converge clusterRead health into cluster health (#19063)

cmd/erasure-server-pool.go

@@ -2280,55 +2280,23 @@ type HealthOptions struct {
 // was queried
 type HealthResult struct {
 	Healthy       bool
+	HealthyRead   bool
 	HealingDrives int
 	ESHealth      []struct {
 		Maintenance   bool
 		PoolID, SetID int
 		Healthy       bool
+		HealthyRead   bool
 		HealthyDrives int
 		HealingDrives int
 		ReadQuorum    int
 		WriteQuorum   int
 	}
 	WriteQuorum   int
+	ReadQuorum    int
 	UsingDefaults bool
 }
 
-// ReadHealth returns if the cluster can serve read requests
-func (z *erasureServerPools) ReadHealth(ctx context.Context) bool {
-	erasureSetUpCount := make([][]int, len(z.serverPools))
-	for i := range z.serverPools {
-		erasureSetUpCount[i] = make([]int, len(z.serverPools[i].sets))
-	}
-
-	diskIDs := globalNotificationSys.GetLocalDiskIDs(ctx)
-	diskIDs = append(diskIDs, getLocalDiskIDs(z))
-
-	for _, localDiskIDs := range diskIDs {
-		for _, id := range localDiskIDs {
-			poolIdx, setIdx, _, err := z.getPoolAndSet(id)
-			if err != nil {
-				logger.LogIf(ctx, err)
-				continue
-			}
-			erasureSetUpCount[poolIdx][setIdx]++
-		}
-	}
-
-	b := z.BackendInfo()
-	poolReadQuorums := make([]int, len(b.StandardSCData))
-	copy(poolReadQuorums, b.StandardSCData)
-
-	for poolIdx := range erasureSetUpCount {
-		for setIdx := range erasureSetUpCount[poolIdx] {
-			if erasureSetUpCount[poolIdx][setIdx] < poolReadQuorums[poolIdx] {
-				return false
-			}
-		}
-	}
-	return true
-}
-
 // Health - returns current status of the object layer health,
 // provides if write access exists across sets, additionally
 // can be used to query scenarios if health may be lost
@@ -2397,9 +2365,20 @@ func (z *erasureServerPools) Health(ctx context.Context, opts HealthOptions) Hea
 		}
 	}
 
+	var maximumReadQuorum int
+	for _, readQuorum := range poolReadQuorums {
+		if maximumReadQuorum == 0 {
+			maximumReadQuorum = readQuorum
+		}
+		if readQuorum > maximumReadQuorum {
+			maximumReadQuorum = readQuorum
+		}
+	}
+
 	result := HealthResult{
 		Healthy:       true,
 		WriteQuorum:   maximumWriteQuorum,
+		ReadQuorum:    maximumReadQuorum,
 		UsingDefaults: usingDefaults, // indicates if config was not initialized and we are using defaults on this node.
 	}
 
@@ -2409,6 +2388,7 @@ func (z *erasureServerPools) Health(ctx context.Context, opts HealthOptions) Hea
 				Maintenance                  bool
 				PoolID, SetID                int
 				Healthy                      bool
+				HealthyRead                  bool
 				HealthyDrives, HealingDrives int
 				ReadQuorum, WriteQuorum      int
 			}{
@@ -2416,6 +2396,7 @@ func (z *erasureServerPools) Health(ctx context.Context, opts HealthOptions) Hea
 				SetID:         setIdx,
 				PoolID:        poolIdx,
 				Healthy:       erasureSetUpCount[poolIdx][setIdx].online >= poolWriteQuorums[poolIdx],
+				HealthyRead:   erasureSetUpCount[poolIdx][setIdx].online >= poolReadQuorums[poolIdx],
 				HealthyDrives: erasureSetUpCount[poolIdx][setIdx].online,
 				HealingDrives: erasureSetUpCount[poolIdx][setIdx].healing,
 				ReadQuorum:    poolReadQuorums[poolIdx],
@@ -2428,6 +2409,12 @@ func (z *erasureServerPools) Health(ctx context.Context, opts HealthOptions) Hea
 					fmt.Errorf("Write quorum may be lost on pool: %d, set: %d, expected write quorum: %d",
 						poolIdx, setIdx, poolWriteQuorums[poolIdx]))
 			}
+			result.HealthyRead = erasureSetUpCount[poolIdx][setIdx].online >= poolReadQuorums[poolIdx]
+			if !result.HealthyRead {
+				logger.LogIf(logger.SetReqInfo(ctx, reqInfo),
+					fmt.Errorf("Read quorum may be lost on pool: %d, set: %d, expected read quorum: %d",
+						poolIdx, setIdx, poolReadQuorums[poolIdx]))
+			}
 		}
 	}
 

cmd/healthcheck-handler.go

@@ -81,12 +81,28 @@ func ClusterReadCheckHandler(w http.ResponseWriter, r *http.Request) {
 	ctx, cancel := context.WithTimeout(ctx, globalAPIConfig.getClusterDeadline())
 	defer cancel()
 
