Пример #1
0
func RecoverDeadMaster(analysisEntry inst.ReplicationAnalysis, skipProcesses bool) (promotedSlave *inst.Instance, lostSlaves [](*inst.Instance), err error) {
	failedInstanceKey := &analysisEntry.AnalyzedInstanceKey
	if ok, err := AttemptRecoveryRegistration(&analysisEntry); !ok {
		log.Debugf("topology_recovery: found an active or recent recovery on %+v. Will not issue another RecoverDeadMaster.", *failedInstanceKey)
		return nil, lostSlaves, err
	}

	inst.AuditOperation("recover-dead-master", failedInstanceKey, "problem found; will recover")
	if !skipProcesses {
		if err := executeProcesses(config.Config.PreFailoverProcesses, "PreFailoverProcesses", analysisEntry, nil, emptySlavesList, true); err != nil {
			return nil, lostSlaves, err
		}
	}

	log.Debugf("topology_recovery: RecoverDeadMaster: will recover %+v", *failedInstanceKey)

	var masterRecoveryType MasterRecoveryType = MasterRecoveryPseudoGTID
	if (analysisEntry.OracleGTIDImmediateTopology || analysisEntry.MariaDBGTIDImmediateTopology) && !analysisEntry.PseudoGTIDImmediateTopology {
		masterRecoveryType = MasterRecoveryGTID
	} else if analysisEntry.BinlogServerImmediateTopology {
		masterRecoveryType = MasterRecoveryBinlogServer
	}
	log.Debugf("topology_recovery: RecoverDeadMaster: masterRecoveryType=%+v", masterRecoveryType)

	switch masterRecoveryType {
	case MasterRecoveryGTID:
		{
			lostSlaves, _, promotedSlave, err = inst.RegroupSlavesGTID(failedInstanceKey, true, nil)
		}
	case MasterRecoveryPseudoGTID:
		{
			lostSlaves, _, _, promotedSlave, err = inst.RegroupSlavesIncludingSubSlavesOfBinlogServers(failedInstanceKey, true, nil)
		}
	case MasterRecoveryBinlogServer:
		{
			promotedSlave, err = inst.RegroupSlavesBinlogServers(failedInstanceKey, true, nil)
		}
	}
	if promotedSlave != nil && len(lostSlaves) > 0 && config.Config.DetachLostSlavesAfterMasterFailover {
		log.Debugf("topology_recovery: - RecoverDeadMaster: lost %+v slaves during recovery process; detaching them", len(lostSlaves))
		go func() {
			for _, slave := range lostSlaves {
				slave := slave
				inst.DetachSlaveOperation(&slave.Key)
			}
		}()
	}
	if config.Config.MasterFailoverLostInstancesDowntimeMinutes > 0 {
		inst.BeginDowntime(failedInstanceKey, inst.GetMaintenanceOwner(), "RecoverDeadMaster indicates this instance is lost", config.Config.MasterFailoverLostInstancesDowntimeMinutes*60)
		for _, slave := range lostSlaves {
			slave := slave
			inst.BeginDowntime(&slave.Key, inst.GetMaintenanceOwner(), "RecoverDeadMaster indicates this instance is lost", config.Config.MasterFailoverLostInstancesDowntimeMinutes*60)
		}
	}

	if promotedSlave == nil {
		log.Debugf("topology_recovery: - RecoverDeadMaster: Failure: no slave promoted.")
		inst.AuditOperation("recover-dead-master", failedInstanceKey, "Failure: no slave promoted.")
	} else {
		log.Debugf("topology_recovery: - RecoverDeadMaster: promoted slave is %+v", promotedSlave.Key)
		inst.AuditOperation("recover-dead-master", failedInstanceKey, fmt.Sprintf("master: %+v", promotedSlave.Key))
	}
	return promotedSlave, lostSlaves, err
}
Пример #2
0
// Cli initiates a command line interface, executing requested command.
func Cli(command string, strict bool, instance string, destination string, owner string, reason string, duration string, pattern string, clusterAlias string, pool string, hostnameFlag string) {
	if instance != "" && !strings.Contains(instance, ":") {
		instance = fmt.Sprintf("%s:%d", instance, config.Config.DefaultInstancePort)
	}
	instanceKey, err := inst.ParseInstanceKey(instance)
	if err != nil {
		instanceKey = nil
	}
	rawInstanceKey, err := inst.NewRawInstanceKey(instance)
	if err != nil {
		rawInstanceKey = nil
	}

	if destination != "" && !strings.Contains(destination, ":") {
		destination = fmt.Sprintf("%s:%d", destination, config.Config.DefaultInstancePort)
	}
	destinationKey, err := inst.ParseInstanceKey(destination)
	if err != nil {
		destinationKey = nil
	}

	if hostname, err := os.Hostname(); err == nil {
		thisInstanceKey = &inst.InstanceKey{Hostname: hostname, Port: int(config.Config.DefaultInstancePort)}
	}
	postponedFunctionsContainer := inst.NewPostponedFunctionsContainer()

	if len(owner) == 0 {
		// get os username as owner
		usr, err := user.Current()
		if err != nil {
			log.Fatale(err)
		}
		owner = usr.Username
	}
	inst.SetMaintenanceOwner(owner)

	skipDatabaseCommands := false
	switch command {
	case "reset-internal-db-deployment":
		skipDatabaseCommands = true
	case "help":
		skipDatabaseCommands = true
	}

