Add StalledDiskPrimary analysis and recovery to vtorc (#17470)

Signed-off-by: Manan Gupta <manan@planetscale.com>

Author: GuptaManan100

changelog/22.0/22.0.0/summary.md

@@ -12,6 +12,7 @@
   - **[Support for More Efficient JSON Replication](#efficient-json-replication)**
   - **[Support for LAST_INSERT_ID(x)](#last-insert-id)**
   - **[Support for Maximum Idle Connections in the Pool](#max-idle-connections)**
+  - **[Stalled Disk Recovery in VTOrc](#stall-disk-recovery)**
 - **[Minor Changes](#minor-changes)**
   - **[VTTablet Flags](#flags-vttablet)**
   - **[Topology read concurrency behaviour changes](#topo-read-concurrency-changes)**
@@ -100,6 +101,11 @@ You can control idle connection retention for the query server’s query pool, s
 
 This feature ensures that, during traffic spikes, idle connections are available for faster responses, while minimizing overhead in low-traffic periods by limiting the number of idle connections retained. It helps strike a balance between performance, efficiency, and cost.
 
+### <a id="stall-disk-recovery"/>Stalled Disk Recovery in VTOrc</a>
+VTOrc can now identify and recover from stalled disk errors. VTTablets test whether the disk is writable and they send this information in the full status output to VTOrc. If the disk is not writable on the primary tablet, VTOrc will attempt to recover the cluster by promoting a new primary. This is useful in scenarios where the disk is stalled and the primary vttablet is unable to accept writes because of it.
+
+To opt into this feature, `--enable-primary-disk-stalled-recovery` flag has to be specified on VTOrc, and `--disk-write-dir` flag has to be specified on the vttablets. `--disk-write-interval` and `--disk-write-timeout` flags can be used to configure the polling interval and timeout respectively. 
+
 ## <a id="minor-changes"/>Minor Changes</a>
 
 #### <a id="flags-vttablet"/>VTTablet Flags</a>

go/flags/endtoend/vtcombo.txt

@@ -102,6 +102,9 @@ Flags:
       --ddl_strategy string                                              Set default strategy for DDL statements. Override with @@ddl_strategy session variable (default "direct")
       --default_tablet_type topodatapb.TabletType                        The default tablet type to set for queries, when one is not explicitly selected. (default PRIMARY)
       --degraded_threshold duration                                      replication lag after which a replica is considered degraded (default 30s)
+      --disk-write-dir string                                            if provided, tablet will attempt to write a file to this directory to check if the disk is stalled
+      --disk-write-interval duration                                     how often to write to the disk to check whether it is stalled (default 5s)
+      --disk-write-timeout duration                                      if writes exceed this duration, the disk is considered stalled (default 30s)
       --emit_stats                                                       If set, emit stats to push-based monitoring and stats backends
       --enable-consolidator                                              Synonym to -enable_consolidator (default true)
       --enable-consolidator-replicas                                     Synonym to -enable_consolidator_replicas

go/flags/endtoend/vtorc.txt

@@ -33,6 +33,7 @@ Flags:
       --config-type string                                          Config file type (omit to infer config type from file extension).
       --consul_auth_static_file string                              JSON File to read the topos/tokens from.
       --emit_stats                                                  If set, emit stats to push-based monitoring and stats backends
+      --enable-primary-disk-stalled-recovery                        Whether VTOrc should detect a stalled disk on the primary and failover
       --grpc-dial-concurrency-limit int                             Maximum concurrency of grpc dial operations. This should be less than the golang max thread limit of 10000. (default 1024)
       --grpc_auth_static_client_creds string                        When using grpc_static_auth in the server, this file provides the credentials to use to authenticate with server.
       --grpc_compression string                                     Which protocol to use for compressing gRPC. Default: nothing. Supported: snappy

go/flags/endtoend/vttablet.txt

@@ -133,6 +133,9 @@ Flags:
       --dba_idle_timeout duration                                        Idle timeout for dba connections (default 1m0s)
       --dba_pool_size int                                                Size of the connection pool for dba connections (default 20)
       --degraded_threshold duration                                      replication lag after which a replica is considered degraded (default 30s)
+      --disk-write-dir string                                            if provided, tablet will attempt to write a file to this directory to check if the disk is stalled
+      --disk-write-interval duration                                     how often to write to the disk to check whether it is stalled (default 5s)
+      --disk-write-timeout duration                                      if writes exceed this duration, the disk is considered stalled (default 30s)
       --emit_stats                                                       If set, emit stats to push-based monitoring and stats backends
       --enable-consolidator                                              Synonym to -enable_consolidator (default true)
       --enable-consolidator-replicas                                     Synonym to -enable_consolidator_replicas

