Example #1
0
func (s *TestSuite) TestReadTopologySlave(c *C) {
	key := slave3Key
	i, _ := inst.ReadTopologyInstance(&key)
	c.Assert(i.Key.Hostname, Equals, key.Hostname)
	c.Assert(i.IsSlave(), Equals, true)
	c.Assert(len(i.SlaveHosts), Equals, 0)
}
Example #2
0
func (s *TestSuite) TestReadTopologyAndInstanceSlave(c *C) {
	i, _ := inst.ReadTopologyInstance(&slave1Key)
	iRead, found, _ := inst.ReadInstance(&slave1Key)
	c.Assert(found, Equals, true)
	c.Assert(iRead.Key.Hostname, Equals, i.Key.Hostname)
	c.Assert(iRead.Version, Equals, i.Version)
}
Example #3
0
func (s *TestSuite) TestBeginMaintenance(c *C) {
	clearTestMaintenance()
	_, _ = inst.ReadTopologyInstance(&masterKey)
	_, err := inst.BeginMaintenance(&masterKey, "unittest", "TestBeginMaintenance")

	c.Assert(err, IsNil)
}
Example #4
0
func (s *TestSuite) TestMakeCoMasterAndBack(c *C) {
	clearTestMaintenance()

	slave1, err := inst.MakeCoMaster(&slave1Key)
	c.Assert(err, IsNil)

	// Now master & slave1 expected to be co-masters. Check!
	master, _ := inst.ReadTopologyInstance(&masterKey)
	c.Assert(master.IsSlaveOf(slave1), Equals, true)
	c.Assert(slave1.IsSlaveOf(master), Equals, true)

	// reset - restore to original state
	master, err = inst.ResetSlaveOperation(&masterKey)
	slave1, _ = inst.ReadTopologyInstance(&slave1Key)
	c.Assert(err, IsNil)
	c.Assert(master.MasterKey.Hostname, Equals, "_")
}
Example #5
0
func (s *TestSuite) TestForgetMaster(c *C) {
	_, _ = inst.ReadTopologyInstance(&masterKey)
	_, found, _ := inst.ReadInstance(&masterKey)
	c.Assert(found, Equals, true)
	inst.ForgetInstance(&masterKey)
	_, found, _ = inst.ReadInstance(&masterKey)
	c.Assert(found, Equals, false)
}
Example #6
0
func (s *TestSuite) TestReadTopologyAndInstanceMaster(c *C) {
	i, _ := inst.ReadTopologyInstance(&masterKey)
	iRead, found, _ := inst.ReadInstance(&masterKey)
	c.Assert(found, Equals, true)
	c.Assert(iRead.Key.Hostname, Equals, i.Key.Hostname)
	c.Assert(iRead.Version, Equals, i.Version)
	c.Assert(len(iRead.SlaveHosts), Equals, len(i.SlaveHosts))
}
Example #7
0
func (s *TestSuite) TestGetMasterOfASlave(c *C) {
	i, err := inst.ReadTopologyInstance(&slave1Key)
	c.Assert(err, IsNil)
	master, err := inst.GetInstanceMaster(i)
	c.Assert(err, IsNil)
	c.Assert(master.IsSlave(), Equals, false)
	c.Assert(master.Key.Port, Equals, 22987)
}
Example #8
0
func (s *TestSuite) TestReadTopologyUnexisting(c *C) {
	key := inst.InstanceKey{
		Hostname: "127.0.0.1",
		Port:     22999,
	}
	_, err := inst.ReadTopologyInstance(&key)

