mysqlctl,vttablet,vtbackup: support restore tablet and vtbackup restore phase with CLONE

go/cmd/vtbackup/cli/vtbackup.go

@@ -53,6 +53,7 @@ import (
 
 	tabletmanagerdatapb "vitess.io/vitess/go/vt/proto/tabletmanagerdata"
 	topodatapb "vitess.io/vitess/go/vt/proto/topodata"
+	"vitess.io/vitess/go/vt/proto/vtrpc"
 )
 
 const (
@@ -93,6 +94,7 @@ var (
 	initShard          string
 	concurrency        = 4
 	incrementalFromPos string
+	restoreWithClone   bool
 
 	// mysqlctld-like flags
 	mysqlPort            = 3306
@@ -157,7 +159,7 @@ When run periodically for each shard, vtbackup can ensure these configurable pol
  * Old backups for the shard are removed.
 
 Whatever system launches vtbackup is responsible for the following:
- - Running vtbackup with similar flags that would be used for a vttablet and 
+ - Running vtbackup with similar flags that would be used for a vttablet and
    mysqlctld in the target shard to be backed up.
 
  - Provisioning as much disk space for vtbackup as would be given to vttablet.
@@ -226,6 +228,7 @@ func init() {
 	utils.SetFlagStringVar(Main.Flags(), &initShard, "init-shard", initShard, "(init parameter) shard to use for this tablet")
 	Main.Flags().IntVar(&concurrency, "concurrency", concurrency, "(init restore parameter) how many concurrent files to restore at once")
 	utils.SetFlagStringVar(Main.Flags(), &incrementalFromPos, "incremental-from-pos", incrementalFromPos, "Position, or name of backup from which to create an incremental backup. Default: empty. If given, then this backup becomes an incremental backup from given position or given backup. If value is 'auto', this backup will be taken from the last successful backup position.")
+	Main.Flags().BoolVar(&restoreWithClone, "restore-with-clone", restoreWithClone, "(init parameter) will perform the restore phase with MySQL CLONE, requires either --clone-from-primary or --clone-from-tablet")
 
 	// mysqlctld-like flags
 	utils.SetFlagIntVar(Main.Flags(), &mysqlPort, "mysql-port", mysqlPort, "MySQL port")
@@ -457,42 +460,49 @@ func takeBackup(ctx, backgroundCtx context.Context, topoServer *topo.Server, bac
 		return nil
 	}
 