	if !skipDatabaseCommands {
		process.ContinuousRegistration(string(process.OrchestratorExecutionCliMode), command)
	}
	// begin commands
	switch command {
	// smart mode
	case registerCliCommand("relocate", "Smart relocation", `Relocate a slave beneath another instance`), registerCliCommand("relocate-below", "Smart relocation", `Synonym to 'relocate', will be deprecated`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			_, err := inst.RelocateBelow(instanceKey, destinationKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case registerCliCommand("relocate-slaves", "Smart relocation", `Relocates all or part of the slaves of a given instance under another instance`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			slaves, _, err, errs := inst.RelocateSlaves(instanceKey, destinationKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range slaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case registerCliCommand("regroup-slaves", "Smart relocation", `Given an instance, pick one of its slave and make it local master of its siblings`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			lostSlaves, equalSlaves, aheadSlaves, promotedSlave, err := inst.RegroupSlaves(instanceKey, false, func(candidateSlave *inst.Instance) { fmt.Println(candidateSlave.Key.DisplayString()) }, postponedFunctionsContainer)
			postponedFunctionsContainer.InvokePostponed()
			if promotedSlave == nil {
				log.Fatalf("Could not regroup slaves of %+v; error: %+v", *instanceKey, err)
			}
			fmt.Println(fmt.Sprintf("%s lost: %d, trivial: %d, pseudo-gtid: %d",
				promotedSlave.Key.DisplayString(), len(lostSlaves), len(equalSlaves), len(aheadSlaves)))
			if err != nil {
				log.Fatale(err)
			}
		}
		// General replication commands
		// move, binlog file:pos
	case registerCliCommand("move-up", "Classic file:pos relocation", `Move a slave one level up the topology`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			instance, err := inst.MoveUp(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), instance.MasterKey.DisplayString()))
		}
	case registerCliCommand("move-up-slaves", "Classic file:pos relocation", `Moves slaves of the given instance one level up the topology`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			movedSlaves, _, err, errs := inst.MoveUpSlaves(instanceKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range movedSlaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case registerCliCommand("move-below", "Classic file:pos relocation", `Moves a slave beneath its sibling. Both slaves must be actively replicating from same master.`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination/sibling:", destination)
			}
			_, err := inst.MoveBelow(instanceKey, destinationKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case registerCliCommand("move-equivalent", "Classic file:pos relocation", `Moves a slave beneath another server, based on previously recorded "equivalence coordinates"`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			_, err := inst.MoveEquivalent(instanceKey, destinationKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case registerCliCommand("repoint", "Classic file:pos relocation", `Make the given instance replicate from another instance without changing the binglog coordinates. Use with care`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			// destinationKey can be null, in which case the instance repoints to its existing master
			instance, err := inst.Repoint(instanceKey, destinationKey, inst.GTIDHintNeutral)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), instance.MasterKey.DisplayString()))
		}
	case registerCliCommand("repoint-slaves", "Classic file:pos relocation", `Repoint all slaves of given instance to replicate back from the instance. Use with care`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			repointedSlaves, err, errs := inst.RepointSlavesTo(instanceKey, pattern, destinationKey)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range repointedSlaves {
					fmt.Println(fmt.Sprintf("%s<%s", slave.Key.DisplayString(), instanceKey.DisplayString()))
				}
			}
		}
	case registerCliCommand("enslave-siblings", "Classic file:pos relocation", `Turn all siblings of a slave into its sub-slaves.`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, _, err := inst.EnslaveSiblings(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("enslave-master", "Classic file:pos relocation", `Turn an instance into a master of its own master; essentially switch the two.`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.EnslaveMaster(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("make-co-master", "Classic file:pos relocation", `Create a master-master replication. Given instance is a slave which replicates directly from a master.`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.MakeCoMaster(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("get-candidate-slave", "Classic file:pos relocation", `Information command suggesting the most up-to-date slave of a given instance that is good for promotion`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			instance, _, _, _, err := inst.GetCandidateSlave(instanceKey, false)
			if err != nil {
				log.Fatale(err)
			} else {
				fmt.Println(instance.Key.DisplayString())
			}
		}
	case registerCliCommand("regroup-slaves-bls", "Binlog server relocation", `Regroup Binlog Server slaves of a given instance`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			_, promotedBinlogServer, err := inst.RegroupSlavesBinlogServers(instanceKey, false)
			if promotedBinlogServer == nil {
				log.Fatalf("Could not regroup binlog server slaves of %+v; error: %+v", *instanceKey, err)
			}
			fmt.Println(promotedBinlogServer.Key.DisplayString())
			if err != nil {
				log.Fatale(err)
			}
		}
	// move, GTID
	case registerCliCommand("move-gtid", "GTID relocation", `Move a slave beneath another instance.`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			_, err := inst.MoveBelowGTID(instanceKey, destinationKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case registerCliCommand("move-slaves-gtid", "GTID relocation", `Moves all slaves of a given instance under another (destination) instance using GTID`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			movedSlaves, _, err, errs := inst.MoveSlavesGTID(instanceKey, destinationKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range movedSlaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case registerCliCommand("regroup-slaves-gtid", "GTID relocation", `Given an instance, pick one of its slave and make it local master of its siblings, using GTID.`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			lostSlaves, movedSlaves, promotedSlave, err := inst.RegroupSlavesGTID(instanceKey, false, func(candidateSlave *inst.Instance) { fmt.Println(candidateSlave.Key.DisplayString()) })
			if promotedSlave == nil {
				log.Fatalf("Could not regroup slaves of %+v; error: %+v", *instanceKey, err)
			}
			fmt.Println(fmt.Sprintf("%s lost: %d, moved: %d",
				promotedSlave.Key.DisplayString(), len(lostSlaves), len(movedSlaves)))
			if err != nil {
				log.Fatale(err)
			}
		}
		// Pseudo-GTID
	case registerCliCommand("match", "Pseudo-GTID relocation", `Matches a slave beneath another (destination) instance using Pseudo-GTID`),
		registerCliCommand("match-below", "Pseudo-GTID relocation", `Synonym to 'match', will be deprecated`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			_, _, err := inst.MatchBelow(instanceKey, destinationKey, true)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case registerCliCommand("match-up", "Pseudo-GTID relocation", `Transport the slave one level up the hierarchy, making it child of its grandparent, using Pseudo-GTID`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			instance, _, err := inst.MatchUp(instanceKey, true)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), instance.MasterKey.DisplayString()))
		}
	case registerCliCommand("rematch", "Pseudo-GTID relocation", `Reconnect a slave onto its master, via PSeudo-GTID.`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			instance, _, err := inst.RematchSlave(instanceKey, true)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), instance.MasterKey.DisplayString()))
		}
	case registerCliCommand("match-slaves", "Pseudo-GTID relocation", `Matches all slaves of a given instance under another (destination) instance using Pseudo-GTID`),
		registerCliCommand("multi-match-slaves", "Pseudo-GTID relocation", `Synonym to 'match-slaves', will be deprecated`):
		{
			// Move all slaves of "instance" beneath "destination"
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}

			matchedSlaves, _, err, errs := inst.MultiMatchSlaves(instanceKey, destinationKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range matchedSlaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case registerCliCommand("match-up-slaves", "Pseudo-GTID relocation", `Matches slaves of the given instance one level up the topology, making them siblings of given instance, using Pseudo-GTID`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			matchedSlaves, _, err, errs := inst.MatchUpSlaves(instanceKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range matchedSlaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case registerCliCommand("regroup-slaves-pgtid", "Pseudo-GTID relocation", `Given an instance, pick one of its slave and make it local master of its siblings, using Pseudo-GTID.`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			lostSlaves, equalSlaves, aheadSlaves, promotedSlave, err := inst.RegroupSlavesPseudoGTID(instanceKey, false, func(candidateSlave *inst.Instance) { fmt.Println(candidateSlave.Key.DisplayString()) }, postponedFunctionsContainer)
			postponedFunctionsContainer.InvokePostponed()
			if promotedSlave == nil {
				log.Fatalf("Could not regroup slaves of %+v; error: %+v", *instanceKey, err)
			}
			fmt.Println(fmt.Sprintf("%s lost: %d, trivial: %d, pseudo-gtid: %d",
				promotedSlave.Key.DisplayString(), len(lostSlaves), len(equalSlaves), len(aheadSlaves)))
			if err != nil {
				log.Fatale(err)
			}
		}
		// General replication commands
	case registerCliCommand("enable-gtid", "Replication, general", `If possible, turn on GTID replication`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.EnableGTID(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("disable-gtid", "Replication, general", `Turn off GTID replication, back to file:pos replication`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.DisableGTID(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("reset-master-gtid-remove-own-uuid", "Replication, general", `Reset master on instance, remove GTID entries generated by instance`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.ResetMasterGTIDOperation(instanceKey, true, "")
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("skip-query", "Replication, general", `Skip a single statement on a slave; either when running with GTID or without`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.SkipQuery(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("stop-slave", "Replication, general", `Issue a STOP SLAVE on an instance`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.StopSlave(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("start-slave", "Replication, general", `Issue a START SLAVE on an instance`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.StartSlave(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("restart-slave", "Replication, general", `STOP and START SLAVE on an instance`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.RestartSlave(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("reset-slave", "Replication, general", `Issues a RESET SLAVE command; use with care`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.ResetSlaveOperation(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("detach-slave", "Replication, general", `Stops replication and modifies binlog position into an impossible, yet reversible, value.`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.DetachSlaveOperation(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("reattach-slave", "Replication, general", `Undo a detach-slave operation`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.ReattachSlaveOperation(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("detach-slave-master-host", "Replication, general", `Stops replication and modifies Master_Host into an impossible, yet reversible, value.`):
		{
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.DetachSlaveMasterHost(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("reattach-slave-master-host", "Replication, general", `Undo a detach-slave-master-host operation`):
		{
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.ReattachSlaveMasterHost(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("master-pos-wait", "Replication, general", `Wait until slave reaches given replication coordinates (--binlog=file:pos)`):
		{
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatalf("Unresolved instance")
			}
			instance, err := inst.ReadTopologyInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			var binlogCoordinates *inst.BinlogCoordinates