go/vt/vtorc/config/config.go

@@ -174,6 +174,15 @@ var (
 			Dynamic:  true,
 		},
 	)
+
+	enablePrimaryDiskStalledRecovery = viperutil.Configure(
+		"enable-primary-disk-stalled-recovery",
+		viperutil.Options[bool]{
+			FlagName: "enable-primary-disk-stalled-recovery",
+			Default:  false,
+			Dynamic:  true,
+		},
+	)
 )
 
 func init() {
@@ -197,6 +206,7 @@ func registerFlags(fs *pflag.FlagSet) {
 	fs.Duration("recovery-poll-duration", recoveryPollDuration.Default(), "Timer duration on which VTOrc polls its database to run a recovery")
 	fs.Bool("allow-emergency-reparent", ersEnabled.Default(), "Whether VTOrc should be allowed to run emergency reparent operation when it detects a dead primary")
 	fs.Bool("change-tablets-with-errant-gtid-to-drained", convertTabletsWithErrantGTIDs.Default(), "Whether VTOrc should be changing the type of tablets with errant GTIDs to DRAINED")
+	fs.Bool("enable-primary-disk-stalled-recovery", enablePrimaryDiskStalledRecovery.Default(), "Whether VTOrc should detect a stalled disk on the primary and failover")
 
 	viperutil.BindFlags(fs,
 		instancePollTime,
@@ -214,6 +224,7 @@ func registerFlags(fs *pflag.FlagSet) {
 		recoveryPollDuration,
 		ersEnabled,
 		convertTabletsWithErrantGTIDs,
+		enablePrimaryDiskStalledRecovery,
 	)
 }
 
@@ -332,6 +343,11 @@ func SetConvertTabletWithErrantGTIDs(val bool) {
 	convertTabletsWithErrantGTIDs.Set(val)
 }
 
+// GetStalledDiskPrimaryRecovery reports whether VTOrc is allowed to check for and recovery stalled disk problems.
+func GetStalledDiskPrimaryRecovery() bool {
+	return enablePrimaryDiskStalledRecovery.Get()
+}
+
 // MarkConfigurationLoaded is called once configuration has first been loaded.
 // Listeners on ConfigurationLoaded will get a notification
 func MarkConfigurationLoaded() {

go/vt/vtorc/db/generate_base.go

@@ -105,6 +105,7 @@ CREATE TABLE database_instance (
 	semi_sync_primary_status TINYint NOT NULL DEFAULT 0,
 	semi_sync_replica_status TINYint NOT NULL DEFAULT 0,
 	semi_sync_primary_clients int NOT NULL DEFAULT 0,
+	is_disk_stalled TINYint NOT NULL DEFAULT 0,
 	PRIMARY KEY (alias)
 )`,
 	`

go/vt/vtorc/inst/analysis.go

@@ -56,6 +56,7 @@ const (
 	LockedSemiSyncPrimaryHypothesis        AnalysisCode = "LockedSemiSyncPrimaryHypothesis"
 	LockedSemiSyncPrimary                  AnalysisCode = "LockedSemiSyncPrimary"
 	ErrantGTIDDetected                     AnalysisCode = "ErrantGTIDDetected"
+	PrimaryDiskStalled                     AnalysisCode = "PrimaryDiskStalled"
 )
 
 type StructureAnalysisCode string
@@ -129,6 +130,7 @@ type ReplicationAnalysis struct {
 	MaxReplicaGTIDMode                        string
 	MaxReplicaGTIDErrant                      string
 	IsReadOnly                                bool
+	IsDiskStalled                             bool
 }
 