-	result := objLayer.ReadHealth(ctx)
+	opts := HealthOptions{
-	if !result {
+		Maintenance:    r.Form.Get("maintenance") == "true",
-		writeResponse(w, http.StatusServiceUnavailable, nil, mimeNone)
+		DeploymentType: r.Form.Get("deployment-type"),
+	}
+	result := objLayer.Health(ctx, opts)
+	w.Header().Set(xhttp.MinIOReadQuorum, strconv.Itoa(result.ReadQuorum))
+	w.Header().Set(xhttp.MinIOStorageClassDefaults, strconv.FormatBool(result.UsingDefaults))
+	// return how many drives are being healed if any
+	if result.HealingDrives > 0 {
+		w.Header().Set(xhttp.MinIOHealingDrives, strconv.Itoa(result.HealingDrives))
+	}
+	if !result.HealthyRead {
+		// As a maintenance call we are purposefully asked to be taken
+		// down, this is for orchestrators to know if we can safely
+		// take this server down, return appropriate error.
+		if opts.Maintenance {
+			writeResponse(w, http.StatusPreconditionFailed, nil, mimeNone)
+		} else {
+			writeResponse(w, http.StatusServiceUnavailable, nil, mimeNone)
+		}
 		return
 	}
-
 	writeResponse(w, http.StatusOK, nil, mimeNone)
 }
 

cmd/notification.go

@@ -1104,24 +1104,6 @@ func (sys *NotificationSys) ServerInfo(metrics bool) []madmin.ServerProperties {
 	return reply
 }
 
-// GetLocalDiskIDs - return disk ids of the local disks of the peers.
-func (sys *NotificationSys) GetLocalDiskIDs(ctx context.Context) (localDiskIDs [][]string) {
-	localDiskIDs = make([][]string, len(sys.peerClients))
-	var wg sync.WaitGroup
-	for idx, client := range sys.peerClients {
-		if client == nil {
-			continue
-		}
-		wg.Add(1)
-		go func(idx int, client *peerRESTClient) {
-			defer wg.Done()
-			localDiskIDs[idx] = client.GetLocalDiskIDs(ctx)
-		}(idx, client)
-	}
-	wg.Wait()
-	return localDiskIDs
-}
-
 // returns all the peers that are currently online.
 func (sys *NotificationSys) getOnlinePeers() []*peerRESTClient {
 	var peerClients []*peerRESTClient

cmd/object-api-interface.go

@@ -287,7 +287,6 @@ type ObjectLayer interface {
 
 	// Returns health of the backend
 	Health(ctx context.Context, opts HealthOptions) HealthResult
-	ReadHealth(ctx context.Context) bool
 
 	// Metadata operations
 	PutObjectMetadata(context.Context, string, string, ObjectOptions) (ObjectInfo, error)

cmd/peer-rest-client.go

@@ -601,21 +601,6 @@ func (client *peerRESTClient) BackgroundHealStatus() (madmin.BgHealState, error)
 	return state, err
 }
 
-// GetLocalDiskIDs - get a peer's local disks' IDs.
-func (client *peerRESTClient) GetLocalDiskIDs(ctx context.Context) (diskIDs []string) {
-	conn := client.gridConn()
-	if conn == nil {
-		return
-	}
-
-	resp, err := getLocalDiskIDsHandler.Call(ctx, conn, grid.NewMSS())
-	if err != nil {
-		return
-	}
-
-	return resp.IDs
-}
-
 // GetMetacacheListing - get a new or existing metacache.
 func (client *peerRESTClient) GetMetacacheListing(ctx context.Context, o listPathOptions) (*metacache, error) {
 	if client == nil {

cmd/peer-rest-server.go

@@ -715,59 +715,11 @@ func (s *peerRESTServer) PutBucketNotificationHandler(w http.ResponseWriter, r *
 	globalEventNotifier.AddRulesMap(bucketName, rulesMap)
 }
 
