::io::Error::other)
 }
-
-fn start_inventory_runtime(
-    config: Option<HeartbeatConfig>,
-    node_count: Option<usize>,
-    expected_drive_count: Option<usize>,
-    store: Arc<ECStore>,
-    shutdown: &CancellationToken,
-) -> Result<Option<InventoryRuntime>> {
-    let Some(config) = config else {
-        return Ok(None);
-    };
-    let node_count = node_count.unwrap_or_default();
-    let expected_drive_count = expected_drive_count.unwrap_or_default();
-    spawn_inventory_runtime(Some(config), InventorySchedule::default(), shutdown, move || {
-        let store = store.clone();
-        async move {
-            let info = StorageAdminApi::storage_info(store.as_ref()).await;
-            inventory_snapshot(node_count, expected_drive_count, info)
-        }
-    })
-    .map_err(std::io::Error::other)
-}
-
-fn inventory_snapshot(
-    node_count: usize,
-    expected_drive_count: usize,
-    info: rustfs_madmin::StorageInfo,
-) -> std::result::Result<InventorySnapshot, InventoryError> {
-    if info.disks.len() != expected_drive_count {
-        return Err(InventoryError::SnapshotIncomplete {
-            expected: expected_drive_count,
-            observed: info.disks.len(),
-        });
-    }
-    let total = crate::app::storage_api::capacity::get_total_usable_capacity(&info.disks, &info) as u64;
-    let free = crate::app::storage_api::capacity::get_total_usable_capacity_free(&info.disks, &info) as u64;
-    let mut flags = Vec::with_capacity(3);
-    if info.disks.iter().any(|disk| disk.state == rustfs_madmin::ITEM_OFFLINE) {
-        flags.extend([InventoryFlag::ClusterDegraded, InventoryFlag::DriveOffline]);
-    }
-    if info.disks.iter().any(|disk| disk.healing) {
-        flags.push(InventoryFlag::ClusterHealing);
-    }
-    InventorySnapshot::current(node_count, info.disks.len(), total, free, flags)
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn inventory_rejects_a_partial_startup_storage_snapshot() {
-        let info = rustfs_madmin::StorageInfo {
-            disks: vec![rustfs_madmin::Disk::default()],
-            ..Default::default()
-        };
-
-        assert!(matches!(
-            inventory_snapshot(2, 2, info),
-            Err(InventoryError::SnapshotIncomplete {
-                expected: 2,
-                observed: 1
-            })
-        ));
-    }
-}

rustfs/src/storage_api.rs

@@ -279,7 +279,6 @@ pub(crate) mod startup {
     }
 
     pub(crate) mod services {
-        pub(crate) use super::super::storage_contracts::StorageAdminApi;
         pub(crate) use crate::storage::storage_api::{ECStore, EndpointServerPools, ServerContextSlot};
     }
 

rustfs/tests/connect_inventory.rs

@@ -1,669 +0,0 @@
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-use std::collections::VecDeque;
-use std::fs;
-use std::sync::atomic::{AtomicUsize, Ordering};
-use std::sync::{Arc, Mutex};
-use std::time::Duration;
-
-use bytes::Bytes;
-use http_body_util::{BodyExt as _, Full};
-use hyper::service::service_fn;
-use hyper::{Request, Response, StatusCode};
-use hyper_util::rt::TokioIo;
-use rcgen::{
-    BasicConstraints, CertificateParams, DistinguishedName, DnType, ExtendedKeyUsagePurpose, IsCa, Issuer, KeyPair,
-    KeyUsagePurpose, SanType,
-};
-use rustfs::connect::{
-    CredentialStore, DeviceCredential, HeartbeatConfig, HeartbeatSchedule, IdentityStore, InventoryFlag, InventoryOsVersion,
-    InventorySchedule, InventorySnapshot, InventoryStatus, OperatingSystemFamily, spawn_inventory_runtime,
-};
-use rustls::RootCertStore;
-use rustls::pki_types::{CertificateDer, PrivateKeyDer, PrivatePkcs8KeyDer};
-use rustls::server::WebPkiClientVerifier;
-use serde_json::{Value, json};
-use time::OffsetDateTime;
-use tokio::net::TcpListener;
-use tokio::sync::watch;
-use tokio_rustls::TlsAcceptor;
