func getClusterName(clusterAlias string, instanceKey *inst.InstanceKey) (clusterName string) { var err error if clusterAlias != "" { clusterName, err = inst.ReadClusterByAlias(clusterAlias) if err != nil { log.Fatale(err) } } else { // deduce cluster by instance if instanceKey == nil { instanceKey = thisInstanceKey } if instanceKey == nil { log.Fatalf("Unable to get cluster instances: unresolved instance") } instance, _, err := inst.ReadInstance(instanceKey) if err != nil { log.Fatale(err) } if instance == nil { log.Fatalf("Instance not found: %+v", *instanceKey) } clusterName = instance.ClusterName } if clusterName == "" { log.Fatalf("Unable to determine cluster name") } return clusterName }
// deployIfNotAlreadyDeployed will issue given sql queries that are not already known to be deployed. // This iterates both lists (to-run and already-deployed) and also verifies no contraditions. func deployIfNotAlreadyDeployed(db *sql.DB, queries []string, deployedQueries []string, deploymentType string, fatalOnError bool) error { tx, err := db.Begin() if err != nil { log.Fatale(err) } // Ugly workaround ahead. // Origin of this workaround is the existence of some "timestamp NOT NULL," column definitions, // where in NO_ZERO_IN_DATE,NO_ZERO_DATE sql_mode are invalid (since default is implicitly "0") // This means installation of orchestrator fails on such configured servers, and in particular on 5.7 // where this setting is the dfault. // For purpose of backwards compatability, what we do is force sql_mode to be more relaxed, create the schemas // along with the "invalid" definition, and then go ahead and fix those definitions via following ALTER statements. // My bad. originalSqlMode := "" err = tx.QueryRow(`select @@session.sql_mode`).Scan(&originalSqlMode) if _, err := tx.Exec(`set @@session.sql_mode=REPLACE(@@session.sql_mode, 'NO_ZERO_DATE', '')`); err != nil { log.Fatale(err) } if _, err := tx.Exec(`set @@session.sql_mode=REPLACE(@@session.sql_mode, 'NO_ZERO_IN_DATE', '')`); err != nil { log.Fatale(err) } for i, query := range queries { queryAlreadyExecuted := false // While iterating 'queries', also iterate 'deployedQueries'. Expect identity if len(deployedQueries) > i { if deployedQueries[i] != query { log.Fatalf("initOrchestratorDB() PANIC: non matching %s queries between deployment requests and _orchestrator_db_deployment. Please execute 'orchestrator -c reset-internal-db-deployment'", deploymentType) } queryAlreadyExecuted = true } if queryAlreadyExecuted { continue } if i == 0 { log.Debugf("sql_mode is: %+v", originalSqlMode) } if config.Config.SmartOrchestratorDatabaseUpdate { log.Debugf("initOrchestratorDB executing: %.80s", strings.TrimSpace(strings.Replace(query, "\n", "", -1))) } if fatalOnError { if _, err := tx.Exec(query); err != nil { return log.Fatalf("Cannot initiate orchestrator: %+v", err) } } else { tx.Exec(query) // And ignore any error } writeInternalDeployment(db, deploymentType, query, i) } if _, err := tx.Exec(`set session sql_mode=?`, originalSqlMode); err != nil { log.Fatale(err) } if err := tx.Commit(); err != nil { log.Fatale(err) } return nil }
// ResetInternalDeployment will clear existing deployment history (and the effect will be a complete re-deployment // the next run). This is made available for possible operational errors, a red button func ResetInternalDeployment() error { db, err := OpenOrchestrator() if err != nil { log.Fatalf("Cannot reset orchestrator internal deployment: %+v", err) } if _, err := execInternal(db, `delete from _orchestrator_db_deployment`); err != nil { log.Fatalf("Unable to clear _orchestrator_db_deployment: %+v", err) } return nil }
func init() { var err error ThisHostname, err = os.Hostname() if err != nil { log.Fatalf("Cannot resolve self hostname; required. Aborting. %+v", err) } }