-// Return disk IDs of all the local disks.
-func getLocalDiskIDs(z *erasureServerPools) []string {
-	var ids []string
-
-	for poolIdx := range z.serverPools {
-		for _, set := range z.serverPools[poolIdx].sets {
-			disks := set.getDisks()
-			for _, disk := range disks {
-				if disk == nil {
-					continue
-				}
-				if disk.IsLocal() {
-					id, err := disk.GetDiskID()
-					if err != nil {
-						continue
-					}
-					if id == "" {
-						continue
-					}
-					ids = append(ids, id)
-				}
-			}
-		}
-	}
-
-	return ids
-}
-
 // HealthHandler - returns true of health
 func (s *peerRESTServer) HealthHandler(w http.ResponseWriter, r *http.Request) {
 	s.IsValid(w, r)
 }
 
-var getLocalDiskIDsHandler = grid.NewSingleHandler[*grid.MSS, *LocalDiskIDs](grid.HandlerGetLocalDiskIDs, grid.NewMSS, func() *LocalDiskIDs {
-	return &LocalDiskIDs{}
-})
-
-// GetLocalDiskIDs - Return disk IDs of all the local disks.
-func (s *peerRESTServer) GetLocalDiskIDs(mss *grid.MSS) (*LocalDiskIDs, *grid.RemoteErr) {
-	objLayer := newObjectLayerFn()
-	// Service not initialized yet
-	if objLayer == nil {
-		return nil, grid.NewRemoteErr(errServerNotInitialized)
-	}
-
-	z, ok := objLayer.(*erasureServerPools)
-	if !ok {
-		return nil, grid.NewRemoteErr(errServerNotInitialized)
-	}
-
-	return &LocalDiskIDs{IDs: getLocalDiskIDs(z)}, nil
-}
-
 // VerifyBinary - verifies the downloaded binary is in-tact
 func (s *peerRESTServer) VerifyBinaryHandler(w http.ResponseWriter, r *http.Request) {
 	if !s.IsValid(w, r) {
@@ -1591,7 +1543,6 @@ func registerPeerRESTHandlers(router *mux.Router, gm *grid.Manager) {
 	logger.FatalIf(reloadSiteReplicationConfigHandler.Register(gm, server.ReloadSiteReplicationConfigHandler), "unable to register handler")
 	logger.FatalIf(loadBucketMetadataHandler.Register(gm, server.LoadBucketMetadataHandler), "unable to register handler")
 	logger.FatalIf(deleteBucketMetadataHandler.Register(gm, server.DeleteBucketMetadataHandler), "unable to register handler")
-	logger.FatalIf(getLocalDiskIDsHandler.Register(gm, server.GetLocalDiskIDs), "unable to register handler")
 	logger.FatalIf(listenHandler.RegisterNoInput(gm, server.ListenHandler), "unable to register handler")
 	logger.FatalIf(gm.RegisterStreamingHandler(grid.HandlerTrace, grid.StreamHandler{
 		Handle:      server.TraceHandler,

internal/grid/handlers.go

@@ -62,7 +62,6 @@ const (
 	HandlerServerVerify
 	HandlerTrace
 	HandlerListen
-	HandlerGetLocalDiskIDs
 	HandlerDeleteBucketMetadata
 	HandlerLoadBucketMetadata
 	HandlerReloadSiteReplicationConfig
@@ -119,7 +118,6 @@ var handlerPrefixes = [handlerLast]string{
 	HandlerServerVerify:                bootstrapPrefix,
 	HandlerTrace:                       peerPrefix,
 	HandlerListen:                      peerPrefix,
-	HandlerGetLocalDiskIDs:             peerPrefix,
 	HandlerDeleteBucketMetadata:        peerPrefix,
 	HandlerLoadBucketMetadata:          peerPrefix,
 	HandlerReloadSiteReplicationConfig: peerPrefix,