-use tokio_util::sync::CancellationToken;
-
-const ORGANIZATION_UID: &str = "0198f4b0-1a00-7c10-8d21-2e3f4a5b6c70";
-const CLUSTER_UID: &str = "0198f4b0-2b00-7d20-9e31-3f4a5b6c7d81";
-const DEVICE_UID: &str = "0198f4b0-3c00-7e30-8f41-4a5b6c7d8e92";
-const SNAPSHOT_UID: &str = "0198f4b0-4d00-7f40-9051-5b6c7d8e9fa3";
-
-struct TestPki {
-    root_params: CertificateParams,
-    root_key: KeyPair,
-    root_der: CertificateDer<'static>,
-    root_pem: String,
-    server_der: CertificateDer<'static>,
-    server_key: PrivatePkcs8KeyDer<'static>,
-}
-
-impl TestPki {
-    fn new() -> Self {
-        let now = OffsetDateTime::now_utc();
-        let root_key = KeyPair::generate().expect("generate root key");
-        let mut root_params = CertificateParams::default();
-        root_params.is_ca = IsCa::Ca(BasicConstraints::Unconstrained);
-        root_params.not_before = now - time::Duration::days(30);
-        root_params.not_after = now + time::Duration::days(30);
-        root_params.key_usages = vec![KeyUsagePurpose::KeyCertSign, KeyUsagePurpose::DigitalSignature];
-        let root = root_params.self_signed(&root_key).expect("sign root");
-
-        let server_key = KeyPair::generate().expect("generate server key");
-        let mut server_params = CertificateParams::default();
-        server_params.not_before = now - time::Duration::hours(1);
-        server_params.not_after = now + time::Duration::days(2);
-        server_params
-            .subject_alt_names
-            .push(SanType::DnsName("localhost".try_into().expect("valid DNS name")));
-        server_params.extended_key_usages = vec![ExtendedKeyUsagePurpose::ServerAuth];
-        let server = server_params
-            .signed_by(&server_key, &Issuer::from_params(&root_params, &root_key))
-            .expect("sign server certificate");
-        Self {
-            root_params,
-            root_key,
-            root_der: root.der().clone(),
-            root_pem: root.pem(),
-            server_der: server.der().clone(),
-            server_key: PrivatePkcs8KeyDer::from(server_key.serialize_der()),
-        }
-    }
-
-    fn server_config(&self) -> rustls::ServerConfig {
-        let mut roots = RootCertStore::empty();
-        roots.add(self.root_der.clone()).expect("add client root");
-        let verifier = WebPkiClientVerifier::builder(Arc::new(roots))
-            .build()
-            .expect("client verifier");
-        rustls::ServerConfig::builder()
-            .with_client_cert_verifier(verifier)
-            .with_single_cert(vec![self.server_der.clone()], PrivateKeyDer::Pkcs8(self.server_key.clone_key()))
-            .expect("server TLS")
-    }
-
-    fn stores(&self, temp: &tempfile::TempDir) -> (IdentityStore, CredentialStore) {
-        let identity_store = IdentityStore::new(temp.path().join("identity"));
-        let identity = identity_store.load_or_create().expect("create identity");
-        let private_key = PrivatePkcs8KeyDer::from(identity.to_pkcs8_der().expect("serialize key").to_vec());
-        let device_key = KeyPair::from_pkcs8_der_and_sign_algo(&private_key, &rcgen::PKCS_ECDSA_P256_SHA256).expect("device key");
-        let now = OffsetDateTime::now_utc();
-        let mut params = CertificateParams::default();
-        params.not_before = now - time::Duration::hours(1);
-        params.not_after = now + time::Duration::hours(23);
-        params.serial_number = Some(vec![1; 16].into());
-        params.key_usages = vec![KeyUsagePurpose::DigitalSignature];
-        params.extended_key_usages = vec![ExtendedKeyUsagePurpose::ClientAuth];
-        params.distinguished_name = DistinguishedName::new();
-        params.distinguished_name.push(DnType::CommonName, DEVICE_UID);
-        params.subject_alt_names.push(SanType::URI(