-	phase.Set(phaseNameRestoreLastBackup, int64(1))
-	defer phase.Set(phaseNameRestoreLastBackup, int64(0))
-	backupDir := mysqlctl.GetBackupDir(initKeyspace, initShard)
-	log.Infof("Restoring latest backup from directory %v", backupDir)
-	restoreAt := time.Now()
-	params := mysqlctl.RestoreParams{
-		Cnf:                  mycnf,
-		Mysqld:               mysqld,
-		Logger:               logutil.NewConsoleLogger(),
-		Concurrency:          concurrency,
-		HookExtraEnv:         extraEnv,
-		DeleteBeforeRestore:  true,
-		DbName:               dbName,
-		Keyspace:             initKeyspace,
-		Shard:                initShard,
-		Stats:                backupstats.RestoreStats(),
-		MysqlShutdownTimeout: mysqlShutdownTimeout,
-	}
-	backupManifest, err := mysqlctl.Restore(ctx, params)
 	var restorePos replication.Position
-	switch err {
-	case nil:
-		// if err is nil, we expect backupManifest to be non-nil
-		restorePos = backupManifest.Position
-		log.Infof("Successfully restored from backup at replication position %v", restorePos)
-	case mysqlctl.ErrNoBackup:
-		// There is no backup found, but we may be taking the initial backup of a shard
-		if !allowFirstBackup {
-			return errors.New("no backup found; not starting up empty since --initial_backup flag was not enabled")
-		}
-		restorePos = replication.Position{}
-	default:
-		return fmt.Errorf("can't restore from backup: %v", err)
-	}
-	deprecatedDurationByPhase.Set("RestoreLastBackup", int64(time.Since(restoreAt).Seconds()))
-	phase.Set(phaseNameRestoreLastBackup, int64(0))
+	if restoreWithClone {
+		restorePos, err = mysqlctl.CloneFromDonor(ctx, topoServer, mysqld, initKeyspace, initShard)
+		if err != nil {
+			return vterrors.Wrap(err, "restore with clone failed")
+		}
+	} else {
+		phase.Set(phaseNameRestoreLastBackup, int64(1))
+		defer phase.Set(phaseNameRestoreLastBackup, int64(0))
+		backupDir := mysqlctl.GetBackupDir(initKeyspace, initShard)
+		log.Infof("Restoring latest backup from directory %v", backupDir)
+		restoreAt := time.Now()
+		params := mysqlctl.RestoreParams{
+			Cnf:                  mycnf,
+			Mysqld:               mysqld,
+			Logger:               logutil.NewConsoleLogger(),
+			Concurrency:          concurrency,
+			HookExtraEnv:         extraEnv,
+			DeleteBeforeRestore:  true,
+			DbName:               dbName,
+			Keyspace:             initKeyspace,
+			Shard:                initShard,
+			Stats:                backupstats.RestoreStats(),
+			MysqlShutdownTimeout: mysqlShutdownTimeout,
+		}
+		backupManifest, err := mysqlctl.Restore(ctx, params)
+		switch err {
+		case nil:
+			// if err is nil, we expect backupManifest to be non-nil
+			restorePos = backupManifest.Position
+			log.Infof("Successfully restored from backup at replication position %v", restorePos)
+		case mysqlctl.ErrNoBackup:
+			// There is no backup found, but we may be taking the initial backup of a shard
+			if !allowFirstBackup {
+				return vterrors.New(vtrpc.Code_FAILED_PRECONDITION, "no backup found; not starting up empty since --initial_backup flag was not enabled")
+			}
+			restorePos = replication.Position{}
+		default:
+			return vterrors.Wrap(err, "can't restore from backup")
+		}
+		deprecatedDurationByPhase.Set("RestoreLastBackup", int64(time.Since(restoreAt).Seconds()))
+		phase.Set(phaseNameRestoreLastBackup, int64(0))
+	}
 
 	// As of MySQL 8.0.21, you can disable redo logging using the ALTER INSTANCE
 	// DISABLE INNODB REDO_LOG statement. This functionality is intended for

go/cmd/vttablet/cli/cli_test.go

@@ -69,5 +69,5 @@ func TestRunFailsToStartTabletManager(t *testing.T) {
 	defer cancel()
 
 	err := Main.ExecuteContext(ctx)
-	require.ErrorContains(t, err, "you cannot enable --restore-from-backup without a my.cnf file")
+	require.ErrorContains(t, err, "you cannot enable --restore-from-backup or --restore-with-clone without a my.cnf file")
 }

go/flags/endtoend/mysqlctl.txt

@@ -70,7 +70,6 @@ Flags:
       --log_dir string                                              If non-empty, write log files in this directory
       --logtostderr                                                 log to standard error instead of files
       --max-stack-size int                                          configure the maximum stack size in bytes (default 67108864)
-      --mysql-clone-enabled                                         Enable MySQL CLONE plugin and user for backup/replica provisioning (requires MySQL 8.0.17+)
       --mysql-port int                                              MySQL port. (default 3306)
       --mysql-server-version string                                 MySQL server version to advertise. (default "8.4.6-Vitess")
       --mysql-socket string                                         Path to the mysqld socket file.

go/flags/endtoend/mysqlctld.txt

@@ -99,7 +99,6 @@ Flags:
       --log_dir string                                                   If non-empty, write log files in this directory
       --logtostderr                                                      log to standard error instead of files
       --max-stack-size int                                               configure the maximum stack size in bytes (default 67108864)
-      --mysql-clone-enabled                                              Enable MySQL CLONE plugin and user for backup/replica provisioning (requires MySQL 8.0.17+)
       --mysql-port int                                                   MySQL port (default 3306)
       --mysql-server-version string                                      MySQL server version to advertise. (default "8.4.6-Vitess")
       --mysql-socket string                                              Path to the mysqld socket file

go/flags/endtoend/vtbackup.txt

@@ -6,7 +6,7 @@ When run periodically for each shard, vtbackup can ensure these configurable pol
  * Old backups for the shard are removed.
 