	c.Assert(err, Not(IsNil))
}
// Force a re-read of a topology instance; this is done because we need to substantiate a suspicion that we may have a failover
// scenario. we want to speed up rading the complete picture.
func emergentlyReadTopologyInstance(instanceKey *inst.InstanceKey, analysisCode inst.AnalysisCode) {
	if err := emergencyReadTopologyInstanceMap.Add(instanceKey.DisplayString(), true, 0); err == nil {
		emergencyReadTopologyInstanceMap.Set(instanceKey.DisplayString(), true, 0)
		go inst.ExecuteOnTopology(func() {
			inst.ReadTopologyInstance(instanceKey)
			inst.AuditOperation("emergently-read-topology-instance", instanceKey, string(analysisCode))
		})
	}
}
Example #10
0
func (s *TestSuite) TestReadTopologyMaster(c *C) {
	key := masterKey
	i, _ := inst.ReadTopologyInstance(&key)

	c.Assert(i.Key.Hostname, Equals, key.Hostname)
	c.Assert(i.IsSlave(), Equals, false)
	c.Assert(len(i.SlaveHosts), Equals, 3)
	c.Assert(len(i.SlaveHosts.GetInstanceKeys()), Equals, len(i.SlaveHosts))
}
Example #11
0
func (s *TestSuite) TestFailEndMaintenanceTwice(c *C) {
	clearTestMaintenance()
	_, _ = inst.ReadTopologyInstance(&masterKey)
	k, err := inst.BeginMaintenance(&masterKey, "unittest", "TestFailEndMaintenanceTwice")
	c.Assert(err, IsNil)
	err = inst.EndMaintenance(k)
	c.Assert(err, IsNil)
	err = inst.EndMaintenance(k)
	c.Assert(err, Not(IsNil))
}
Example #12
0
// Force a re-read of a topology instance; this is done because we need to substantiate a suspicion that we may have a failover
// scenario. we want to speed up reading the complete picture.
func emergentlyReadTopologyInstance(instanceKey *inst.InstanceKey, analysisCode inst.AnalysisCode) {
	if existsInCacheError := emergencyReadTopologyInstanceMap.Add(instanceKey.DisplayString(), true, cache.DefaultExpiration); existsInCacheError != nil {
		// Just recently attempted
		return
	}
	go inst.ExecuteOnTopology(func() {
		inst.ReadTopologyInstance(instanceKey)
		inst.AuditOperation("emergently-read-topology-instance", instanceKey, string(analysisCode))
	})
}
Example #13
0
func (s *TestSuite) TestFailMoveBelowUponMaintenance(c *C) {
	clearTestMaintenance()
	_, _ = inst.ReadTopologyInstance(&slave1Key)
	k, err := inst.BeginMaintenance(&slave1Key, "unittest", "TestBeginEndMaintenance")
	c.Assert(err, IsNil)

	_, err = inst.MoveBelow(&slave1Key, &slave2Key)
	c.Assert(err, Not(IsNil))

	err = inst.EndMaintenance(k)
	c.Assert(err, IsNil)
}
Example #14
0
func (s *TestSuite) TestStopStartSlave(c *C) {

	i, _ := inst.ReadTopologyInstance(&slave1Key)
	c.Assert(i.SlaveRunning(), Equals, true)
	i, _ = inst.StopSlaveNicely(&i.Key, 0)

	c.Assert(i.SlaveRunning(), Equals, false)
	c.Assert(i.SQLThreadUpToDate(), Equals, true)

	i, _ = inst.StartSlave(&i.Key)
	c.Assert(i.SlaveRunning(), Equals, true)
}
Example #15
0
func (s *TestSuite) TestFailMoveBelowUponOtherSlaveStopped(c *C) {
	clearTestMaintenance()

	slave1, _ := inst.ReadTopologyInstance(&slave1Key)
	c.Assert(slave1.SlaveRunning(), Equals, true)
	slave1, _ = inst.StopSlaveNicely(&slave1.Key, 0)
	c.Assert(slave1.SlaveRunning(), Equals, false)

	_, err := inst.MoveBelow(&slave2Key, &slave1Key)
	c.Assert(err, Not(IsNil))

	_, _ = inst.StartSlave(&slave1.Key)
}
Example #16
0
func (s *TestSuite) TestMoveBelowAndBackComplex(c *C) {
	clearTestMaintenance()

	// become child
	slave1, _ := inst.MoveBelow(&slave1Key, &slave2Key)

	c.Assert(slave1.MasterKey.Equals(&slave2Key), Equals, true)
	c.Assert(slave1.SlaveRunning(), Equals, true)

	// Now let's have fun. Stop slave2 (which is now parent of slave1), execute queries on master,
	// move s1 back under master, start all, verify queries.

