CHANGELOG.md

@@ -13,6 +13,7 @@ Changelog for FrostFS Node
 - Reload config for pprof and metrics on SIGHUP in `neofs-node` (#1868)
 - Multiple configs support (#44)
 - Parameters `nns-name` and `nns-zone` for command `frostfs-cli container create` (#37)
+- Tree service now saves the last synchronization height which persists across restarts (#82)
 
 ### Changed
 - Change `frostfs_node_engine_container_size` to counting sizes of logical objects

pkg/services/tree/service.go

@@ -31,10 +31,8 @@ type Service struct {
 	syncChan chan struct{}
 	syncPool *ants.Pool
 
-	// cnrMap maps contrainer and tree ID to the minimum height which was fetched from _each_ client.
-	// This allows us to better handle split-brain scenario, because we always synchronize
-	// from the last seen height. The inner map is read-only and should not be modified in-place.
-	cnrMap map[cidSDK.ID]map[string]uint64
+	// cnrMap contains existing (used) container IDs.
+	cnrMap map[cidSDK.ID]struct{}
 	// cnrMapMtx protects cnrMap
 	cnrMapMtx sync.Mutex
 }
@@ -63,7 +61,7 @@ func New(opts ...Option) *Service {
 	s.replicateLocalCh = make(chan applyOp)
 	s.replicationTasks = make(chan replicationTask, s.replicatorWorkerCount)
 	s.containerCache.init(s.containerCacheSize)
-	s.cnrMap = make(map[cidSDK.ID]map[string]uint64)
+	s.cnrMap = make(map[cidSDK.ID]struct{})
 	s.syncChan = make(chan struct{})
 	s.syncPool, _ = ants.NewPool(defaultSyncWorkerCount)
 

pkg/services/tree/sync.go

@@ -86,31 +86,24 @@ func (s *Service) synchronizeAllTrees(ctx context.Context, cid cid.ID) error {
 		return fmt.Errorf("could not fetch tree ID list: %w", outErr)
 	}
 
-	s.cnrMapMtx.Lock()
-	oldStatus := s.cnrMap[cid]
-	s.cnrMapMtx.Unlock()
-
-	syncStatus := map[string]uint64{}
-	for i := range treesToSync {
-		syncStatus[treesToSync[i]] = 0
-	}
-	for tid := range oldStatus {
-		if _, ok := syncStatus[tid]; ok {
-			syncStatus[tid] = oldStatus[tid]
-		}
-	}
-
 	for _, tid := range treesToSync {
-		h := s.synchronizeTree(ctx, d, syncStatus[tid], tid, nodes)
-		if syncStatus[tid] < h {
-			syncStatus[tid] = h
+		h, err := s.forest.TreeLastSyncHeight(d.CID, tid)
+		if err != nil && !errors.Is(err, pilorama.ErrTreeNotFound) {
+			s.log.Warn("could not get last synchronized height for a tree",
+				zap.Stringer("cid", d.CID),
+				zap.String("tree", tid))
+			continue
+		}
+		newHeight := s.synchronizeTree(ctx, d, h, tid, nodes)
+		if h < newHeight {
+			if err := s.forest.TreeUpdateLastSyncHeight(d.CID, tid, newHeight); err != nil {
+				s.log.Warn("could not update last synchronized height for a tree",
+					zap.Stringer("cid", d.CID),
+					zap.String("tree", tid))
+			}
 		}
 	}
 
-	s.cnrMapMtx.Lock()
-	s.cnrMap[cid] = syncStatus
-	s.cnrMapMtx.Unlock()
-
 	return nil
 }