-            format!("urn:rustfs:connect:device:{DEVICE_UID}")
-                .try_into()
-                .expect("device URI"),
-        ));
-        let certificate = params
-            .signed_by(&device_key, &Issuer::from_params(&self.root_params, &self.root_key))
-            .expect("device certificate");
-        let cluster = format!("organizations/{ORGANIZATION_UID}/clusters/{CLUSTER_UID}");
-        let credential = DeviceCredential {
-            name: format!("{cluster}/clusterDevices/{DEVICE_UID}"),
-            uid: DEVICE_UID.to_owned(),
-            protocol_version: "v1".to_owned(),
-            key_id: format!("x509-{}", "01".repeat(16)),
-            certificate_serial: "01".repeat(16),
-            certificate: certificate.pem(),
-            certificate_chain: certificate.pem(),
-            not_before_unix: (now - time::Duration::hours(1)).unix_timestamp(),
-            not_after_unix: (now + time::Duration::hours(23)).unix_timestamp(),
-        };
-        let directory = temp.path().join("credential");
-        fs::create_dir_all(&directory).expect("credential directory");
-        let path = directory.join("device.crt.json");
-        fs::write(&path, serde_json::to_vec(&credential).expect("credential JSON")).expect("write credential");
-        private_mode(&path);
-        (identity_store, CredentialStore::new(directory))
-    }
-}
-
-#[derive(Clone)]
-struct Reply {
-    status: StatusCode,
-    body: Value,
-    retry_after: Option<&'static str>,
-}
-
-impl Reply {
-    fn ok(content_hash: &str) -> Self {
-        Self {
-            status: StatusCode::OK,
-            body: json!({
-                "name": format!("organizations/{ORGANIZATION_UID}/clusters/{CLUSTER_UID}/inventorySnapshots/{SNAPSHOT_UID}"),
-                "uid": SNAPSHOT_UID,
-                "contentHash": content_hash,
-                "receivedAt": "2026-08-22T01:02:03Z",
-                "futureField": true
-            }),
-            retry_after: None,
-        }
-    }
-
-    fn error(status: StatusCode, reason: &str) -> Self {
-        Self {
-            status,
-            body: json!({"details": [{"reason": reason}]}),
-            retry_after: None,
-        }
-    }
-}
-
-struct TestServer {
-    endpoint: String,
-    seen: Arc<Mutex<Vec<Value>>>,
-    task: tokio::task::JoinHandle<()>,
-}
-
-impl Drop for TestServer {
-    fn drop(&mut self) {
-        self.task.abort();
-    }
-}
-
-async fn server(pki: &TestPki, replies: Vec<Reply>) -> TestServer {
-    let listener = TcpListener::bind("127.0.0.1:0").await.expect("bind server");
-    let address = listener.local_addr().expect("server address");
-    let acceptor = TlsAcceptor::from(Arc::new(pki.server_config()));
-    let replies = Arc::new(Mutex::new(VecDeque::from(replies)));
-    let seen = Arc::new(Mutex::new(Vec::new()));
-    let captured = seen.clone();
-    let task = tokio::spawn(async move {
-        while let Ok((stream, _)) = listener.accept().await {
-            let acceptor = acceptor.clone();
-            let replies = replies.clone();
-            let seen = captured.clone();
-            tokio::spawn(async move {
-                let Ok(stream) = acceptor.accept(stream).await else { return };
-                let service = service_fn(move |request: Request<hyper::body::Incoming>| {
-                    let replies = replies.clone();
-                    let seen = seen.clone();
-                    async move {
-                        assert_eq!(request.uri().path(), format!("/agent/clusters/{CLUSTER_UID}/inventorySnapshots"));
-                        let body = request.into_body().collect().await.expect("request body").to_bytes();
-                        seen.lock()
-                            .expect("seen lock")