	_, err := inst.StopSlave(&slave2Key)
	c.Assert(err, IsNil)

	randValue := rand.Int()
	_, err = inst.ExecInstance(&masterKey, `replace into orchestrator_test.test_table (name, value) values ('TestMoveBelowAndBackComplex', ?)`, randValue)
	c.Assert(err, IsNil)
	master, err := inst.ReadTopologyInstance(&masterKey)
	c.Assert(err, IsNil)

	// And back; keep topology intact
	slave1, err = inst.MoveUp(&slave1Key)
	c.Assert(err, IsNil)
	_, err = inst.MasterPosWait(&slave1Key, &master.SelfBinlogCoordinates)
	c.Assert(err, IsNil)
	slave2, err := inst.ReadTopologyInstance(&slave2Key)
	c.Assert(err, IsNil)
	_, err = inst.MasterPosWait(&slave2Key, &master.SelfBinlogCoordinates)
	c.Assert(err, IsNil)
	// Now check for value!
	var value1, value2 int
	inst.ScanInstanceRow(&slave1Key, `select value from orchestrator_test.test_table where name='TestMoveBelowAndBackComplex'`, &value1)
	inst.ScanInstanceRow(&slave2Key, `select value from orchestrator_test.test_table where name='TestMoveBelowAndBackComplex'`, &value2)

	c.Assert(inst.InstancesAreSiblings(slave1, slave2), Equals, true)
	c.Assert(value1, Equals, randValue)
	c.Assert(value2, Equals, randValue)
}
Example #17
0
// discoverInstance will attempt discovering an instance (unless it is already up to date) and will
// list down its master and slaves (if any) for further discovery.
func discoverInstance(instanceKey inst.InstanceKey) {
	start := time.Now()

	instanceKey.Formalize()
	if !instanceKey.IsValid() {
		return
	}

	if existsInCacheError := recentDiscoveryOperationKeys.Add(instanceKey.DisplayString(), true, cache.DefaultExpiration); existsInCacheError != nil {
		// Just recently attempted
		return
	}

	instance, found, err := inst.ReadInstance(&instanceKey)
	if found && instance.IsUpToDate && instance.IsLastCheckValid {
		// we've already discovered this one. Skip!
		return
	}

	discoveriesCounter.Inc(1)

	// First we've ever heard of this instance. Continue investigation:
	instance, err = inst.ReadTopologyInstance(&instanceKey)
	// panic can occur (IO stuff). Therefore it may happen
	// that instance is nil. Check it.
	if instance == nil {
		failedDiscoveriesCounter.Inc(1)
		log.Warningf("discoverInstance(%+v) instance is nil in %.3fs, error=%+v", instanceKey, time.Since(start).Seconds(), err)
		return
	}

	log.Debugf("Discovered host: %+v, master: %+v, version: %+v in %.3fs", instance.Key, instance.MasterKey, instance.Version, time.Since(start).Seconds())

	if atomic.LoadInt64(&isElectedNode) == 0 {
		// Maybe this node was elected before, but isn't elected anymore.
		// If not elected, stop drilling up/down the topology
		return
	}