			if binlogCoordinates, err = inst.ParseBinlogCoordinates(*config.RuntimeCLIFlags.BinlogFile); err != nil {
				log.Fatalf("Expecing --binlog argument as file:pos")
			}
			_, err = inst.MasterPosWait(instanceKey, binlogCoordinates)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
		// Pool
	case registerCliCommand("set-read-only", "Instance", `Turn an instance read-only, via SET GLOBAL read_only := 1`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.SetReadOnly(instanceKey, true)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("set-writeable", "Instance", `Turn an instance writeable, via SET GLOBAL read_only := 0`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			_, err := inst.SetReadOnly(instanceKey, false)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
		// Binary log operations
	case registerCliCommand("flush-binary-logs", "Binary logs", `Flush binary logs on an instance`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			var err error
			if *config.RuntimeCLIFlags.BinlogFile == "" {
				_, err = inst.FlushBinaryLogs(instanceKey, 1)
			} else {
				_, err = inst.FlushBinaryLogsTo(instanceKey, *config.RuntimeCLIFlags.BinlogFile)
			}
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("purge-binary-logs", "Binary logs", `Purge binary logs of an instance`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			var err error
			if *config.RuntimeCLIFlags.BinlogFile == "" {
				log.Fatal("expecting --binlog value")
			}

			_, err = inst.PurgeBinaryLogsTo(instanceKey, *config.RuntimeCLIFlags.BinlogFile)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("last-pseudo-gtid", "Binary logs", `Find latest Pseudo-GTID entry in instance's binary logs`):
		{
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatalf("Unresolved instance")
			}
			instance, err := inst.ReadTopologyInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			coordinates, text, err := inst.FindLastPseudoGTIDEntry(instance, instance.RelaylogCoordinates, nil, strict, nil)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%+v:%s", *coordinates, text))
		}
	case registerCliCommand("find-binlog-entry", "Binary logs", `Get binlog file:pos of entry given by --pattern (exact full match, not a regular expression) in a given instance`):
		{
			if pattern == "" {
				log.Fatal("No pattern given")
			}
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatalf("Unresolved instance")
			}
			instance, err := inst.ReadTopologyInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			coordinates, err := inst.SearchEntryInInstanceBinlogs(instance, pattern, false)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%+v", *coordinates))
		}
	case registerCliCommand("correlate-binlog-pos", "Binary logs", `Given an instance (-i) and binlog coordinates (--binlog=file:pos), find the correlated coordinates in another instance (-d)`):
		{
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatalf("Unresolved instance")
			}
			instance, err := inst.ReadTopologyInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			if !instance.LogBinEnabled {
				log.Fatalf("Instance does not have binary logs: %+v", *instanceKey)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce target instance:", destination)
			}
			otherInstance, err := inst.ReadTopologyInstance(destinationKey)
			if err != nil {
				log.Fatale(err)
			}
			if otherInstance == nil {
				log.Fatalf("Instance not found: %+v", *destinationKey)
			}
			var binlogCoordinates *inst.BinlogCoordinates
			if *config.RuntimeCLIFlags.BinlogFile == "" {
				binlogCoordinates = &instance.SelfBinlogCoordinates
			} else {
				if binlogCoordinates, err = inst.ParseBinlogCoordinates(*config.RuntimeCLIFlags.BinlogFile); err != nil {
					log.Fatalf("Expecing --binlog argument as file:pos")
				}
			}