-                            .push(serde_json::from_slice(&body).expect("request JSON"));
-                        let reply = replies
-                            .lock()
-                            .expect("reply lock")
-                            .pop_front()
-                            .unwrap_or_else(|| Reply::error(StatusCode::SERVICE_UNAVAILABLE, "UNAVAILABLE"));
-                        let mut builder = Response::builder()
-                            .status(reply.status)
-                            .header("content-type", "application/json");
-                        if let Some(value) = reply.retry_after {
-                            builder = builder.header("retry-after", value);
-                        }
-                        Ok::<_, hyper::Error>(
-                            builder
-                                .body(Full::new(Bytes::from(serde_json::to_vec(&reply.body).expect("reply JSON"))))
-                                .expect("reply"),
-                        )
-                    }
-                });
-                let _ = hyper::server::conn::http1::Builder::new()
-                    .serve_connection(TokioIo::new(stream), service)
-                    .await;
-            });
-        }
-    });
-    TestServer {
-        endpoint: format!("https://localhost:{}/agent/", address.port()),
-        seen,
-        task,
-    }
-}
-
-fn config(temp: &tempfile::TempDir, pki: &TestPki, server: &TestServer) -> HeartbeatConfig {
-    let (identity_store, credential_store) = pki.stores(temp);
-    HeartbeatConfig {
-        endpoint: server.endpoint.clone(),
-        root_ca_pem: pki.root_pem.as_bytes().to_vec(),
-        identity_store,
-        credential_store,
-        state_path: temp.path().join("private-config-secret/heartbeat/state.json"),
-        schedule: HeartbeatSchedule {
-            cadence: Duration::from_secs(30),
-            jitter: Duration::ZERO,
-            timeout: Duration::from_millis(200),
-            initial_backoff: Duration::from_millis(20),
-            max_backoff: Duration::from_millis(80),
-        },
-    }
-}
-
-fn schedule() -> InventorySchedule {
-    InventorySchedule {
-        cadence: Duration::from_secs(60),
-        jitter: Duration::ZERO,
-    }
-}
-
-fn snapshot() -> InventorySnapshot {
-    InventorySnapshot::new(
-        "1.4.2",
-        Some(InventoryOsVersion::new(OperatingSystemFamily::Linux, 6, 8).expect("valid operating-system version")),
-        8,
-        96,
-        1_099_511_627_776,
-        412_316_860_416,
-        [InventoryFlag::ClusterDegraded, InventoryFlag::DriveOffline],
-    )
-    .expect("valid inventory")
-}
-
-fn collect_strings(value: &Value, strings: &mut Vec<String>) {
-    match value {
-        Value::String(value) => strings.push(value.clone()),
-        Value::Array(values) => values.iter().for_each(|value| collect_strings(value, strings)),
-        Value::Object(values) => values.values().for_each(|value| collect_strings(value, strings)),
-        _ => {}
-    }
-}
-
-async fn wait_for(
-    status: &mut watch::Receiver<InventoryStatus>,
-    predicate: impl Fn(&InventoryStatus) -> bool,
-) -> InventoryStatus {
-    tokio::time::timeout(Duration::from_secs(3), async {
-        loop {
-            let current = status.borrow_and_update().clone();
-            if predicate(&current) {
-                return current;
-            }
-            status.changed().await.expect("status channel");
-        }
-    })
-    .await
-    .expect("inventory status timeout")
-}
-
-#[test]
-fn connect_inventory_frozen_vector_has_the_exact_canonical_hash_and_no_open_ended_fields() {
-    let fixtures: Value = serde_json::from_str(include_str!("../../protocol/agent/v1/fixtures/inventory/valid-vectors.json"))
-        .expect("valid fixture JSON");