	// Investigate slaves:
	for _, slaveKey := range instance.SlaveHosts.GetInstanceKeys() {
		slaveKey := slaveKey
		if slaveKey.IsValid() {
			discoveryQueue.Push(slaveKey)
		}
	}
	// Investigate master:
	if instance.MasterKey.IsValid() {
		discoveryQueue.Push(instance.MasterKey)
	}
}
Example #18
0
func (s *TestSuite) TestDiscover(c *C) {
	var err error
	_, err = db.ExecOrchestrator("delete from database_instance where hostname = ? and port = ?", masterKey.Hostname, masterKey.Port)
	_, err = db.ExecOrchestrator("delete from database_instance where hostname = ? and port = ?", slave1Key.Hostname, slave1Key.Port)
	_, err = db.ExecOrchestrator("delete from database_instance where hostname = ? and port = ?", slave2Key.Hostname, slave2Key.Port)
	_, err = db.ExecOrchestrator("delete from database_instance where hostname = ? and port = ?", slave3Key.Hostname, slave3Key.Port)
	_, found, _ := inst.ReadInstance(&masterKey)
	c.Assert(found, Equals, false)
	_, _ = inst.ReadTopologyInstance(&slave1Key)
	logic.StartDiscovery(slave1Key)
	_, found, err = inst.ReadInstance(&slave1Key)
	c.Assert(found, Equals, true)
	c.Assert(err, IsNil)
}
Example #19
0
// If a mysql port is available, try to discover against it
func DiscoverAgentInstance(hostname string, port int) error {
	agent, err := GetAgent(hostname)
	if err != nil {
		log.Errorf("Couldn't get agent for %s: %v", hostname, err)
		return err
	}

	instanceKey := agent.GetInstance()
	instance, err := inst.ReadTopologyInstance(instanceKey)
	if err != nil {
		log.Errorf("Failed to read topology for %v", instanceKey)
		return err
	}
	log.Infof("Discovered Agent Instance: %v", instance.Key)
	return nil
}
Example #20
0
func (s *TestSuite) TestMoveBelowAndBack(c *C) {
	clearTestMaintenance()
	// become child
	slave1, err := inst.MoveBelow(&slave1Key, &slave2Key)
	c.Assert(err, IsNil)

	c.Assert(slave1.MasterKey.Equals(&slave2Key), Equals, true)
	c.Assert(slave1.SlaveRunning(), Equals, true)

	// And back; keep topology intact
	slave1, _ = inst.MoveUp(&slave1Key)
	slave2, _ := inst.ReadTopologyInstance(&slave2Key)

	c.Assert(inst.InstancesAreSiblings(slave1, slave2), Equals, true)
	c.Assert(slave1.SlaveRunning(), Equals, true)

}
Example #21
0
// discoverInstance will attempt discovering an instance (unless it is already up to date) and will
// list down its master and slaves (if any) for further discovery.
func discoverInstance(instanceKey inst.InstanceKey) {
	instanceKey.Formalize()
	if !instanceKey.IsValid() {
		return
	}

	if existsInCacheError := recentDiscoveryOperationKeys.Add(instanceKey.DisplayString(), true, cache.DefaultExpiration); existsInCacheError != nil {
		// Just recently attempted
		return
	}

	instance, found, err := inst.ReadInstance(&instanceKey)

	if found && instance.IsUpToDate && instance.IsLastCheckValid {
		// we've already discovered this one. Skip!
		return
	}
	discoveriesCounter.Inc(1)
	// First we've ever heard of this instance. Continue investigation:
	instance, err = inst.ReadTopologyInstance(&instanceKey)
	// panic can occur (IO stuff). Therefore it may happen
	// that instance is nil. Check it.
	if instance == nil {
		failedDiscoveriesCounter.Inc(1)
		log.Warningf("instance is nil in discoverInstance. key=%+v, error=%+v", instanceKey, err)
		return
	}

	log.Debugf("Discovered host: %+v, master: %+v", instance.Key, instance.MasterKey)

	if !isElectedNode {
		// Maybe this node was elected before, but isn't elected anymore.
		// If not elected, stop drilling down to further investigate slaves.
		return
	}