 func (replicationAnalysis *ReplicationAnalysis) MarshalJSON() ([]byte, error) {

go/vt/vtorc/inst/analysis_dao.go

@@ -79,7 +79,7 @@ func GetReplicationAnalysis(keyspace string, shard string, hints *ReplicationAna
 		vitess_keyspace.durability_policy AS durability_policy,
 		vitess_shard.primary_timestamp AS shard_primary_term_timestamp,
 		primary_instance.read_only AS read_only,
-		MIN(primary_instance.gtid_errant) AS gtid_errant, 
+		MIN(primary_instance.gtid_errant) AS gtid_errant,
 		MIN(primary_instance.alias) IS NULL AS is_invalid,
 		MIN(primary_instance.binary_log_file) AS binary_log_file,
 		MIN(primary_instance.binary_log_pos) AS binary_log_pos,
@@ -233,7 +233,8 @@ func GetReplicationAnalysis(keyspace string, shard string, hints *ReplicationAna
 		COUNT(
 			DISTINCT case when replica_instance.log_bin
 			AND replica_instance.log_replica_updates then replica_instance.major_version else NULL end
-		) AS count_distinct_logging_major_versions
+		) AS count_distinct_logging_major_versions,
+		primary_instance.is_disk_stalled != 0 AS is_disk_stalled
 	FROM
 		vitess_tablet
 		JOIN vitess_keyspace ON (
@@ -354,6 +355,7 @@ func GetReplicationAnalysis(keyspace string, shard string, hints *ReplicationAna
 		a.HeartbeatInterval = m.GetFloat64("heartbeat_interval")
 
 		a.IsReadOnly = m.GetUint("read_only") == 1
+		a.IsDiskStalled = m.GetBool("is_disk_stalled")
 
 		if !a.LastCheckValid {
 			analysisMessage := fmt.Sprintf("analysis: Alias: %+v, Keyspace: %+v, Shard: %+v, IsPrimary: %+v, LastCheckValid: %+v, LastCheckPartialSuccess: %+v, CountReplicas: %+v, CountValidReplicas: %+v, CountValidReplicatingReplicas: %+v, CountLaggingReplicas: %+v, CountDelayedReplicas: %+v",
@@ -401,6 +403,10 @@ func GetReplicationAnalysis(keyspace string, shard string, hints *ReplicationAna
 		} else if isInvalid {
 			a.Analysis = InvalidReplica
 			a.Description = "VTOrc hasn't been able to reach the replica even once since restart/shutdown"
+		} else if a.IsClusterPrimary && !a.LastCheckValid && a.IsDiskStalled {
+			a.Analysis = PrimaryDiskStalled
+			a.Description = "Primary has a stalled disk"
+			ca.hasClusterwideAction = true
 		} else if a.IsClusterPrimary && !a.LastCheckValid && a.CountReplicas == 0 {
 			a.Analysis = DeadPrimaryWithoutReplicas
 			a.Description = "Primary cannot be reached by vtorc and has no replica"