// Create a TLS configuration from the config supplied CA, Certificate, and Private key. // Register the TLS config with the mysql drivers as the "orchestrator" config // Modify the supplied URI to call the TLS config func SetupMySQLOrchestratorTLS(uri string) (string, error) { if !orchestratorTLSConfigured { tlsConfig, err := ssl.NewTLSConfig(config.Config.MySQLOrchestratorSSLCAFile, true) // Drop to TLS 1.0 for talking to MySQL tlsConfig.MinVersion = tls.VersionTLS10 if err != nil { return "", log.Fatalf("Can't create TLS configuration for Orchestrator connection %s: %s", uri, err) } tlsConfig.InsecureSkipVerify = config.Config.MySQLOrchestratorSSLSkipVerify if err = ssl.AppendKeyPair(tlsConfig, config.Config.MySQLOrchestratorSSLCertFile, config.Config.MySQLOrchestratorSSLPrivateKeyFile); err != nil { return "", log.Fatalf("Can't setup TLS key pairs for %s: %s", uri, err) } if err = mysql.RegisterTLSConfig("orchestrator", tlsConfig); err != nil { return "", log.Fatalf("Can't register mysql TLS config for orchestrator: %s", err) } orchestratorTLSConfigured = true } return fmt.Sprintf("%s&tls=orchestrator", uri), nil }
// readInternalDeployments reads orchestrator db deployment statements that are known to have been executed func readInternalDeployments() (baseDeployments []string, patchDeployments []string, err error) { if !config.Config.SmartOrchestratorDatabaseUpdate { return baseDeployments, patchDeployments, nil } query := fmt.Sprintf(` select deployment_type, sql_statement from _orchestrator_db_deployment order by deployment_id `) db, err := OpenOrchestrator() if err != nil { log.Fatalf("Cannot initiate orchestrator internal deployment: %+v", err) } err = sqlutils.QueryRowsMap(db, query, func(m sqlutils.RowMap) error { deploymentType := m.GetString("deployment_type") sqlStatement := m.GetString("sql_statement") if deploymentType == "base" { baseDeployments = append(baseDeployments, sqlStatement) } else if deploymentType == "patch" { patchDeployments = append(patchDeployments, sqlStatement) } else { log.Fatalf("Unknown deployment type (%+v) encountered in _orchestrator_db_deployment", deploymentType) } return nil }) if err != nil { log.Debugf("Deploying internal orchestrator tables to fix the above; this is a one time operation") // Table does not exist? Create it for first time for _, query := range internalDBDeploymentSQL { if _, err = execInternal(db, query); err != nil { log.Fatalf("Cannot initiate orchestrator internal deployment: %+v", err) } } } return baseDeployments, patchDeployments, nil }
func postReadAdjustments() { if Config.MySQLOrchestratorCredentialsConfigFile != "" { mySQLConfig := struct { Client struct { User string Password string } }{} err := gcfg.ReadFileInto(&mySQLConfig, Config.MySQLOrchestratorCredentialsConfigFile) if err != nil { log.Fatalf("Failed to parse gcfg data from file: %+v", err) } else { log.Debugf("Parsed orchestrator credentials from %s", Config.MySQLOrchestratorCredentialsConfigFile) Config.MySQLOrchestratorUser = mySQLConfig.Client.User Config.MySQLOrchestratorPassword = mySQLConfig.Client.Password } } if Config.MySQLTopologyCredentialsConfigFile != "" { mySQLConfig := struct { Client struct { User string Password string } }{} err := gcfg.ReadFileInto(&mySQLConfig, Config.MySQLTopologyCredentialsConfigFile) if err != nil { log.Fatalf("Failed to parse gcfg data from file: %+v", err) } else { log.Debugf("Parsed topology credentials from %s", Config.MySQLTopologyCredentialsConfigFile) Config.MySQLTopologyUser = mySQLConfig.Client.User Config.MySQLTopologyPassword = mySQLConfig.Client.Password } } if Config.RecoveryPeriodBlockSeconds == 0 && Config.RecoveryPeriodBlockMinutes > 0 { // RecoveryPeriodBlockSeconds is a newer addition that overrides RecoveryPeriodBlockMinutes // The code does not consider RecoveryPeriodBlockMinutes anymore, but RecoveryPeriodBlockMinutes // still supported in config file for backwards compatibility Config.RecoveryPeriodBlockSeconds = Config.RecoveryPeriodBlockMinutes * 60 } }