 Whatever system launches vtbackup is responsible for the following:
- - Running vtbackup with similar flags that would be used for a vttablet and 
+ - Running vtbackup with similar flags that would be used for a vttablet and
    mysqlctld in the target shard to be backed up.
 
  - Provisioning as much disk space for vtbackup as would be given to vttablet.
@@ -68,6 +68,9 @@ Flags:
       --builtinbackup-mysqld-timeout duration                       how long to wait for mysqld to shutdown at the start of the backup. (default 10m0s)
       --builtinbackup-progress duration                             how often to send progress updates when backing up large files. (default 5s)
       --ceph-backup-storage-config string                           Path to JSON config file for ceph backup storage. (default "ceph_backup_config.json")
+      --clone-from-primary                                          Clone data from the primary tablet in the shard using MySQL CLONE REMOTE instead of restoring from backup. Requires MySQL 8.0.17+. Mutually exclusive with --clone-from-tablet.
+      --clone-from-tablet string                                    Clone data from this tablet using MySQL CLONE REMOTE instead of restoring from backup (tablet alias, e.g., zone1-123). Requires MySQL 8.0.17+. Mutually exclusive with --clone-from-primary.
+      --clone-restart-wait-timeout duration                         Timeout for waiting for MySQL to restart after CLONE REMOTE. (default 5m0s)
       --compression-engine-name string                              compressor engine used for compression. (default "pargzip")
       --compression-level int                                       what level to pass to the compressor. (default 1)
       --concurrency int                                             (init restore parameter) how many concurrent files to restore at once (default 4)
@@ -189,6 +192,7 @@ Flags:
       --mycnf-slow-log-path string                                  mysql slow query log path
       --mycnf-socket-file string                                    mysql socket file
       --mycnf-tmp-dir string                                        mysql tmp directory
+      --mysql-clone-enabled                                         Enable MySQL CLONE plugin and user for backup/replica provisioning (requires MySQL 8.0.17+)
       --mysql-port int                                              MySQL port (default 3306)
       --mysql-server-version string                                 MySQL server version to advertise. (default "8.4.6-Vitess")
       --mysql-shell-backup-location string                          location where the backup will be stored
@@ -207,6 +211,7 @@ Flags:
       --purge-logs-interval duration                                how often try to remove old logs (default 1h0m0s)
       --remote-operation-timeout duration                           time to wait for a remote operation (default 15s)
       --restart-before-backup                                       Perform a mysqld clean/full restart after applying binlogs, but before taking the backup. Only makes sense to work around xtrabackup bugs.
+      --restore-with-clone                                          (init parameter) will perform the restore phase with MySQL CLONE, requires either --clone-from-primary or --clone-from-tablet
       --s3-backup-aws-endpoint string                               endpoint of the S3 backend (region must be provided).
       --s3-backup-aws-min-partsize int                              Minimum part size to use, defaults to 5MiB but can be increased due to the dataset size. (default 5242880)
       --s3-backup-aws-region string                                 AWS region to use. (default "us-east-1")