			coordinates, _, err := inst.CorrelateBinlogCoordinates(instance, binlogCoordinates, otherInstance)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%+v", *coordinates))
		}
		// Pool
	case registerCliCommand("submit-pool-instances", "Pools", `Submit a pool name with a list of instances in that pool`):
		{
			if pool == "" {
				log.Fatal("Please submit --pool")
			}
			err := inst.ApplyPoolInstances(pool, instance)
			if err != nil {
				log.Fatale(err)
			}
		}
	case registerCliCommand("cluster-pool-instances", "Pools", `List all pools and their associated instances`):
		{
			clusterPoolInstances, err := inst.ReadAllClusterPoolInstances()
			if err != nil {
				log.Fatale(err)
			}
			for _, clusterPoolInstance := range clusterPoolInstances {
				fmt.Println(fmt.Sprintf("%s\t%s\t%s\t%s:%d", clusterPoolInstance.ClusterName, clusterPoolInstance.ClusterAlias, clusterPoolInstance.Pool, clusterPoolInstance.Hostname, clusterPoolInstance.Port))
			}
		}
		// Information
	case registerCliCommand("find", "Information", `Find instances whose hostname matches given regex pattern`):
		{
			if pattern == "" {
				log.Fatal("No pattern given")
			}
			instances, err := inst.FindInstances(pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, instance := range instances {
					fmt.Println(instance.Key.DisplayString())
				}
			}
		}
	case registerCliCommand("clusters", "Information", `List all clusters known to orchestrator`):
		{
			clusters, err := inst.ReadClusters()
			if err != nil {
				log.Fatale(err)
			} else {
				fmt.Println(strings.Join(clusters, "\n"))
			}
		}
	case registerCliCommand("topology", "Information", `Show an ascii-graph of a replication topology, given a member of that topology`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			output, err := inst.ASCIITopology(instanceKey, pattern)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(output)
		}
	case registerCliCommand("which-instance", "Information", `Output the fully-qualified hostname:port representation of the given instance, or error if unknown`):
		{
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatalf("Unable to get master: unresolved instance")
			}
			instance, _, err := inst.ReadInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			fmt.Println(instance.Key.DisplayString())
		}
	case registerCliCommand("which-cluster", "Information", `Output the name of the cluster an instance belongs to, or error if unknown to orchestrator`):
		{
			clusterName := getClusterName(clusterAlias, instanceKey)
			fmt.Println(clusterName)
		}
	case registerCliCommand("which-cluster-instances", "Information", `Output the list of instances participating in same cluster as given instance`):
		{
			clusterName := getClusterName(clusterAlias, instanceKey)
			instances, err := inst.ReadClusterInstances(clusterName)
			if err != nil {
				log.Fatale(err)
			}
			for _, clusterInstance := range instances {
				fmt.Println(clusterInstance.Key.DisplayString())
			}
		}
	case registerCliCommand("which-cluster-osc-slaves", "Information", `Output a list of slaves in same cluster as given instance, that could serve as a pt-online-schema-change operation control slaves`):
		{
			clusterName := getClusterName(clusterAlias, instanceKey)
			instances, err := inst.GetClusterOSCSlaves(clusterName)
			if err != nil {
				log.Fatale(err)
			}
			for _, clusterInstance := range instances {
				fmt.Println(clusterInstance.Key.DisplayString())
			}
		}
	case registerCliCommand("which-master", "Information", `Output the fully-qualified hostname:port representation of a given instance's master`):
		{
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatalf("Unable to get master: unresolved instance")
			}
			instance, _, err := inst.ReadInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			fmt.Println(instance.MasterKey.DisplayString())
		}
	case registerCliCommand("which-slaves", "Information", `Output the fully-qualified hostname:port list of slaves of a given instance`):
		{
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatalf("Unable to get slaves: unresolved instance")
			}
			slaves, err := inst.ReadSlaveInstances(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			for _, slave := range slaves {
				fmt.Println(slave.Key.DisplayString())
			}
		}
	case registerCliCommand("instance-status", "Information", `Output short status on a given instance`):
		{
			if instanceKey == nil {
				instanceKey = assignThisInstanceKey()
			}
			if instanceKey == nil {
				log.Fatalf("Unable to get status: unresolved instance")
			}
			instance, _, err := inst.ReadInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			fmt.Println(instance.HumanReadableDescription())
		}
	case registerCliCommand("get-cluster-heuristic-lag", "Information", `For a given cluster (indicated by an instance or alias), output a heuristic "representative" lag of that cluster`):
		{
			clusterName := getClusterName(clusterAlias, instanceKey)
			lag, err := inst.GetClusterHeuristicLag(clusterName)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(lag)
		}
		// Instance management
	case registerCliCommand("discover", "Instance management", `Lookup an instance, investigate it`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			instance, err := inst.ReadTopologyInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instance.Key.DisplayString())
		}
	case registerCliCommand("forget", "Instance management", `Forget about an instance's existence`):
		{
			if rawInstanceKey == nil {
				rawInstanceKey = assignThisInstanceKey()
			}
			if rawInstanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			err := inst.ForgetInstance(rawInstanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(rawInstanceKey.DisplayString())
		}
	case registerCliCommand("begin-maintenance", "Instance management", `Request a maintenance lock on an instance`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			if reason == "" {
				log.Fatal("--reason option required")
			}
			var durationSeconds int = 0
			if duration != "" {
				durationSeconds, err = util.SimpleTimeToSeconds(duration)
				if err != nil {
					log.Fatale(err)
				}
				if durationSeconds < 0 {
					log.Fatalf("Duration value must be non-negative. Given value: %d", durationSeconds)
				}
			}
			maintenanceKey, err := inst.BeginBoundedMaintenance(instanceKey, inst.GetMaintenanceOwner(), reason, uint(durationSeconds))
			if err == nil {
				log.Infof("Maintenance key: %+v", maintenanceKey)
				log.Infof("Maintenance duration: %d seconds", durationSeconds)
			}
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("end-maintenance", "Instance management", `Remove maintenance lock from an instance`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			err := inst.EndMaintenanceByInstanceKey(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("begin-downtime", "Instance management", `Mark an instance as downtimed`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			if reason == "" {
				log.Fatal("--reason option required")
			}
			var durationSeconds int = 0
			if duration != "" {
				durationSeconds, err = util.SimpleTimeToSeconds(duration)
				if err != nil {
					log.Fatale(err)
				}
				if durationSeconds < 0 {
					log.Fatalf("Duration value must be non-negative. Given value: %d", durationSeconds)
				}
			}
			err := inst.BeginDowntime(instanceKey, inst.GetMaintenanceOwner(), reason, uint(durationSeconds))
			if err == nil {
				log.Infof("Downtime duration: %d seconds", durationSeconds)
			} else {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("end-downtime", "Instance management", `Indicate an instance is no longer downtimed`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			err := inst.EndDowntime(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
		// Recovery & analysis
	case registerCliCommand("recover", "Recovery", `Do auto-recovery given a dead instance`), registerCliCommand("recover-lite", "Recovery", `Do auto-recovery given a dead instance. Orchestrator chooses the best course of actionwithout executing external processes`):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			recoveryAttempted, promotedInstanceKey, err := logic.CheckAndRecover(instanceKey, destinationKey, (command == "recover-lite"))
			if err != nil {
				log.Fatale(err)
			}
			if recoveryAttempted {
				if promotedInstanceKey == nil {
					log.Fatalf("Recovery attempted yet no slave promoted")
				}
				fmt.Println(promotedInstanceKey.DisplayString())
			}
		}
	case registerCliCommand("replication-analysis", "Recovery", `Request an analysis of potential crash incidents in all known topologies`):
		{
			analysis, err := inst.GetReplicationAnalysis("", false, false)
			if err != nil {
				log.Fatale(err)
			}
			for _, entry := range analysis {
				fmt.Println(fmt.Sprintf("%s (cluster %s): %s", entry.AnalyzedInstanceKey.DisplayString(), entry.ClusterDetails.ClusterName, entry.Analysis))
			}
		}
	case registerCliCommand("ack-cluster-recoveries", "Recovery", `Acknowledge recoveries for a given cluster; this unblocks pending future recoveries`):
		{
			if reason == "" {
				log.Fatal("--reason option required (comment your ack)")
			}
			clusterName := getClusterName(clusterAlias, instanceKey)
			countRecoveries, err := logic.AcknowledgeClusterRecoveries(clusterName, inst.GetMaintenanceOwner(), reason)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%d recoveries acknowldged", countRecoveries))
		}
	case registerCliCommand("ack-instance-recoveries", "Recovery", `Acknowledge recoveries for a given instance; this unblocks pending future recoveries`):
		{
			if reason == "" {
				log.Fatal("--reason option required (comment your ack)")
			}
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)

			countRecoveries, err := logic.AcknowledgeInstanceRecoveries(instanceKey, inst.GetMaintenanceOwner(), reason)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%d recoveries acknowldged", countRecoveries))
		}
	// Instance meta
	case registerCliCommand("register-candidate", "Instance, meta", `Indicate that a specific instance is a preferred candidate for master promotion`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			err := inst.RegisterCandidateInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("register-hostname-unresolve", "Instance, meta", `Assigns the given instance a virtual (aka "unresolved") name`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			err := inst.RegisterHostnameUnresolve(instanceKey, hostnameFlag)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case registerCliCommand("deregister-hostname-unresolve", "Instance, meta", `Explicitly deregister/dosassociate a hostname with an "unresolved" name`):
		{
			instanceKey = deduceInstanceKeyIfNeeded(instance, instanceKey)
			err := inst.DeregisterHostnameUnresolve(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
		// meta
	case registerCliCommand("snapshot-topologies", "Meta", `Take a snapshot of existing topologies.`):
		{
			err := inst.SnapshotTopologies()
			if err != nil {
				log.Fatale(err)
			}
		}
	case registerCliCommand("continuous", "Meta", `Enter continuous mode, and actively poll for instances, diagnose problems, do maintenance`):
		{
			logic.ContinuousDiscovery()
		}
	case registerCliCommand("resolve", "Meta", `Resolve given hostname`):
		{
			if rawInstanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if conn, err := net.Dial("tcp", rawInstanceKey.DisplayString()); err == nil {
				log.Debugf("tcp test is good; got connection %+v", conn)
				conn.Close()
			} else {
				log.Fatale(err)
			}
			if cname, err := inst.GetCNAME(rawInstanceKey.Hostname); err == nil {
				log.Debugf("GetCNAME() %+v, %+v", cname, err)
				rawInstanceKey.Hostname = cname
				fmt.Println(rawInstanceKey.DisplayString())
			} else {
				log.Fatale(err)
			}
		}
	case registerCliCommand("reset-hostname-resolve-cache", "Meta", `Clear the hostname resolve cache`):
		{
			err := inst.ResetHostnameResolveCache()
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println("hostname resolve cache cleared")
		}
	case registerCliCommand("reset-internal-db-deployment", "Meta, internal", `Clear internal db deployment history, use if somehow corrupted internal deployment history`):
		{
			config.Config.SkipOrchestratorDatabaseUpdate = true
			db.ResetInternalDeployment()
			fmt.Println("Internal db deployment history reset. Next orchestrator execution will rebuild internal db structure (no data will be lost)")
		}
		// Help
	case "help":
		{
			fmt.Fprintf(os.Stderr, availableCommandsUsage())
		}
	default:
		log.Fatalf("Unknown command: \"%s\". %s", command, availableCommandsUsage())
	}
}
Пример #3
0
// Cli initiates a command line interface, executing requested command.
func Cli(command string, strict bool, instance string, destination string, owner string, reason string, duration string, pattern string, clusterAlias string, pool string, hostnameFlag string) {