	// Investigate slaves:
	for _, slaveKey := range instance.SlaveHosts.GetInstanceKeys() {
		discoveryInstanceKeys <- slaveKey
	}
	// Investigate master:
	discoveryInstanceKeys <- instance.MasterKey
}
Example #22
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"))
	}
}
Example #23
0
func discoverSeededAgents() {
	for seededAgent := range agent.SeededAgents {
		instanceKey := &inst.InstanceKey{Hostname: seededAgent.Hostname, Port: int(seededAgent.MySQLPort)}
		go inst.ReadTopologyInstance(instanceKey)
	}
}
Example #24
0
func (s *TestSuite) TestInstanceIsMasterOf(c *C) {
	i0, _ := inst.ReadTopologyInstance(&masterKey)
	i1, _ := inst.ReadTopologyInstance(&slave1Key)
	c.Assert(inst.InstanceIsMasterOf(i0, i1), Equals, true)
}
Example #25
0
func (s *TestSuite) TestNonSiblings(c *C) {
	i0, _ := inst.ReadTopologyInstance(&masterKey)
	i1, _ := inst.ReadTopologyInstance(&slave1Key)
	c.Assert(inst.InstancesAreSiblings(i0, i1), Not(Equals), true)
}
Example #26
0
// DiscoverInstance will attempt discovering an instance (unless it is already up to date) and will
// list down its master and slaves (if any) for further discovery.
func DiscoverInstance(instanceKey inst.InstanceKey) {
	instanceKey.Formalize()
	if !instanceKey.IsValid() {
		return
	}

	instance, found, err := inst.ReadInstance(&instanceKey)

	if found && instance.IsUpToDate && instance.IsLastCheckValid {
		// we've already discovered this one. Skip!
		goto Cleanup
	}
	// First we've ever heard of this instance. Continue investigation:
	instance, err = inst.ReadTopologyInstance(&instanceKey)
	// panic can occur (IO stuff). Therefore it may happen
	// that instance is nil. Check it.
	if err != nil || instance == nil {
		log.Warningf("instance is nil in DiscoverInstance. key=%+v, error=%+v", instanceKey, err)
		goto Cleanup
	}

	log.Debugf("Discovered host: %+v, master: %+v", instance.Key, instance.MasterKey)

	// Investigate slaves:
	for _, slaveKey := range instance.SlaveHosts.GetInstanceKeys() {
		discoveryInstanceKeys <- slaveKey
	}
	// Investigate master:
	discoveryInstanceKeys <- instance.MasterKey

Cleanup:
}

// Start discovery begins a one time asynchronuous discovery process for the given
// instance and all of its topology connected instances.
// That is, the instance will be investigated for master and slaves, and the routines will follow on
// each and every such found master/slave.
// In essense, assuming all slaves in a replication topology are running, and given a single instance
// in such topology, this function will detect the entire topology.
func StartDiscovery(instanceKey inst.InstanceKey) {
	log.Infof("Starting discovery at %+v", instanceKey)
	pendingTokens := make(chan bool, maxConcurrency)
	completedTokens := make(chan bool, maxConcurrency)

	AccountedDiscoverInstance(instanceKey, pendingTokens, completedTokens)
	go handleDiscoveryRequests(pendingTokens, completedTokens)

	// Block until all are complete
	for {
		select {
		case <-pendingTokens:
			<-completedTokens
		default:
			inst.AuditOperation("start-discovery", &instanceKey, "")
			return
		}
	}
}