-    let expected = &fixtures["vectors"][0]["expected"];
-    let snapshot = snapshot();
-
-    assert_eq!(snapshot.content_hash().expect("content hash"), expected["contentHash"]);
-    assert_eq!(InventorySchedule::default().cadence, Duration::from_secs(6 * 60 * 60));
-    assert_eq!(InventorySchedule::default().jitter, Duration::from_secs(30 * 60));
-    let encoded = serde_json::to_value(snapshot).expect("snapshot JSON");
-    assert_eq!(
-        encoded,
-        json!({
-            "rustfsVersion": "1.4.2",
-            "osVersion": {"family": "linux", "major": 6, "minor": 8},
-            "nodeCount": 8,
-            "driveCount": 96,
-            "capacityTotalBytes": 1099511627776_u64,
-            "capacityUsedBytes": 412316860416_u64,
-            "coarseFlags": ["cluster.degraded", "drive.offline"]
-        })
-    );
-
-    let fixtures: Value =
-        serde_json::from_str(include_str!("../../protocol/agent/v1/fixtures/inventory/secret-like-vectors.json"))
-            .expect("valid secret-like fixture JSON");
-    let known_fields = [
-        "protocolVersion",
-        "rustfsVersion",
-        "osVersion",
-        "nodeCount",
-        "driveCount",
-        "capacityTotalBytes",
-        "capacityUsedBytes",
-        "coarseFlags",
-    ];
-    let known_flags = ["cluster.degraded", "drive.offline"];
-    let mut excluded = Vec::new();
-    for vector in fixtures["vectors"].as_array().expect("fixture vectors") {
-        let input = vector["input"].as_object().expect("fixture input");
-        for (name, value) in input {
-            if !known_fields.contains(&name.as_str()) {
-                collect_strings(value, &mut excluded);
-            }
-        }
-        for (name, value) in input["osVersion"].as_object().expect("fixture OS version") {
-            if !["family", "major", "minor"].contains(&name.as_str()) {
-                collect_strings(value, &mut excluded);
-            }
-        }
-        for flag in input["coarseFlags"].as_array().expect("fixture coarse flags") {
-            let flag = flag.as_str().expect("fixture coarse flag");
-            if !known_flags.contains(&flag) {
-                excluded.push(flag.to_owned());
-            }
-        }
-    }
-    let encoded = serde_json::to_string(&encoded).expect("encoded snapshot");
-    for value in excluded {
-        assert!(!encoded.contains(&value), "snapshot exposed fixture value {value}");
-    }
-}
-
-#[test]
-fn connect_inventory_bounds_fail_instead_of_truncating_or_inventing_values() {
-    assert!(matches!(
-        InventorySnapshot::current(0, 0, 0, 0, []),
-        Err(rustfs::connect::InventoryError::NodeCount)
-    ));
-    assert!(matches!(
-        InventorySnapshot::current(1, 1_048_577, 0, 0, []),
-        Err(rustfs::connect::InventoryError::DriveCount)
-    ));
-    assert!(matches!(
-        InventorySnapshot::current(1, 0, 9_007_199_254_740_992, 0, []),
-        Err(rustfs::connect::InventoryError::Capacity)
-    ));
-    assert!(matches!(
-        InventorySnapshot::current(1, 0, 10, 11, []),
-        Err(rustfs::connect::InventoryError::Capacity)
-    ));
-    assert!(matches!(
-        InventorySnapshot::new("1.0.0-private.1", None, 1, 0, 0, 0, []),
-        Err(rustfs::connect::InventoryError::RustfsVersion)
-    ));
-}
-
-#[tokio::test]
-async fn connect_inventory_restart_replays_the_pending_request_and_then_skips_unchanged_inventory() {
-    let pki = TestPki::new();
-    let content_hash = snapshot().content_hash().expect("content hash");
-    let first_server = server(&pki, vec![Reply::error(StatusCode::SERVICE_UNAVAILABLE, "UNAVAILABLE")]).await;
-    let temp = tempfile::tempdir().expect("tempdir");
-    let shutdown = CancellationToken::new();
-    let samples = Arc::new(AtomicUsize::new(0));