	if instance != "" && !strings.Contains(instance, ":") {
		instance = fmt.Sprintf("%s:%d", instance, config.Config.DefaultInstancePort)
	}
	instanceKey, err := inst.ParseInstanceKey(instance)
	if err != nil {
		instanceKey = nil
	}
	rawInstanceKey, err := inst.NewRawInstanceKey(instance)
	if err != nil {
		rawInstanceKey = nil
	}

	if destination != "" && !strings.Contains(destination, ":") {
		destination = fmt.Sprintf("%s:%d", destination, config.Config.DefaultInstancePort)
	}
	destinationKey, err := inst.ParseInstanceKey(destination)
	if err != nil {
		destinationKey = nil
	}

	if hostname, err := os.Hostname(); err == nil {
		thisInstanceKey = &inst.InstanceKey{Hostname: hostname, Port: int(config.Config.DefaultInstancePort)}
	}

	if len(owner) == 0 {
		// get os username as owner
		usr, err := user.Current()
		if err != nil {
			log.Fatale(err)
		}
		owner = usr.Username
	}
	inst.SetMaintenanceOwner(owner)

	// begin commands
	switch command {
	// Instance meta
	case cliCommand("discover"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			instance, err := inst.ReadTopologyInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instance.Key.DisplayString())
		}
	case cliCommand("forget"):
		{
			if rawInstanceKey == nil {
				rawInstanceKey = thisInstanceKey
			}
			if rawInstanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			err := inst.ForgetInstance(rawInstanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(rawInstanceKey.DisplayString())
		}
	case cliCommand("resolve"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if conn, err := net.Dial("tcp", instanceKey.DisplayString()); err == nil {
				conn.Close()
			} else {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("register-hostname-unresolve"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			err := inst.RegisterHostnameUnresolve(instanceKey, hostnameFlag)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("deregister-hostname-unresolve"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			err := inst.DeregisterHostnameUnresolve(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("register-candidate"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			err := inst.RegisterCandidateInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
		// Instance
	case cliCommand("begin-maintenance"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if reason == "" {
				log.Fatal("--reason option required")
			}
			var durationSeconds int = 0
			if duration != "" {
				durationSeconds, err = util.SimpleTimeToSeconds(duration)
				if err != nil {
					log.Fatale(err)
				}
				if durationSeconds < 0 {
					log.Fatalf("Duration value must be non-negative. Given value: %d", durationSeconds)
				}
			}
			maintenanceKey, err := inst.BeginBoundedMaintenance(instanceKey, inst.GetMaintenanceOwner(), reason, uint(durationSeconds))
			if err == nil {
				log.Infof("Maintenance key: %+v", maintenanceKey)
				log.Infof("Maintenance duration: %d seconds", durationSeconds)
			}
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("end-maintenance"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			err := inst.EndMaintenanceByInstanceKey(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("begin-downtime"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if reason == "" {
				log.Fatal("--reason option required")
			}
			var durationSeconds int = 0
			if duration != "" {
				durationSeconds, err = util.SimpleTimeToSeconds(duration)
				if err != nil {
					log.Fatale(err)
				}
				if durationSeconds < 0 {
					log.Fatalf("Duration value must be non-negative. Given value: %d", durationSeconds)
				}
			}
			err := inst.BeginDowntime(instanceKey, inst.GetMaintenanceOwner(), reason, uint(durationSeconds))
			if err == nil {
				log.Infof("Downtime duration: %d seconds", durationSeconds)
			} else {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("end-downtime"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			err := inst.EndDowntime(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("set-read-only"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.SetReadOnly(instanceKey, true)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("set-writeable"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.SetReadOnly(instanceKey, false)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("flush-binary-logs"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			var err error
			if *config.RuntimeCLIFlags.BinlogFile == "" {
				err = inst.FlushBinaryLogs(instanceKey, 1)
			} else {
				_, err = inst.FlushBinaryLogsTo(instanceKey, *config.RuntimeCLIFlags.BinlogFile)
			}
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("last-pseudo-gtid"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatalf("Unresolved instance")
			}
			instance, err := inst.ReadTopologyInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			coordinates, text, err := inst.FindLastPseudoGTIDEntry(instance, instance.RelaylogCoordinates, strict, nil)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%+v:%s", *coordinates, text))
		}
		// replication
	case cliCommand("stop-slave"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.StopSlave(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("start-slave"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.StartSlave(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("restart-slave"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.RestartSlave(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("reset-slave"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.ResetSlaveOperation(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("detach-slave"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.DetachSlaveOperation(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("reattach-slave"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.ReattachSlaveOperation(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("enable-gtid"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.EnableGTID(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("disable-gtid"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.DisableGTID(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("skip-query"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.SkipQuery(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
		// move
	case cliCommand("relocate"), cliCommand("relocate-below"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			_, err := inst.RelocateBelow(instanceKey, destinationKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case cliCommand("relocate-slaves"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			slaves, _, err, errs := inst.RelocateSlaves(instanceKey, destinationKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range slaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case cliCommand("move-up"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			instance, err := inst.MoveUp(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), instance.MasterKey.DisplayString()))
		}
	case cliCommand("move-up-slaves"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			movedSlaves, _, err, errs := inst.MoveUpSlaves(instanceKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range movedSlaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case cliCommand("move-below"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce sibling:", destination)
			}
			_, err := inst.MoveBelow(instanceKey, destinationKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case cliCommand("move-equivalent"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce sibling:", destination)
			}
			_, err := inst.MoveEquivalent(instanceKey, destinationKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case cliCommand("move-gtid"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce sibling:", destination)
			}
			_, err := inst.MoveBelowGTID(instanceKey, destinationKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case cliCommand("move-slaves-gtid"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			movedSlaves, _, err, errs := inst.MoveSlavesGTID(instanceKey, destinationKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range movedSlaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case cliCommand("repoint"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			// destinationKey can be null, in which case the instance repoints to its existing master
			instance, err := inst.Repoint(instanceKey, destinationKey, inst.GTIDHintNeutral)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), instance.MasterKey.DisplayString()))
		}
	case cliCommand("repoint-slaves"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			repointedSlaves, err, errs := inst.RepointSlavesTo(instanceKey, pattern, destinationKey)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range repointedSlaves {
					fmt.Println(fmt.Sprintf("%s<%s", slave.Key.DisplayString(), instanceKey.DisplayString()))
				}
			}
		}
	case cliCommand("enslave-siblings"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, _, err := inst.EnslaveSiblings(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("enslave-master"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.EnslaveMaster(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
	case cliCommand("make-co-master"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			_, err := inst.MakeCoMaster(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(instanceKey.DisplayString())
		}
		// Pseudo-GTID
	case cliCommand("match"), cliCommand("match-below"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}
			_, _, err := inst.MatchBelow(instanceKey, destinationKey, true)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), destinationKey.DisplayString()))
		}
	case cliCommand("match-up"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			instance, _, err := inst.MatchUp(instanceKey, true)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), instance.MasterKey.DisplayString()))
		}
	case cliCommand("rematch"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			instance, _, err := inst.RematchSlave(instanceKey, true)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(fmt.Sprintf("%s<%s", instanceKey.DisplayString(), instance.MasterKey.DisplayString()))
		}
	case cliCommand("get-candidate-slave"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			instance, _, _, _, err := inst.GetCandidateSlave(instanceKey, false)
			if err != nil {
				log.Fatale(err)
			} else {
				fmt.Println(instance.Key.DisplayString())
			}
		}
	case cliCommand("match-slaves"), cliCommand("multi-match-slaves"):
		{
			// Move all slaves of "instance" beneath "destination"
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			if destinationKey == nil {
				log.Fatal("Cannot deduce destination:", destination)
			}