internal/grid/handlers_string.go

@@ -32,34 +32,33 @@ func _() {
 	_ = x[HandlerServerVerify-21]
 	_ = x[HandlerTrace-22]
 	_ = x[HandlerListen-23]
-	_ = x[HandlerGetLocalDiskIDs-24]
+	_ = x[HandlerDeleteBucketMetadata-24]
-	_ = x[HandlerDeleteBucketMetadata-25]
+	_ = x[HandlerLoadBucketMetadata-25]
-	_ = x[HandlerLoadBucketMetadata-26]
+	_ = x[HandlerReloadSiteReplicationConfig-26]
-	_ = x[HandlerReloadSiteReplicationConfig-27]
+	_ = x[HandlerReloadPoolMeta-27]
-	_ = x[HandlerReloadPoolMeta-28]
+	_ = x[HandlerStopRebalance-28]
-	_ = x[HandlerStopRebalance-29]
+	_ = x[HandlerLoadRebalanceMeta-29]
-	_ = x[HandlerLoadRebalanceMeta-30]
+	_ = x[HandlerLoadTransitionTierConfig-30]
-	_ = x[HandlerLoadTransitionTierConfig-31]
+	_ = x[HandlerDeletePolicy-31]
-	_ = x[HandlerDeletePolicy-32]
+	_ = x[HandlerLoadPolicy-32]
-	_ = x[HandlerLoadPolicy-33]
+	_ = x[HandlerLoadPolicyMapping-33]
-	_ = x[HandlerLoadPolicyMapping-34]
+	_ = x[HandlerDeleteServiceAccount-34]
-	_ = x[HandlerDeleteServiceAccount-35]
+	_ = x[HandlerLoadServiceAccount-35]
-	_ = x[HandlerLoadServiceAccount-36]
+	_ = x[HandlerDeleteUser-36]
-	_ = x[HandlerDeleteUser-37]
+	_ = x[HandlerLoadUser-37]
-	_ = x[HandlerLoadUser-38]
+	_ = x[HandlerLoadGroup-38]
-	_ = x[HandlerLoadGroup-39]
+	_ = x[HandlerHealBucket-39]
-	_ = x[HandlerHealBucket-40]
+	_ = x[HandlerMakeBucket-40]
-	_ = x[HandlerMakeBucket-41]
+	_ = x[HandlerHeadBucket-41]
-	_ = x[HandlerHeadBucket-42]
+	_ = x[HandlerDeleteBucket-42]
-	_ = x[HandlerDeleteBucket-43]
+	_ = x[handlerTest-43]
-	_ = x[handlerTest-44]
+	_ = x[handlerTest2-44]
-	_ = x[handlerTest2-45]
+	_ = x[handlerLast-45]
-	_ = x[handlerLast-46]
 }
 
-const _HandlerID_name = "handlerInvalidLockLockLockRLockLockUnlockLockRUnlockLockRefreshLockForceUnlockWalkDirStatVolDiskInfoNSScannerReadXLReadVersionDeleteFileDeleteVersionUpdateMetadataWriteMetadataCheckPartsRenameDataRenameFileReadAllServerVerifyTraceListenGetLocalDiskIDsDeleteBucketMetadataLoadBucketMetadataReloadSiteReplicationConfigReloadPoolMetaStopRebalanceLoadRebalanceMetaLoadTransitionTierConfigDeletePolicyLoadPolicyLoadPolicyMappingDeleteServiceAccountLoadServiceAccountDeleteUserLoadUserLoadGroupHealBucketMakeBucketHeadBucketDeleteBuckethandlerTesthandlerTest2handlerLast"
+const _HandlerID_name = "handlerInvalidLockLockLockRLockLockUnlockLockRUnlockLockRefreshLockForceUnlockWalkDirStatVolDiskInfoNSScannerReadXLReadVersionDeleteFileDeleteVersionUpdateMetadataWriteMetadataCheckPartsRenameDataRenameFileReadAllServerVerifyTraceListenDeleteBucketMetadataLoadBucketMetadataReloadSiteReplicationConfigReloadPoolMetaStopRebalanceLoadRebalanceMetaLoadTransitionTierConfigDeletePolicyLoadPolicyLoadPolicyMappingDeleteServiceAccountLoadServiceAccountDeleteUserLoadUserLoadGroupHealBucketMakeBucketHeadBucketDeleteBuckethandlerTesthandlerTest2handlerLast"
 
-var _HandlerID_index = [...]uint16{0, 14, 22, 31, 41, 52, 63, 78, 85, 92, 100, 109, 115, 126, 136, 149, 163, 176, 186, 196, 206, 213, 225, 230, 236, 251, 271, 289, 316, 330, 343, 360, 384, 396, 406, 423, 443, 461, 471, 479, 488, 498, 508, 518, 530, 541, 553, 564}
+var _HandlerID_index = [...]uint16{0, 14, 22, 31, 41, 52, 63, 78, 85, 92, 100, 109, 115, 126, 136, 149, 163, 176, 186, 196, 206, 213, 225, 230, 236, 256, 274, 301, 315, 328, 345, 369, 381, 391, 408, 428, 446, 456, 464, 473, 483, 493, 503, 515, 526, 538, 549}
 
 func (i HandlerID) String() string {
 	if i >= HandlerID(len(_HandlerID_index)-1) {

internal/http/headers.go

@@ -192,6 +192,9 @@ const (
 	// Writes expected write quorum
 	MinIOWriteQuorum = "x-minio-write-quorum"
 
+	// Reads expected read quorum
+	MinIOReadQuorum = "x-minio-read-quorum"
+
 	// Indicates if we are using default storage class and there was problem loading config
 	// if this header is set to "true"
 	MinIOStorageClassDefaults = "x-minio-storage-class-defaults"