-    let sampled = samples.clone();
-    let runtime = spawn_inventory_runtime(Some(config(&temp, &pki, &first_server)), schedule(), &shutdown, move || {
-        sampled.fetch_add(1, Ordering::Relaxed);
-        std::future::ready(Ok(snapshot()))
-    })
-    .expect("start inventory")
-    .expect("configured inventory");
-    let mut status = runtime.status();
-
-    assert!(matches!(
-        wait_for(&mut status, |status| matches!(status, InventoryStatus::BackingOff { .. })).await,
-        InventoryStatus::BackingOff { delay } if delay == Duration::from_millis(20)
-    ));
-    assert_eq!(samples.load(Ordering::Relaxed), 1);
-    let original = first_server.seen.lock().expect("seen lock")[0].clone();
-    runtime.shutdown().await;
-
-    let mut limited = Reply::error(StatusCode::TOO_MANY_REQUESTS, "RATE_LIMITED");
-    limited.retry_after = Some("0");
-    let restart_server = server(&pki, vec![limited, Reply::ok(&content_hash)]).await;
-    let restart_config = config(&temp, &pki, &restart_server);
-    let restart_samples = Arc::new(AtomicUsize::new(0));
-    let sampled = restart_samples.clone();
-    let restart = spawn_inventory_runtime(Some(restart_config.clone()), schedule(), &shutdown, move || {
-        sampled.fetch_add(1, Ordering::Relaxed);
-        std::future::ready(Ok(snapshot()))
-    })
-    .expect("restart inventory")
-    .expect("configured inventory");
-    let mut status = restart.status();
-
-    assert!(matches!(
-        wait_for(&mut status, |status| matches!(status, InventoryStatus::Online { .. })).await,
-        InventoryStatus::Online { content_hash: accepted, received_at }
-            if accepted == content_hash && received_at == "2026-08-22T01:02:03Z"
-    ));
-    assert_eq!(restart_samples.load(Ordering::Relaxed), 0);
-    let delivered = restart_server.seen.lock().expect("seen lock").clone();
-    assert_eq!(delivered, vec![original.clone(), original.clone()]);
-    assert_eq!(original["sequence"], 0);
-    let encoded = serde_json::to_string(&original).expect("request JSON");
-    for forbidden in [
-        "private-config-secret",
-        "BEGIN CERTIFICATE",
-        "AKIAIOSFODNN7EXAMPLE",
-        "bucket",
-        "object",
-        "path",
-    ] {
-        assert!(!encoded.contains(forbidden), "request exposed {forbidden}");
-    }
-    assert_eq!(original.as_object().expect("request object").len(), 10);
-    restart.shutdown().await;
-
-    let unchanged_samples = Arc::new(AtomicUsize::new(0));
-    let sampled = unchanged_samples.clone();
-    let unchanged = spawn_inventory_runtime(Some(restart_config), schedule(), &shutdown, move || {
-        sampled.fetch_add(1, Ordering::Relaxed);
-        std::future::ready(Ok(snapshot()))
-    })
-    .expect("restart inventory")
-    .expect("configured inventory");
-    let mut unchanged_status = unchanged.status();
-    assert!(matches!(
-        wait_for(&mut unchanged_status, |status| matches!(status, InventoryStatus::Unchanged { .. })).await,
-        InventoryStatus::Unchanged { content_hash: unchanged } if unchanged == content_hash
-    ));
-    assert_eq!(unchanged_samples.load(Ordering::Relaxed), 1);
-    assert_eq!(restart_server.seen.lock().expect("seen lock").len(), 2);
-    unchanged.shutdown().await;
-}
-
-#[tokio::test]
-async fn connect_inventory_disconnect_retries_without_resampling() {
-    let pki = TestPki::new();
-    let unavailable = server(&pki, Vec::new()).await;
-    let temp = tempfile::tempdir().expect("tempdir");
-    let config = config(&temp, &pki, &unavailable);
-    drop(unavailable);
-    let shutdown = CancellationToken::new();
-    let samples = Arc::new(AtomicUsize::new(0));
-    let sampled = samples.clone();
-    let runtime = spawn_inventory_runtime(Some(config), schedule(), &shutdown, move || {