// writeInternalDeployment will persist a successful deployment func writeInternalDeployment(db *sql.DB, deploymentType string, sqlStatement string, statementIndex int) error { if !config.Config.SmartOrchestratorDatabaseUpdate { return nil } query := ` insert ignore into _orchestrator_db_deployment ( deployment_type, sql_statement, statement_digest, statement_index) VALUES ( ?, ?, CONCAT(SHA1(?), ':', LEFT(REPLACE(REPLACE(REPLACE(?, ' ', ''), '\n', ' '), '\t', ''), 60)), ?) ` if _, err := execInternal(db, query, deploymentType, sqlStatement, sqlStatement, sqlStatement, statementIndex); err != nil { log.Fatalf("Unable to write to _orchestrator_db_deployment: %+v", err) } return nil }
// main is the application's entry point. It will either spawn a CLI or HTTP itnerfaces. func main() { configFile := flag.String("config", "", "config file name") command := flag.String("c", "", "command, required. See full list of commands via 'orchestrator -c help'") strict := flag.Bool("strict", false, "strict mode (more checks, slower)") instance := flag.String("i", "", "instance, host_fqdn[:port] (e.g. db.company.com:3306, db.company.com)") sibling := flag.String("s", "", "sibling instance, host_fqdn[:port]") destination := flag.String("d", "", "destination instance, host_fqdn[:port] (synonym to -s)") owner := flag.String("owner", "", "operation owner") reason := flag.String("reason", "", "operation reason") duration := flag.String("duration", "", "maintenance duration (format: 59s, 59m, 23h, 6d, 4w)") pattern := flag.String("pattern", "", "regular expression pattern") clusterAlias := flag.String("alias", "", "cluster alias") pool := flag.String("pool", "", "Pool logical name (applies for pool-related commands)") hostnameFlag := flag.String("hostname", "", "Hostname/fqdn/CNAME/VIP (applies for hostname/resolve related commands)") discovery := flag.Bool("discovery", true, "auto discovery mode") quiet := flag.Bool("quiet", false, "quiet") verbose := flag.Bool("verbose", false, "verbose") debug := flag.Bool("debug", false, "debug mode (very verbose)") stack := flag.Bool("stack", false, "add stack trace upon error") config.RuntimeCLIFlags.Databaseless = flag.Bool("databaseless", false, "EXPERIMENTAL! Work without backend database") config.RuntimeCLIFlags.SkipUnresolve = flag.Bool("skip-unresolve", false, "Do not unresolve a host name") config.RuntimeCLIFlags.SkipUnresolveCheck = flag.Bool("skip-unresolve-check", false, "Skip/ignore checking an unresolve mapping (via hostname_unresolve table) resolves back to same hostname") config.RuntimeCLIFlags.Noop = flag.Bool("noop", false, "Dry run; do not perform destructing operations") config.RuntimeCLIFlags.BinlogFile = flag.String("binlog", "", "Binary log file name") config.RuntimeCLIFlags.GrabElection = flag.Bool("grab-election", false, "Grab leadership (only applies to continuous mode)") flag.Parse() if *destination != "" && *sibling != "" { log.Fatalf("-s and -d are synonyms, yet both were specified. You're probably doing the wrong thing.") } if *destination == "" { *destination = *sibling } log.SetLevel(log.ERROR) if *verbose { log.SetLevel(log.INFO) } if *debug { log.SetLevel(log.DEBUG) } if *stack { log.SetPrintStackTrace(*stack) } log.Info("starting") runtime.GOMAXPROCS(math.MinInt(4, runtime.NumCPU())) if len(*configFile) > 0 { config.ForceRead(*configFile) } else { config.Read("/etc/orchestrator.conf.json", "conf/orchestrator.conf.json", "orchestrator.conf.json") } if *config.RuntimeCLIFlags.Databaseless { config.Config.DatabaselessMode__experimental = true } if config.Config.Debug { log.SetLevel(log.DEBUG) } if *quiet { // Override!! log.SetLevel(log.ERROR) } if config.Config.EnableSyslog { log.EnableSyslogWriter("orchestrator") log.SetSyslogLevel(log.INFO) } if config.Config.AuditToSyslog { inst.EnableAuditSyslog() } if len(flag.Args()) == 0 && *command == "" { // No command, no argument: just prompt fmt.Println(prompt) return } switch { case len(flag.Args()) == 0 || flag.Arg(0) == "cli": app.CliWrapper(*command, *strict, *instance, *destination, *owner, *reason, *duration, *pattern, *clusterAlias, *pool, *hostnameFlag) case flag.Arg(0) == "http": app.Http(*discovery) default: fmt.Fprintln(os.Stderr, `Usage: orchestrator --options... [cli|http] See complete list of commands: orchestrator -c help Full blown documentation: orchestrator `) os.Exit(1) } }
// 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()) } }