// ContinuousDiscovery starts an asynchronuous infinite discovery process where instances are
// periodically investigated and their status captured, and long since unseen instances are
// purged and forgotten.
func ContinuousDiscovery() {
	log.Infof("Starting continuous discovery")
	inst.LoadHostnameResolveCacheFromDatabase()
	go handleDiscoveryRequests(nil, nil)
	tick := time.Tick(time.Duration(config.Config.DiscoveryPollSeconds) * time.Second)
	forgetUnseenTick := time.Tick(time.Minute)
	recoverTick := time.Tick(10 * time.Second)

	var snapshotTopologiesTick <-chan time.Time
	if config.Config.SnapshotTopologiesIntervalHours > 0 {
		snapshotTopologiesTick = time.Tick(time.Duration(config.Config.SnapshotTopologiesIntervalHours) * time.Hour)
	}

	elected := false
	_ = CreateElectionAnchor(false)
	for {
		select {
		case <-tick:
			if elected, _ = AttemptElection(); elected {
				instanceKeys, _ := inst.ReadOutdatedInstanceKeys()
				log.Debugf("outdated keys: %+v", instanceKeys)
				for _, instanceKey := range instanceKeys {
					discoveryInstanceKeys <- instanceKey
				}
			} else {
				log.Debugf("Not elected as active node; polling")
			}
		case <-forgetUnseenTick:
			// See if we should also forget objects (lower frequency)
			go func() {
				if elected {
					inst.ForgetLongUnseenInstances()
					inst.ForgetUnseenInstancesDifferentlyResolved()
					inst.ForgetExpiredHostnameResolves()
					inst.DeleteInvalidHostnameResolves()
					inst.ReviewUnseenInstances()
					inst.InjectUnseenMasters()
					inst.ResolveUnknownMasterHostnameResolves()
					inst.ExpireMaintenance()
					inst.ExpireDowntime()
					inst.ExpireCandidateInstances()
					inst.ExpireHostnameUnresolve()
					inst.ExpireClusterDomainName()
				}
				if !elected {
					// Take this opportunity to refresh yourself
					inst.LoadHostnameResolveCacheFromDatabase()
				}
				inst.ReadClusterAliases()
				HealthTest()
			}()
		case <-recoverTick:
			go func() {
				if elected {
					ClearActiveFailureDetections()
					ClearActiveRecoveries()
					CheckAndRecover(nil, nil, false, false)
				}
			}()
		case <-snapshotTopologiesTick:
			go func() {
				inst.SnapshotTopologies()
			}()
		}
	}
}

func pollAgent(hostname string) error {
	polledAgent, err := agent.GetAgent(hostname)
	agent.UpdateAgentLastChecked(hostname)

	if err != nil {
		return log.Errore(err)
	}

	err = agent.UpdateAgentInfo(hostname, polledAgent)
	if err != nil {
		return log.Errore(err)
	}

	return nil
}

// ContinuousAgentsPoll starts an asynchronuous infinite process where agents are
// periodically investigated and their status captured, and long since unseen agents are
// purged and forgotten.
func ContinuousAgentsPoll() {
	log.Infof("Starting continuous agents poll")

	go discoverSeededAgents()

	tick := time.Tick(time.Duration(config.Config.DiscoveryPollSeconds) * time.Second)
	forgetUnseenTick := time.Tick(time.Hour)
	for _ = range tick {
		agentsHosts, _ := agent.ReadOutdatedAgentsHosts()
		log.Debugf("outdated agents hosts: %+v", agentsHosts)
		for _, hostname := range agentsHosts {
			go pollAgent(hostname)
		}
		// See if we should also forget agents (lower frequency)
		select {
		case <-forgetUnseenTick:
			agent.ForgetLongUnseenAgents()
			agent.FailStaleSeeds()
		default:
		}
	}
}

func discoverSeededAgents() {
	for seededAgent := range agent.SeededAgents {
		instanceKey := inst.InstanceKey{Hostname: seededAgent.Hostname, Port: int(seededAgent.MySQLPort)}
		go StartDiscovery(instanceKey)
	}
}
Example #27
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())
	}
}
Example #28
0
// discoverInstance will attempt discovering an instance (unless it is already up to date) and will
// list down its master and slaves (if any) for further discovery.
func discoverInstance(instanceKey inst.InstanceKey) {
	instanceKey.Formalize()
	if !instanceKey.IsValid() {
		return
	}