-        sampled.fetch_add(1, Ordering::Relaxed);
-        std::future::ready(Ok(snapshot()))
-    })
-    .expect("start inventory")
-    .expect("configured inventory");
-    let mut status = runtime.status();
-
-    assert!(matches!(
-        wait_for(&mut status, |status| {
-            matches!(status, InventoryStatus::BackingOff { delay } if *delay == Duration::from_millis(40))
-        })
-        .await,
-        InventoryStatus::BackingOff { delay } if delay == Duration::from_millis(40)
-    ));
-    assert_eq!(samples.load(Ordering::Relaxed), 1);
-    runtime.shutdown().await;
-}
-
-#[tokio::test]
-async fn connect_inventory_retries_an_incomplete_sample_before_delivery() {
-    let pki = TestPki::new();
-    let content_hash = snapshot().content_hash().expect("content hash");
-    let server = server(&pki, vec![Reply::ok(&content_hash)]).await;
-    let temp = tempfile::tempdir().expect("tempdir");
-    let shutdown = CancellationToken::new();
-    let samples = Arc::new(AtomicUsize::new(0));
-    let sampled = samples.clone();
-    let runtime = spawn_inventory_runtime(Some(config(&temp, &pki, &server)), schedule(), &shutdown, move || {
-        let attempt = sampled.fetch_add(1, Ordering::Relaxed);
-        std::future::ready(if attempt == 0 {
-            Err(rustfs::connect::InventoryError::SnapshotIncomplete {
-                expected: 96,
-                observed: 12,
-            })
-        } else {
-            Ok(snapshot())
-        })
-    })
-    .expect("start inventory")
-    .expect("configured inventory");
-    let mut status = runtime.status();
-
-    assert!(matches!(
-        wait_for(&mut status, |status| matches!(status, InventoryStatus::Online { .. })).await,
-        InventoryStatus::Online { content_hash: accepted, .. } if accepted == content_hash
-    ));
-    assert_eq!(samples.load(Ordering::Relaxed), 2);
-    assert_eq!(server.seen.lock().expect("seen lock").len(), 1);
-    runtime.shutdown().await;
-}
-
-#[tokio::test]
-async fn connect_inventory_unchanged_sample_resets_incomplete_backoff() {
-    let pki = TestPki::new();
-    let content_hash = snapshot().content_hash().expect("content hash");
-    let server = server(&pki, vec![Reply::ok(&content_hash)]).await;
-    let temp = tempfile::tempdir().expect("tempdir");
-    let shutdown = CancellationToken::new();
-    let config = config(&temp, &pki, &server);
-    let seed = spawn_inventory_runtime(Some(config.clone()), schedule(), &shutdown, || std::future::ready(Ok(snapshot())))
-        .expect("start inventory")
-        .expect("configured inventory");
-    let mut seed_status = seed.status();
-
-    assert!(matches!(
-        wait_for(&mut seed_status, |status| matches!(status, InventoryStatus::Online { .. })).await,
-        InventoryStatus::Online { content_hash: accepted, .. } if accepted == content_hash
-    ));
-    seed.shutdown().await;
-
-    let samples = Arc::new(AtomicUsize::new(0));
-    let sampled = samples.clone();
-    let runtime = spawn_inventory_runtime(
-        Some(config),
-        InventorySchedule {
-            cadence: Duration::from_millis(100),
-            jitter: Duration::ZERO,
-        },
-        &shutdown,
-        move || {
-            let attempt = sampled.fetch_add(1, Ordering::Relaxed);
-            std::future::ready(if matches!(attempt, 0 | 1 | 3) {
-                Err(rustfs::connect::InventoryError::SnapshotIncomplete {
-                    expected: 96,
-                    observed: 12,
-                })
-            } else {
-                Ok(snapshot())
-            })
-        },
-    )
-    .expect("restart inventory")
-    .expect("configured inventory");
-    let mut status = runtime.status();
-
-    assert!(matches!(
-        wait_for(&mut status, |status| {