go/vt/vtorc/inst/analysis_dao_test.go

@@ -34,10 +34,10 @@ var (
 	// The initialSQL is a set of insert commands copied from a dump of an actual running VTOrc instances. The relevant insert commands are here.
 	// This is a dump taken from a test running 4 tablets, zone1-101 is the primary, zone1-100 is a replica, zone1-112 is a rdonly and zone2-200 is a cross-cell replica.
 	initialSQL = []string{
-		`INSERT INTO database_instance VALUES('zone1-0000000112','localhost',6747,'2022-12-28 07:26:04','2022-12-28 07:26:04',213696377,'8.0.31','ROW',1,1,'vt-0000000112-bin.000001',15963,'localhost',6714,8,4.0,1,1,'vt-0000000101-bin.000001',15583,'vt-0000000101-bin.000001',15583,0,0,1,'','',1,'vt-0000000112-relay-bin.000002',15815,1,0,'zone1','',0,0,0,1,'729a4cc4-8680-11ed-a104-47706090afbd:1-54','729a5138-8680-11ed-9240-92a06c3be3c2','2022-12-28 07:26:04','',1,0,0,'Homebrew','8.0','FULL',10816929,0,0,'ON',1,'729a4cc4-8680-11ed-a104-47706090afbd','','729a4cc4-8680-11ed-a104-47706090afbd,729a5138-8680-11ed-9240-92a06c3be3c2',1,1,'',1000000000000000000,1,0,0,0);`,
-		`INSERT INTO database_instance VALUES('zone1-0000000100','localhost',6711,'2022-12-28 07:26:04','2022-12-28 07:26:04',1094500338,'8.0.31','ROW',1,1,'vt-0000000100-bin.000001',15963,'localhost',6714,8,4.0,1,1,'vt-0000000101-bin.000001',15583,'vt-0000000101-bin.000001',15583,0,0,1,'','',1,'vt-0000000100-relay-bin.000002',15815,1,0,'zone1','',0,0,0,1,'729a4cc4-8680-11ed-a104-47706090afbd:1-54','729a5138-8680-11ed-acf8-d6b0ef9f4eaa','2022-12-28 07:26:04','',1,0,0,'Homebrew','8.0','FULL',10103920,0,1,'ON',1,'729a4cc4-8680-11ed-a104-47706090afbd','','729a4cc4-8680-11ed-a104-47706090afbd,729a5138-8680-11ed-acf8-d6b0ef9f4eaa',1,1,'',1000000000000000000,1,0,1,0);`,
-		`INSERT INTO database_instance VALUES('zone1-0000000101','localhost',6714,'2022-12-28 07:26:04','2022-12-28 07:26:04',390954723,'8.0.31','ROW',1,1,'vt-0000000101-bin.000001',15583,'',0,0,0,0,0,'',0,'',0,NULL,NULL,0,'','',0,'',0,0,0,'zone1','',0,0,0,1,'729a4cc4-8680-11ed-a104-47706090afbd:1-54','729a4cc4-8680-11ed-a104-47706090afbd','2022-12-28 07:26:04','',0,0,0,'Homebrew','8.0','FULL',11366095,1,1,'ON',1,'','','729a4cc4-8680-11ed-a104-47706090afbd',-1,-1,'',1000000000000000000,1,1,0,2);`,
-		`INSERT INTO database_instance VALUES('zone2-0000000200','localhost',6756,'2022-12-28 07:26:05','2022-12-28 07:26:05',444286571,'8.0.31','ROW',1,1,'vt-0000000200-bin.000001',15963,'localhost',6714,8,4.0,1,1,'vt-0000000101-bin.000001',15583,'vt-0000000101-bin.000001',15583,0,0,1,'','',1,'vt-0000000200-relay-bin.000002',15815,1,0,'zone2','',0,0,0,1,'729a4cc4-8680-11ed-a104-47706090afbd:1-54','729a497c-8680-11ed-8ad4-3f51d747db75','2022-12-28 07:26:05','',1,0,0,'Homebrew','8.0','FULL',10443112,0,1,'ON',1,'729a4cc4-8680-11ed-a104-47706090afbd','','729a4cc4-8680-11ed-a104-47706090afbd,729a497c-8680-11ed-8ad4-3f51d747db75',1,1,'',1000000000000000000,1,0,1,0);`,
+		`INSERT INTO database_instance VALUES('zone1-0000000112','localhost',6747,'2022-12-28 07:26:04','2022-12-28 07:26:04',213696377,'8.0.31','ROW',1,1,'vt-0000000112-bin.000001',15963,'localhost',6714,8,4.0,1,1,'vt-0000000101-bin.000001',15583,'vt-0000000101-bin.000001',15583,0,0,1,'','',1,'vt-0000000112-relay-bin.000002',15815,1,0,'zone1','',0,0,0,1,'729a4cc4-8680-11ed-a104-47706090afbd:1-54','729a5138-8680-11ed-9240-92a06c3be3c2','2022-12-28 07:26:04','',1,0,0,'Homebrew','8.0','FULL',10816929,0,0,'ON',1,'729a4cc4-8680-11ed-a104-47706090afbd','','729a4cc4-8680-11ed-a104-47706090afbd,729a5138-8680-11ed-9240-92a06c3be3c2',1,1,'',1000000000000000000,1,0,0,0,false);`,
+		`INSERT INTO database_instance VALUES('zone1-0000000100','localhost',6711,'2022-12-28 07:26:04','2022-12-28 07:26:04',1094500338,'8.0.31','ROW',1,1,'vt-0000000100-bin.000001',15963,'localhost',6714,8,4.0,1,1,'vt-0000000101-bin.000001',15583,'vt-0000000101-bin.000001',15583,0,0,1,'','',1,'vt-0000000100-relay-bin.000002',15815,1,0,'zone1','',0,0,0,1,'729a4cc4-8680-11ed-a104-47706090afbd:1-54','729a5138-8680-11ed-acf8-d6b0ef9f4eaa','2022-12-28 07:26:04','',1,0,0,'Homebrew','8.0','FULL',10103920,0,1,'ON',1,'729a4cc4-8680-11ed-a104-47706090afbd','','729a4cc4-8680-11ed-a104-47706090afbd,729a5138-8680-11ed-acf8-d6b0ef9f4eaa',1,1,'',1000000000000000000,1,0,1,0,false);`,