	instance, found, err := inst.ReadInstance(&instanceKey)

	if found && instance.IsUpToDate && instance.IsLastCheckValid {
		// we've already discovered this one. Skip!
		goto Cleanup
	}
	discoveriesCounter.Inc(1)
	// First we've ever heard of this instance. Continue investigation:
	instance, err = inst.ReadTopologyInstance(&instanceKey)
	// panic can occur (IO stuff). Therefore it may happen
	// that instance is nil. Check it.
	if err != nil || instance == nil {
		failedDiscoveriesCounter.Inc(1)
		log.Warningf("instance is nil in discoverInstance. key=%+v, error=%+v", instanceKey, err)
		goto Cleanup
	}

	log.Debugf("Discovered host: %+v, master: %+v", instance.Key, instance.MasterKey)

	if !isElectedNode {
		// Maybe this node was elected before, but isn't elected anymore.
		// If not elected, stop drilling down to further investigate slaves.
		return
	}

	// Investigate slaves:
	for _, slaveKey := range instance.SlaveHosts.GetInstanceKeys() {
		discoveryInstanceKeys <- slaveKey
	}
	// Investigate master:
	discoveryInstanceKeys <- instance.MasterKey

Cleanup:
}

// Start discovery begins a one time asynchronuous discovery process for the given
// instance and all of its topology connected instances.
// That is, the instance will be investigated for master and slaves, and the routines will follow on
// each and every such found master/slave.
// In essense, assuming all slaves in a replication topology are running, and given a single instance
// in such topology, this function will detect the entire topology.
func StartDiscovery(instanceKey inst.InstanceKey) {
	log.Infof("Starting discovery at %+v", instanceKey)
	pendingTokens := make(chan bool, maxConcurrency)
	completedTokens := make(chan bool, maxConcurrency)

	accountedDiscoverInstance(instanceKey, pendingTokens, completedTokens)
	go handleDiscoveryRequests(pendingTokens, completedTokens)

	// Block until all are complete
	for {
		select {
		case <-pendingTokens:
			<-completedTokens
		default:
			inst.AuditOperation("start-discovery", &instanceKey, "")
			return
		}
	}
}

func initGraphiteMetrics() error {
	if config.Config.GraphiteAddr == "" {
		return nil
	}
	if config.Config.GraphitePath == "" {
		return log.Errorf("No graphite path provided (see GraphitePath config variable). Will not log to graphite")
	}
	addr, err := net.ResolveTCPAddr("tcp", config.Config.GraphiteAddr)
	if err != nil {
		return log.Errore(err)
	}
	graphitePathHostname := ThisHostname
	if config.Config.GraphiteConvertHostnameDotsToUnderscores {
		graphitePathHostname = strings.Replace(graphitePathHostname, ".", "_", -1)
	}
	graphitePath := config.Config.GraphitePath
	graphitePath = strings.Replace(graphitePath, "{hostname}", graphitePathHostname, -1)

	log.Debugf("Will log to graphite on %+v, %+v", config.Config.GraphiteAddr, graphitePath)
	go graphite.Graphite(metrics.DefaultRegistry, 1*time.Minute, graphitePath, addr)

	return nil

}

// ContinuousDiscovery starts an asynchronuous infinite discovery process where instances are
// periodically investigated and their status captured, and long since unseen instances are
// purged and forgotten.
func ContinuousDiscovery() {
	log.Infof("Starting continuous discovery")
	inst.LoadHostnameResolveCacheFromDatabase()
	go handleDiscoveryRequests(nil, nil)
	tick := time.Tick(time.Duration(config.Config.DiscoveryPollSeconds) * time.Second)
	forgetUnseenTick := time.Tick(time.Minute)
	recoverTick := time.Tick(10 * time.Second)

	var snapshotTopologiesTick <-chan time.Time
	if config.Config.SnapshotTopologiesIntervalHours > 0 {
		snapshotTopologiesTick = time.Tick(time.Duration(config.Config.SnapshotTopologiesIntervalHours) * time.Hour)
	}