			matchedSlaves, _, err, errs := inst.MultiMatchSlaves(instanceKey, destinationKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range matchedSlaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case cliCommand("match-up-slaves"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			matchedSlaves, _, err, errs := inst.MatchUpSlaves(instanceKey, pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, e := range errs {
					log.Errore(e)
				}
				for _, slave := range matchedSlaves {
					fmt.Println(slave.Key.DisplayString())
				}
			}
		}
	case cliCommand("regroup-slaves"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			lostSlaves, equalSlaves, aheadSlaves, promotedSlave, err := inst.RegroupSlaves(instanceKey, false, func(candidateSlave *inst.Instance) { fmt.Println(candidateSlave.Key.DisplayString()) })
			if promotedSlave == nil {
				log.Fatalf("Could not regroup slaves of %+v; error: %+v", *instanceKey, err)
			}
			fmt.Println(fmt.Sprintf("%s lost: %d, trivial: %d, pseudo-gtid: %d",
				promotedSlave.Key.DisplayString(), len(lostSlaves), len(equalSlaves), len(aheadSlaves)))
			if err != nil {
				log.Fatale(err)
			}
		}
	case cliCommand("regroup-slaves-gtid"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			lostSlaves, movedSlaves, promotedSlave, err := inst.RegroupSlavesGTID(instanceKey, false, func(candidateSlave *inst.Instance) { fmt.Println(candidateSlave.Key.DisplayString()) })
			if promotedSlave == nil {
				log.Fatalf("Could not regroup slaves of %+v; error: %+v", *instanceKey, err)
			}
			fmt.Println(fmt.Sprintf("%s lost: %d, moved: %d",
				promotedSlave.Key.DisplayString(), len(lostSlaves), len(movedSlaves)))
			if err != nil {
				log.Fatale(err)
			}
		}
	case cliCommand("regroup-slaves-bls"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			promotedBinlogServer, err := inst.RegroupSlavesBinlogServers(instanceKey, false, nil)
			if promotedBinlogServer == nil {
				log.Fatalf("Could not regroup binlog server slaves of %+v; error: %+v", *instanceKey, err)
			}
			fmt.Println(promotedBinlogServer.Key.DisplayString())
			if err != nil {
				log.Fatale(err)
			}
		}
		// cluster
	case cliCommand("clusters"):
		{
			clusters, err := inst.ReadClusters()
			if err != nil {
				log.Fatale(err)
			} else {
				fmt.Println(strings.Join(clusters, "\n"))
			}
		}
	case cliCommand("find"):
		{
			if pattern == "" {
				log.Fatal("No pattern given")
			}
			instances, err := inst.FindInstances(pattern)
			if err != nil {
				log.Fatale(err)
			} else {
				for _, instance := range instances {
					fmt.Println(instance.Key.DisplayString())
				}
			}
		}
	case cliCommand("topology"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}
			output, err := inst.ASCIITopology(instanceKey, pattern)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(output)
		}
	case cliCommand("which-instance"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatalf("Unable to get master: unresolved instance")
			}
			instance, _, err := inst.ReadInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			fmt.Println(instance.Key.DisplayString())
		}
	case cliCommand("which-master"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatalf("Unable to get master: unresolved instance")
			}
			instance, _, err := inst.ReadInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			fmt.Println(instance.MasterKey.DisplayString())
		}
	case cliCommand("which-cluster"):
		{
			clusterName := getClusterName(clusterAlias, instanceKey)
			fmt.Println(clusterName)
		}
	case cliCommand("which-cluster-instances"):
		{
			clusterName := getClusterName(clusterAlias, instanceKey)
			instances, err := inst.ReadClusterInstances(clusterName)
			if err != nil {
				log.Fatale(err)
			}
			for _, clusterInstance := range instances {
				fmt.Println(clusterInstance.Key.DisplayString())
			}
		}
	case cliCommand("which-cluster-osc-slaves"):
		{
			clusterName := getClusterName(clusterAlias, instanceKey)
			instances, err := inst.GetClusterOSCSlaves(clusterName)
			if err != nil {
				log.Fatale(err)
			}
			for _, clusterInstance := range instances {
				fmt.Println(clusterInstance.Key.DisplayString())
			}
		}
	case cliCommand("which-slaves"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatalf("Unable to get slaves: unresolved instance")
			}
			slaves, err := inst.ReadSlaveInstances(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			for _, slave := range slaves {
				fmt.Println(slave.Key.DisplayString())
			}
		}
	case cliCommand("instance-status"):
		{
			if instanceKey == nil {
				instanceKey = thisInstanceKey
			}
			if instanceKey == nil {
				log.Fatalf("Unable to get status: unresolved instance")
			}
			instance, _, err := inst.ReadInstance(instanceKey)
			if err != nil {
				log.Fatale(err)
			}
			if instance == nil {
				log.Fatalf("Instance not found: %+v", *instanceKey)
			}
			fmt.Println(instance.HumanReadableDescription())
		}
	case cliCommand("get-cluster-heuristic-lag"):
		{
			clusterName := getClusterName(clusterAlias, instanceKey)
			lag, err := inst.GetClusterHeuristicLag(clusterName)
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println(lag)
		}
		// meta
	case cliCommand("snapshot-topologies"):
		{
			err := inst.SnapshotTopologies()
			if err != nil {
				log.Fatale(err)
			}
		}
	case cliCommand("continuous"):
		{
			logic.ContinuousDiscovery()
		}
	case cliCommand("reset-hostname-resolve-cache"):
		{
			err := inst.ResetHostnameResolveCache()
			if err != nil {
				log.Fatale(err)
			}
			fmt.Println("hostname resolve cache cleared")
		}
		// Recovery & analysis
	case cliCommand("recover"), cliCommand("recover-lite"):
		{
			if instanceKey == nil {
				log.Fatal("Cannot deduce instance:", instance)
			}