+		`INSERT INTO database_instance VALUES('zone1-0000000101','localhost',6714,'2022-12-28 07:26:04','2022-12-28 07:26:04',390954723,'8.0.31','ROW',1,1,'vt-0000000101-bin.000001',15583,'',0,0,0,0,0,'',0,'',0,NULL,NULL,0,'','',0,'',0,0,0,'zone1','',0,0,0,1,'729a4cc4-8680-11ed-a104-47706090afbd:1-54','729a4cc4-8680-11ed-a104-47706090afbd','2022-12-28 07:26:04','',0,0,0,'Homebrew','8.0','FULL',11366095,1,1,'ON',1,'','','729a4cc4-8680-11ed-a104-47706090afbd',-1,-1,'',1000000000000000000,1,1,0,2,false);`,
+		`INSERT INTO database_instance VALUES('zone2-0000000200','localhost',6756,'2022-12-28 07:26:05','2022-12-28 07:26:05',444286571,'8.0.31','ROW',1,1,'vt-0000000200-bin.000001',15963,'localhost',6714,8,4.0,1,1,'vt-0000000101-bin.000001',15583,'vt-0000000101-bin.000001',15583,0,0,1,'','',1,'vt-0000000200-relay-bin.000002',15815,1,0,'zone2','',0,0,0,1,'729a4cc4-8680-11ed-a104-47706090afbd:1-54','729a497c-8680-11ed-8ad4-3f51d747db75','2022-12-28 07:26:05','',1,0,0,'Homebrew','8.0','FULL',10443112,0,1,'ON',1,'729a4cc4-8680-11ed-a104-47706090afbd','','729a4cc4-8680-11ed-a104-47706090afbd,729a497c-8680-11ed-8ad4-3f51d747db75',1,1,'',1000000000000000000,1,0,1,0,false);`,
 		`INSERT INTO vitess_tablet VALUES('zone1-0000000100','localhost',6711,'ks','0','zone1',2,'0001-01-01 00:00:00+00:00',X'616c6961733a7b63656c6c3a227a6f6e653122207569643a3130307d20686f73746e616d653a226c6f63616c686f73742220706f72745f6d61703a7b6b65793a2267727063222076616c75653a363731307d20706f72745f6d61703a7b6b65793a227674222076616c75653a363730397d206b657973706163653a226b73222073686172643a22302220747970653a5245504c494341206d7973716c5f686f73746e616d653a226c6f63616c686f737422206d7973716c5f706f72743a363731312064625f7365727665725f76657273696f6e3a22382e302e3331222064656661756c745f636f6e6e5f636f6c6c6174696f6e3a3435');`,
 		`INSERT INTO vitess_tablet VALUES('zone1-0000000101','localhost',6714,'ks','0','zone1',1,'2022-12-28 07:23:25.129898+00:00',X'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');`,
 		`INSERT INTO vitess_tablet VALUES('zone1-0000000112','localhost',6747,'ks','0','zone1',3,'0001-01-01 00:00:00+00:00',X'616c6961733a7b63656c6c3a227a6f6e653122207569643a3131327d20686f73746e616d653a226c6f63616c686f73742220706f72745f6d61703a7b6b65793a2267727063222076616c75653a363734367d20706f72745f6d61703a7b6b65793a227674222076616c75653a363734357d206b657973706163653a226b73222073686172643a22302220747970653a52444f4e4c59206d7973716c5f686f73746e616d653a226c6f63616c686f737422206d7973716c5f706f72743a363734372064625f7365727665725f76657273696f6e3a22382e302e3331222064656661756c745f636f6e6e5f636f6c6c6174696f6e3a3435');`,
@@ -96,6 +96,29 @@ func TestGetReplicationAnalysisDecision(t *testing.T) {
 			keyspaceWanted: "ks",
 			shardWanted:    "0",
 			codeWanted:     PrimaryTabletDeleted,
+		}, {
+			name: "StalledDiskPrimary",
+			info: []*test.InfoForRecoveryAnalysis{{
+				TabletInfo: &topodatapb.Tablet{
+					Alias:         &topodatapb.TabletAlias{Cell: "zon1", Uid: 100},
+					Hostname:      "localhost",
+					Keyspace:      "ks",
+					Shard:         "0",
+					Type:          topodatapb.TabletType_PRIMARY,
+					MysqlHostname: "localhost",
+					MysqlPort:     6709,
+				},
+				DurabilityPolicy:              "none",
+				LastCheckValid:                0,
+				CountReplicas:                 4,
+				CountValidReplicas:            4,
+				CountValidReplicatingReplicas: 0,
+				IsPrimary:                     1,
+				IsStalledDisk:                 1,
+			}},
+			keyspaceWanted: "ks",
+			shardWanted:    "0",
+			codeWanted:     PrimaryDiskStalled,
 		}, {
 			name: "DeadPrimary",
 			info: []*test.InfoForRecoveryAnalysis{{

go/vt/vtorc/inst/instance.go

@@ -91,6 +91,7 @@ type Instance struct {
 	IsUpToDate           bool
 	IsRecentlyChecked    bool
 	SecondsSinceLastSeen sql.NullInt64
+	StalledDisk          bool
 
 	AllowTLS bool