	go initGraphiteMetrics()

	for {
		select {
		case <-tick:
			go func() {
				if isElectedNode, _ = attemptElection(); isElectedNode {
					instanceKeys, _ := inst.ReadOutdatedInstanceKeys()
					log.Debugf("outdated keys: %+v", instanceKeys)
					for _, instanceKey := range instanceKeys {
						discoveryInstanceKeys <- instanceKey
					}
				} else {
					log.Debugf("Not elected as active node; polling")
				}
				discoveryQueueLengthGauge.Update(int64(len(discoveryInstanceKeys)))
			}()
		case <-forgetUnseenTick:
			// See if we should also forget objects (lower frequency)
			go func() {
				if isElectedNode {
					inst.ForgetLongUnseenInstances()
					inst.ForgetUnseenInstancesDifferentlyResolved()
					inst.ForgetExpiredHostnameResolves()
					inst.DeleteInvalidHostnameResolves()
					inst.ReviewUnseenInstances()
					inst.InjectUnseenMasters()
					inst.ResolveUnknownMasterHostnameResolves()
					inst.ExpireMaintenance()
					inst.ExpireDowntime()
					inst.ExpireCandidateInstances()
					inst.ExpireHostnameUnresolve()
					inst.ExpireClusterDomainName()
					inst.ExpireAudit()
					inst.ExpireMasterPositionEquivalence()
				}
				if !isElectedNode {
					// Take this opportunity to refresh yourself
					inst.LoadHostnameResolveCacheFromDatabase()
				}
				inst.ReadClusterAliases()
				HealthTest()
			}()
		case <-recoverTick:
			go func() {
				if isElectedNode {
					ClearActiveFailureDetections()
					ClearActiveRecoveries()
					CheckAndRecover(nil, nil, false, false)
				}
			}()
		case <-snapshotTopologiesTick:
			go func() {
				inst.SnapshotTopologies()
			}()
		}
	}
}

func pollAgent(hostname string) error {
	polledAgent, err := agent.GetAgent(hostname)
	agent.UpdateAgentLastChecked(hostname)

	if err != nil {
		return log.Errore(err)
	}

	err = agent.UpdateAgentInfo(hostname, polledAgent)
	if err != nil {
		return log.Errore(err)
	}

	return nil
}

// ContinuousAgentsPoll starts an asynchronuous infinite process where agents are
// periodically investigated and their status captured, and long since unseen agents are
// purged and forgotten.
func ContinuousAgentsPoll() {
	log.Infof("Starting continuous agents poll")

	go discoverSeededAgents()

	tick := time.Tick(time.Duration(config.Config.DiscoveryPollSeconds) * time.Second)
	forgetUnseenTick := time.Tick(time.Hour)
	for range tick {
		agentsHosts, _ := agent.ReadOutdatedAgentsHosts()
		log.Debugf("outdated agents hosts: %+v", agentsHosts)
		for _, hostname := range agentsHosts {
			go pollAgent(hostname)
		}
		// See if we should also forget agents (lower frequency)
		select {
		case <-forgetUnseenTick:
			agent.ForgetLongUnseenAgents()
			agent.FailStaleSeeds()
		default:
		}
	}
}

func discoverSeededAgents() {
	for seededAgent := range agent.SeededAgents {
		instanceKey := inst.InstanceKey{Hostname: seededAgent.Hostname, Port: int(seededAgent.MySQLPort)}
		go StartDiscovery(instanceKey)
	}
}
Example #29
0
func (s *TestSuite) TestSlavesAreSiblings(c *C) {
	i0, _ := inst.ReadTopologyInstance(&slave1Key)
	i1, _ := inst.ReadTopologyInstance(&slave2Key)
	c.Assert(inst.InstancesAreSiblings(i0, i1), Equals, true)
}