			actionTaken, promotedInstance, err := logic.CheckAndRecover(instanceKey, destinationKey, true, (command == "recover-lite"))
			if err != nil {
				log.Fatale(err)
			}
			if actionTaken {
				fmt.Println(promotedInstance.Key.DisplayString())
			}
		}
	case cliCommand("replication-analysis"):
		{
			analysis, err := inst.GetReplicationAnalysis(false)
			if err != nil {
				log.Fatale(err)
			}
			for _, entry := range analysis {
				fmt.Println(fmt.Sprintf("%s (cluster %s): %s", entry.AnalyzedInstanceKey.DisplayString(), entry.ClusterDetails.ClusterName, entry.Analysis))
			}
		}
		// pool
	case cliCommand("submit-pool-instances"):
		{
			if pool == "" {
				log.Fatal("Please submit --pool")
			}
			err := inst.ApplyPoolInstances(pool, instance)
			if err != nil {
				log.Fatale(err)
			}
		}
	case cliCommand("cluster-pool-instances"):
		{
			clusterPoolInstances, err := inst.ReadAllClusterPoolInstances()
			if err != nil {
				log.Fatale(err)
			}
			for _, clusterPoolInstance := range clusterPoolInstances {
				fmt.Println(fmt.Sprintf("%s\t%s\t%s\t%s:%d", clusterPoolInstance.ClusterName, clusterPoolInstance.ClusterAlias, clusterPoolInstance.Pool, clusterPoolInstance.Hostname, clusterPoolInstance.Port))
			}
		}
	default:
		log.Fatalf("Unknown command: \"%s\". Available commands (-c):\n\t%v", command, strings.Join(knownCommands, "\n\t"))
	}
}
Пример #4
0
// RecoverDeadCoMaster recovers a dead co-master, complete logic inside
func RecoverDeadCoMaster(topologyRecovery *TopologyRecovery, skipProcesses bool) (promotedSlave *inst.Instance, lostSlaves [](*inst.Instance), err error) {
	analysisEntry := &topologyRecovery.AnalysisEntry
	failedInstanceKey := &analysisEntry.AnalyzedInstanceKey
	otherCoMasterKey := &analysisEntry.AnalyzedInstanceMasterKey
	otherCoMaster, found, _ := inst.ReadInstance(otherCoMasterKey)
	if otherCoMaster == nil || !found {
		return nil, lostSlaves, topologyRecovery.AddError(log.Errorf("RecoverDeadCoMaster: could not read info for co-master %+v of %+v", *otherCoMasterKey, *failedInstanceKey))
	}
	inst.AuditOperation("recover-dead-co-master", failedInstanceKey, "problem found; will recover")
	if !skipProcesses {
		if err := executeProcesses(config.Config.PreFailoverProcesses, "PreFailoverProcesses", topologyRecovery, true); err != nil {
			return nil, lostSlaves, topologyRecovery.AddError(err)
		}
	}

	log.Debugf("topology_recovery: RecoverDeadCoMaster: will recover %+v", *failedInstanceKey)

	var coMasterRecoveryType MasterRecoveryType = MasterRecoveryPseudoGTID
	if analysisEntry.OracleGTIDImmediateTopology || analysisEntry.MariaDBGTIDImmediateTopology {
		coMasterRecoveryType = MasterRecoveryGTID
	}

	log.Debugf("topology_recovery: RecoverDeadCoMaster: coMasterRecoveryType=%+v", coMasterRecoveryType)

	switch coMasterRecoveryType {
	case MasterRecoveryGTID:
		{
			lostSlaves, _, promotedSlave, err = inst.RegroupSlavesGTID(failedInstanceKey, true, nil)
		}
	case MasterRecoveryPseudoGTID:
		{
			lostSlaves, _, _, promotedSlave, err = inst.RegroupSlavesPseudoGTIDIncludingSubSlavesOfBinlogServers(failedInstanceKey, true, nil, &topologyRecovery.PostponedFunctionsContainer)
		}
	}
	topologyRecovery.AddError(err)

	mustPromoteOtherCoMaster := config.Config.CoMasterRecoveryMustPromoteOtherCoMaster
	if !otherCoMaster.ReadOnly {
		log.Debugf("topology_recovery: RecoverDeadCoMaster: other co-master %+v is writeable hence has to be promoted", otherCoMaster.Key)
		mustPromoteOtherCoMaster = true
	}
	log.Debugf("topology_recovery: RecoverDeadCoMaster: mustPromoteOtherCoMaster? %+v", mustPromoteOtherCoMaster)

	if promotedSlave != nil {
		topologyRecovery.ParticipatingInstanceKeys.AddKey(promotedSlave.Key)
		if mustPromoteOtherCoMaster {
			log.Debugf("topology_recovery: mustPromoteOtherCoMaster. Verifying that %+v is/can be promoted", *otherCoMasterKey)
			promotedSlave, err = replacePromotedSlaveWithCandidate(failedInstanceKey, promotedSlave, otherCoMasterKey)
		} else {
			// We are allowed to promote any server
			promotedSlave, err = replacePromotedSlaveWithCandidate(failedInstanceKey, promotedSlave, nil)

			if promotedSlave.DataCenter == otherCoMaster.DataCenter &&
				promotedSlave.PhysicalEnvironment == otherCoMaster.PhysicalEnvironment && false {
				// and _still_ we prefer to promote the co-master! They're in same env & DC so no worries about geo issues!
				promotedSlave, err = replacePromotedSlaveWithCandidate(failedInstanceKey, promotedSlave, otherCoMasterKey)
			}
		}
		topologyRecovery.AddError(err)
	}
	if promotedSlave != nil {
		if mustPromoteOtherCoMaster && !promotedSlave.Key.Equals(otherCoMasterKey) {
			topologyRecovery.AddError(log.Errorf("RecoverDeadCoMaster: could not manage to promote other-co-master %+v; was only able to promote %+v; CoMasterRecoveryMustPromoteOtherCoMaster is true, therefore failing", *otherCoMasterKey, promotedSlave.Key))
			promotedSlave = nil
		}
	}
	if promotedSlave != nil {
		topologyRecovery.ParticipatingInstanceKeys.AddKey(promotedSlave.Key)
	}

	// OK, we may have someone promoted. Either this was the other co-master or another slave.
	// Noting down that we DO NOT attempt to set a new co-master topology. We are good with remaining with a single master.
	// I tried solving the "let's promote a slave and create a new co-master setup" but this turns so complex due to various factors.
	// I see this as risky and not worth the questionable benefit.
	// Maybe future me is a smarter person and finds a simple solution. Unlikely. I'm getting dumber.
	//
	// ...
	// Now that we're convinved, take a look at what we can be left with:
	// Say we started with M1<->M2<-S1, with M2 failing, and we promoted S1.
	// We now have M1->S1 (because S1 is promoted), S1->M2 (because that's what it remembers), M2->M1 (because that's what it remembers)
	// !! This is an evil 3-node circle that must be broken.
	// config.Config.ApplyMySQLPromotionAfterMasterFailover, if true, will cause it to break, because we would RESET SLAVE on S1
	// but we want to make sure the circle is broken no matter what.
	// So in the case we promoted not-the-other-co-master, we issue a detach-slave-master-host, which is a reversible operation
	if promotedSlave != nil && !promotedSlave.Key.Equals(otherCoMasterKey) {
		_, err = inst.DetachSlaveMasterHost(&promotedSlave.Key)
		topologyRecovery.AddError(log.Errore(err))
	}

	if promotedSlave != nil && len(lostSlaves) > 0 && config.Config.DetachLostSlavesAfterMasterFailover {
		postponedFunction := func() error {
			log.Debugf("topology_recovery: - RecoverDeadCoMaster: lost %+v slaves during recovery process; detaching them", len(lostSlaves))
			for _, slave := range lostSlaves {
				slave := slave
				inst.DetachSlaveOperation(&slave.Key)
			}
			return nil
		}
		topologyRecovery.AddPostponedFunction(postponedFunction)
	}
	if config.Config.MasterFailoverLostInstancesDowntimeMinutes > 0 {
		postponedFunction := func() error {
			inst.BeginDowntime(failedInstanceKey, inst.GetMaintenanceOwner(), "RecoverDeadCoMaster indicates this instance is lost", config.Config.MasterFailoverLostInstancesDowntimeMinutes*60)
			for _, slave := range lostSlaves {
				slave := slave
				inst.BeginDowntime(&slave.Key, inst.GetMaintenanceOwner(), "RecoverDeadCoMaster indicates this instance is lost", config.Config.MasterFailoverLostInstancesDowntimeMinutes*60)
			}
			return nil
		}
		topologyRecovery.AddPostponedFunction(postponedFunction)
	}

	return promotedSlave, lostSlaves, err
}
Пример #5
0
// RecoverDeadMaster recovers a dead master, complete logic inside
func RecoverDeadMaster(topologyRecovery *TopologyRecovery, skipProcesses bool) (promotedSlave *inst.Instance, lostSlaves [](*inst.Instance), err error) {
	analysisEntry := &topologyRecovery.AnalysisEntry
	failedInstanceKey := &analysisEntry.AnalyzedInstanceKey