-            matches!(status, InventoryStatus::BackingOff { delay } if *delay == Duration::from_millis(20))
-        })
-        .await,
-        InventoryStatus::BackingOff { delay } if delay == Duration::from_millis(20)
-    ));
-    assert!(matches!(
-        wait_for(&mut status, |status| {
-            matches!(status, InventoryStatus::BackingOff { delay } if *delay == Duration::from_millis(40))
-        })
-        .await,
-        InventoryStatus::BackingOff { delay } if delay == Duration::from_millis(40)
-    ));
-    assert!(matches!(
-        wait_for(&mut status, |status| matches!(status, InventoryStatus::Unchanged { .. })).await,
-        InventoryStatus::Unchanged { content_hash: unchanged } if unchanged == content_hash
-    ));
-    assert!(matches!(
-        wait_for(&mut status, |status| matches!(status, InventoryStatus::BackingOff { .. })).await,
-        InventoryStatus::BackingOff { delay } if delay == Duration::from_millis(20)
-    ));
-    assert_eq!(samples.load(Ordering::Relaxed), 4);
-    assert_eq!(server.seen.lock().expect("seen lock").len(), 1);
-    runtime.shutdown().await;
-}
-
-#[tokio::test]
-async fn connect_inventory_revoked_device_stops_without_retrying() {
-    let pki = TestPki::new();
-    let server = server(&pki, vec![Reply::error(StatusCode::UNAUTHORIZED, "DEVICE_REVOKED")]).await;
-    let temp = tempfile::tempdir().expect("tempdir");
-    let shutdown = CancellationToken::new();
-    let runtime = spawn_inventory_runtime(Some(config(&temp, &pki, &server)), schedule(), &shutdown, || {
-        std::future::ready(Ok(snapshot()))
-    })
-    .expect("start inventory")
-    .expect("configured inventory");
-    let mut status = runtime.status();
-
-    assert!(matches!(
-        wait_for(&mut status, |status| matches!(status, InventoryStatus::AuthenticationStopped { .. })).await,
-        InventoryStatus::AuthenticationStopped { status: 401, reason: Some(reason) } if reason == "DEVICE_REVOKED"
-    ));
-    assert_eq!(server.seen.lock().expect("seen lock").len(), 1);
-    runtime.shutdown().await;
-}
-
-#[tokio::test]
-async fn connect_inventory_sequence_overflow_fails_before_sampling_or_network_delivery() {
-    let pki = TestPki::new();
-    let server = server(&pki, Vec::new()).await;
-    let temp = tempfile::tempdir().expect("tempdir");
-    let config = config(&temp, &pki, &server);
-    let state = temp.path().join("private-config-secret/inventory/state.json");
-    fs::create_dir_all(state.parent().expect("state directory")).expect("create state directory");
-    fs::write(
-        &state,
-        br#"{"nextSequence":9007199254740992,"pending":null,"lastAcceptedContentHash":null}"#,
-    )
-    .expect("write state");
-    private_mode(&state);
-    let samples = Arc::new(AtomicUsize::new(0));
-    let sampled = samples.clone();
-    let shutdown = CancellationToken::new();
-    let runtime = spawn_inventory_runtime(Some(config), schedule(), &shutdown, move || {
-        sampled.fetch_add(1, Ordering::Relaxed);
-        std::future::ready(Ok(snapshot()))
-    })
-    .expect("start inventory")
-    .expect("configured inventory");
-    let mut status = runtime.status();
-
-    assert!(matches!(
-        wait_for(&mut status, |status| matches!(status, InventoryStatus::Failed { .. })).await,
-        InventoryStatus::Failed { reason } if reason.contains("sequence is exhausted")
-    ));
-    assert_eq!(samples.load(Ordering::Relaxed), 0);
-    assert!(server.seen.lock().expect("seen lock").is_empty());
-    runtime.shutdown().await;
-}
-
-#[cfg(unix)]
-fn private_mode(path: &std::path::Path) {
-    use std::os::unix::fs::PermissionsExt as _;
-    fs::set_permissions(path, fs::Permissions::from_mode(0o600)).expect("private permissions");
-}
-
-#[cfg(not(unix))]
-fn private_mode(_path: &std::path::Path) {}