properly reload a fresh drive when found in a failed state during startup (#20145)

When a drive is in a failed state when a single node multiple drives
deployment is started, a replacement of a fresh disk will not be
properly healed unless the user restarts the node.

Fix this by always adding the new fresh disk to globalLocalDrivesMap. Also
remove globalLocalDrives for simplification, a map to store local node
drives can still be used since the order of local drives of a node is
not defined.

cmd/background-newdisks-heal-ops.go

@@ -362,7 +362,7 @@ func initAutoHeal(ctx context.Context, objAPI ObjectLayer) {
 
 func getLocalDisksToHeal() (disksToHeal Endpoints) {
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 	for _, disk := range localDrives {
 		_, err := disk.DiskInfo(context.Background(), DiskInfoOptions{})

cmd/bucket-replication.go

@@ -3553,7 +3553,7 @@ func (p *ReplicationPool) persistToDrive(ctx context.Context, v MRFReplicateEntr
 	}
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, localDrive := range localDrives {
@@ -3620,7 +3620,7 @@ func (p *ReplicationPool) loadMRF() (mrfRec MRFReplicateEntries, err error) {
 	}
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, localDrive := range localDrives {

cmd/erasure-server-pool.go

@@ -168,7 +168,7 @@ func newErasureServerPools(ctx context.Context, endpointServerPools EndpointServ
 
 	if !globalIsDistErasure {
 		globalLocalDrivesMu.Lock()
-		globalLocalDrives = localDrives
+		globalLocalDrivesMap = make(map[string]StorageAPI, len(localDrives))
 		for _, drive := range localDrives {
 			globalLocalDrivesMap[drive.Endpoint().String()] = drive
 		}

cmd/erasure-sets.go

@@ -262,13 +262,7 @@ func (s *erasureSets) connectDisks(log bool) {
 				if globalIsDistErasure {
 					globalLocalSetDrives[s.poolIndex][setIndex][diskIndex] = disk
 				}
-				for i, ldisk := range globalLocalDrives {
-					_, k, l := ldisk.GetDiskLoc()
-					if k == setIndex && l == diskIndex {
-						globalLocalDrives[i] = disk
-						break
-					}
-				}
+				globalLocalDrivesMap[disk.Endpoint().String()] = disk
 				globalLocalDrivesMu.Unlock()
 			}
 			s.erasureDisksMu.Unlock()
@@ -1135,13 +1129,7 @@ func (s *erasureSets) HealFormat(ctx context.Context, dryRun bool) (res madmin.H
 					if globalIsDistErasure {
 						globalLocalSetDrives[s.poolIndex][m][n] = disk
 					}
-					for i, ldisk := range globalLocalDrives {
-						_, k, l := ldisk.GetDiskLoc()
-						if k == m && l == n {
-							globalLocalDrives[i] = disk
-							break
-						}
-					}
+					globalLocalDrivesMap[disk.Endpoint().String()] = disk
 					globalLocalDrivesMu.Unlock()
 				}
 			}

cmd/globals.go

@@ -414,10 +414,9 @@ var (
 	globalServiceFreezeCnt int32
 	globalServiceFreezeMu  sync.Mutex // Updates.
 
-	// List of local drives to this node, this is only set during server startup,
-	// and is only mutated by HealFormat. Hold globalLocalDrivesMu to access.
-	globalLocalDrives    []StorageAPI
-	globalLocalDrivesMap = make(map[string]StorageAPI)
+	// Map of local drives to this node, this is set during server startup,
+	// disk reconnect and mutated by HealFormat. Hold globalLocalDrivesMu to access.
+	globalLocalDrivesMap map[string]StorageAPI
 	globalLocalDrivesMu  sync.RWMutex
 
 	globalDriveMonitoring = env.Get("_MINIO_DRIVE_ACTIVE_MONITORING", config.EnableOn) == config.EnableOn

cmd/metrics-resource.go

@@ -262,7 +262,7 @@ func collectDriveMetrics(m madmin.RealtimeMetrics) {
 	latestDriveStatsMu.Unlock()
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, d := range localDrives {

cmd/peer-rest-server.go

@@ -664,7 +664,7 @@ var errUnsupportedSignal = fmt.Errorf("unsupported signal")
 
 func waitingDrivesNode() map[string]madmin.DiskMetrics {
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	errs := make([]error, len(localDrives))

cmd/peer-s3-server.go

@@ -34,7 +34,7 @@ const (
 
 func healBucketLocal(ctx context.Context, bucket string, opts madmin.HealOpts) (res madmin.HealResultItem, err error) {
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	// Initialize sync waitgroup.
@@ -158,7 +158,7 @@ func healBucketLocal(ctx context.Context, bucket string, opts madmin.HealOpts) (
 
 func listBucketsLocal(ctx context.Context, opts BucketOptions) (buckets []BucketInfo, err error) {
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	quorum := (len(localDrives) / 2)
@@ -204,15 +204,17 @@ func listBucketsLocal(ctx context.Context, opts BucketOptions) (buckets []Bucket
 	return buckets, nil
 }
 
-func cloneDrives(drives []StorageAPI) []StorageAPI {
-	newDrives := make([]StorageAPI, len(drives))
-	copy(newDrives, drives)
-	return newDrives
+func cloneDrives(drives map[string]StorageAPI) []StorageAPI {
+	copyDrives := make([]StorageAPI, 0, len(drives))
+	for _, drive := range drives {
+		copyDrives = append(copyDrives, drive)
+	}
+	return copyDrives
 }
 
 func getBucketInfoLocal(ctx context.Context, bucket string, opts BucketOptions) (BucketInfo, error) {
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	g := errgroup.WithNErrs(len(localDrives)).WithConcurrency(32)
@@ -261,7 +263,7 @@ func getBucketInfoLocal(ctx context.Context, bucket string, opts BucketOptions)
 
 func deleteBucketLocal(ctx context.Context, bucket string, opts DeleteBucketOptions) error {
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	g := errgroup.WithNErrs(len(localDrives)).WithConcurrency(32)
@@ -299,7 +301,7 @@ func deleteBucketLocal(ctx context.Context, bucket string, opts DeleteBucketOpti
 
 func makeBucketLocal(ctx context.Context, bucket string, opts MakeBucketOptions) error {
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	g := errgroup.WithNErrs(len(localDrives)).WithConcurrency(32)

cmd/storage-rest-server.go

@@ -1340,6 +1340,7 @@ func registerStorageRESTHandlers(router *mux.Router, endpointServerPools Endpoin
 		return collectInternodeStats(httpTraceHdrs(f))
 	}
 
+	globalLocalDrivesMap = make(map[string]StorageAPI)
 	globalLocalSetDrives = make([][][]StorageAPI, len(endpointServerPools))
 	for pool := range globalLocalSetDrives {
 		globalLocalSetDrives[pool] = make([][]StorageAPI, endpointServerPools[pool].SetCount)
@@ -1413,7 +1414,6 @@ func registerStorageRESTHandlers(router *mux.Router, endpointServerPools Endpoin
 				globalLocalDrivesMu.Lock()
 				defer globalLocalDrivesMu.Unlock()
 
-				globalLocalDrives = append(globalLocalDrives, storage)
 				globalLocalDrivesMap[endpoint.String()] = storage
 				globalLocalSetDrives[endpoint.PoolIdx][endpoint.SetIdx][endpoint.DiskIdx] = storage
 				return true