	inst.AuditOperation("recover-dead-master", failedInstanceKey, "problem found; will recover")
	if !skipProcesses {
		if err := executeProcesses(config.Config.PreFailoverProcesses, "PreFailoverProcesses", topologyRecovery, true); err != nil {
			return nil, lostSlaves, topologyRecovery.AddError(err)
		}
	}

	log.Debugf("topology_recovery: RecoverDeadMaster: will recover %+v", *failedInstanceKey)

	var masterRecoveryType MasterRecoveryType = MasterRecoveryPseudoGTID
	if analysisEntry.OracleGTIDImmediateTopology || analysisEntry.MariaDBGTIDImmediateTopology {
		masterRecoveryType = MasterRecoveryGTID
	} else if analysisEntry.BinlogServerImmediateTopology {
		masterRecoveryType = MasterRecoveryBinlogServer
	}
	log.Debugf("topology_recovery: RecoverDeadMaster: masterRecoveryType=%+v", masterRecoveryType)

	switch masterRecoveryType {
	case MasterRecoveryGTID:
		{
			lostSlaves, _, promotedSlave, err = inst.RegroupSlavesGTID(failedInstanceKey, true, nil)
		}
	case MasterRecoveryPseudoGTID:
		{
			lostSlaves, _, _, promotedSlave, err = inst.RegroupSlavesPseudoGTIDIncludingSubSlavesOfBinlogServers(failedInstanceKey, true, nil, &topologyRecovery.PostponedFunctionsContainer)
		}
	case MasterRecoveryBinlogServer:
		{
			promotedSlave, err = recoverDeadMasterInBinlogServerTopology(topologyRecovery)
		}
	}
	topologyRecovery.AddError(err)

	if promotedSlave != nil && len(lostSlaves) > 0 && config.Config.DetachLostSlavesAfterMasterFailover {
		postponedFunction := func() error {
			log.Debugf("topology_recovery: - RecoverDeadMaster: lost %+v slaves during recovery process; detaching them", len(lostSlaves))
			for _, slave := range lostSlaves {
				slave := slave
				inst.DetachSlaveOperation(&slave.Key)
			}
			return nil
		}
		topologyRecovery.AddPostponedFunction(postponedFunction)
	}
	if config.Config.MasterFailoverLostInstancesDowntimeMinutes > 0 {
		postponedFunction := func() error {
			inst.BeginDowntime(failedInstanceKey, inst.GetMaintenanceOwner(), "RecoverDeadMaster indicates this instance is lost", config.Config.MasterFailoverLostInstancesDowntimeMinutes*60)
			for _, slave := range lostSlaves {
				slave := slave
				inst.BeginDowntime(&slave.Key, inst.GetMaintenanceOwner(), "RecoverDeadMaster indicates this instance is lost", config.Config.MasterFailoverLostInstancesDowntimeMinutes*60)
			}
			return nil
		}
		topologyRecovery.AddPostponedFunction(postponedFunction)
	}

	if promotedSlave == nil {
		inst.AuditOperation("recover-dead-master", failedInstanceKey, "Failure: no slave promoted.")
	} else {
		inst.AuditOperation("recover-dead-master", failedInstanceKey, fmt.Sprintf("promoted slave: %+v", promotedSlave.Key))
	}
	return promotedSlave, lostSlaves, err
}
Пример #6
0
// RecoverDeadCoMaster recovers a dead co-master, complete logic inside
func RecoverDeadCoMaster(topologyRecovery *TopologyRecovery, skipProcesses bool) (otherCoMaster *inst.Instance, lostSlaves [](*inst.Instance), err error) {
	analysisEntry := &topologyRecovery.AnalysisEntry
	failedInstanceKey := &analysisEntry.AnalyzedInstanceKey
	otherCoMasterKey := &analysisEntry.AnalyzedInstanceMasterKey

	inst.AuditOperation("recover-dead-co-master", failedInstanceKey, "problem found; will recover")
	if !skipProcesses {
		if err := executeProcesses(config.Config.PreFailoverProcesses, "PreFailoverProcesses", topologyRecovery, true); err != nil {
			return nil, lostSlaves, topologyRecovery.AddError(err)
		}
	}

	log.Debugf("topology_recovery: RecoverDeadCoMaster: will recover %+v", *failedInstanceKey)

	var coMasterRecoveryType MasterRecoveryType = MasterRecoveryPseudoGTID
	if (analysisEntry.OracleGTIDImmediateTopology || analysisEntry.MariaDBGTIDImmediateTopology) && !analysisEntry.PseudoGTIDImmediateTopology {
		coMasterRecoveryType = MasterRecoveryGTID
	}

	log.Debugf("topology_recovery: RecoverDeadCoMaster: coMasterRecoveryType=%+v", coMasterRecoveryType)

	var promotedSlave *inst.Instance
	switch coMasterRecoveryType {
	case MasterRecoveryGTID:
		{
			lostSlaves, _, promotedSlave, err = inst.RegroupSlavesGTID(failedInstanceKey, true, nil)
		}
	case MasterRecoveryPseudoGTID:
		{
			lostSlaves, _, _, promotedSlave, err = inst.RegroupSlavesPseudoGTIDIncludingSubSlavesOfBinlogServers(failedInstanceKey, true, nil, &topologyRecovery.PostponedFunctionsContainer)
		}
	}
	topologyRecovery.AddError(err)

	if promotedSlave != nil {
		topologyRecovery.ParticipatingInstanceKeys.AddKey(promotedSlave.Key)
		promotedSlave, err = replacePromotedSlaveWithCandidate(failedInstanceKey, promotedSlave, otherCoMasterKey)
		topologyRecovery.AddError(err)
	}
	if promotedSlave != nil {
		if promotedSlave.Key.Equals(otherCoMasterKey) {
			topologyRecovery.ParticipatingInstanceKeys.AddKey(*otherCoMasterKey)
			otherCoMaster = promotedSlave
		} else {
			err = log.Errorf("RecoverDeadCoMaster: could not manage to promote other-co-master %+v; was only able to promote %+v", *otherCoMasterKey, promotedSlave.Key)
			promotedSlave = nil
		}
	}

	if promotedSlave != nil && len(lostSlaves) > 0 && config.Config.DetachLostSlavesAfterMasterFailover {
		postponedFunction := func() error {
			log.Debugf("topology_recovery: - RecoverDeadCoMaster: lost %+v slaves during recovery process; detaching them", len(lostSlaves))
			for _, slave := range lostSlaves {
				slave := slave
				inst.DetachSlaveOperation(&slave.Key)
			}
			return nil
		}
		topologyRecovery.AddPostponedFunction(postponedFunction)
	}
	if config.Config.MasterFailoverLostInstancesDowntimeMinutes > 0 {
		postponedFunction := func() error {
			inst.BeginDowntime(failedInstanceKey, inst.GetMaintenanceOwner(), "RecoverDeadCoMaster indicates this instance is lost", config.Config.MasterFailoverLostInstancesDowntimeMinutes*60)
			for _, slave := range lostSlaves {
				slave := slave
				inst.BeginDowntime(&slave.Key, inst.GetMaintenanceOwner(), "RecoverDeadCoMaster indicates this instance is lost", config.Config.MasterFailoverLostInstancesDowntimeMinutes*60)
			}
			return nil
		}
		topologyRecovery.AddPostponedFunction(postponedFunction)
	}

	return otherCoMaster, lostSlaves, err
}