go/flags/endtoend/vtcombo.txt

@@ -36,6 +36,9 @@ Flags:
       --builtinbackup-progress duration                                  how often to send progress updates when backing up large files. (default 5s)
       --catch-sigpipe                                                    catch and ignore SIGPIPE on stdout and stderr if specified
       --cell string                                                      cell to use
+      --clone-from-primary                                               Clone data from the primary tablet in the shard using MySQL CLONE REMOTE instead of restoring from backup. Requires MySQL 8.0.17+. Mutually exclusive with --clone-from-tablet.
+      --clone-from-tablet string                                         Clone data from this tablet using MySQL CLONE REMOTE instead of restoring from backup (tablet alias, e.g., zone1-123). Requires MySQL 8.0.17+. Mutually exclusive with --clone-from-primary.
+      --clone-restart-wait-timeout duration                              Timeout for waiting for MySQL to restart after CLONE REMOTE. (default 5m0s)
       --compression-engine-name string                                   compressor engine used for compression. (default "pargzip")
       --compression-level int                                            what level to pass to the compressor. (default 1)
       --config-file string                                               Full path of the config file (with extension) to use. If set, --config-path, --config-type, and --config-name are ignored.
@@ -323,6 +326,7 @@ Flags:
       --restore-from-backup-ts string                                    (init restore parameter) if set, restore the latest backup taken at or before this timestamp. Example: '2021-04-29.133050'
       --restore-to-pos string                                            (init incremental restore parameter) if set, run a point in time recovery that ends with the given position. This will attempt to use one full backup followed by zero or more incremental backups
       --restore-to-timestamp string                                      (init incremental restore parameter) if set, run a point in time recovery that restores up to the given timestamp, if possible. Given timestamp in RFC3339 format. Example: '2006-01-02T15:04:05Z07:00'
+      --restore-with-clone                                               (init restore parameter) will restore from a clone, requires either --clone-from-primary or --clone-from-tablet, mutually exclusive with --restore-from-backup
       --retain-online-ddl-tables duration                                How long should vttablet keep an old migrated table before purging it (default 24h0m0s)
       --sanitize-log-messages                                            Remove potentially sensitive information in tablet INFO, WARNING, and ERROR log messages such as query parameters.
       --schema-change-reload-timeout duration                            query server schema change reload timeout, this is how long to wait for the signaled schema reload operation to complete before giving up (default 30s)

go/flags/endtoend/vttablet.txt

@@ -70,6 +70,9 @@ Flags:
       --builtinbackup-progress duration                                  how often to send progress updates when backing up large files. (default 5s)
       --catch-sigpipe                                                    catch and ignore SIGPIPE on stdout and stderr if specified
       --ceph-backup-storage-config string                                Path to JSON config file for ceph backup storage. (default "ceph_backup_config.json")
+      --clone-from-primary                                               Clone data from the primary tablet in the shard using MySQL CLONE REMOTE instead of restoring from backup. Requires MySQL 8.0.17+. Mutually exclusive with --clone-from-tablet.
+      --clone-from-tablet string                                         Clone data from this tablet using MySQL CLONE REMOTE instead of restoring from backup (tablet alias, e.g., zone1-123). Requires MySQL 8.0.17+. Mutually exclusive with --clone-from-primary.
+      --clone-restart-wait-timeout duration                              Timeout for waiting for MySQL to restart after CLONE REMOTE. (default 5m0s)
       --compression-engine-name string                                   compressor engine used for compression. (default "pargzip")
       --compression-level int                                            what level to pass to the compressor. (default 1)
       --config-file string                                               Full path of the config file (with extension) to use. If set, --config-path, --config-type, and --config-name are ignored.
@@ -314,6 +317,7 @@ Flags:
       --restore-from-backup-ts string                                    (init restore parameter) if set, restore the latest backup taken at or before this timestamp. Example: '2021-04-29.133050'
       --restore-to-pos string                                            (init incremental restore parameter) if set, run a point in time recovery that ends with the given position. This will attempt to use one full backup followed by zero or more incremental backups
       --restore-to-timestamp string                                      (init incremental restore parameter) if set, run a point in time recovery that restores up to the given timestamp, if possible. Given timestamp in RFC3339 format. Example: '2006-01-02T15:04:05Z07:00'
+      --restore-with-clone                                               (init restore parameter) will restore from a clone, requires either --clone-from-primary or --clone-from-tablet, mutually exclusive with --restore-from-backup
       --retain-online-ddl-tables duration                                How long should vttablet keep an old migrated table before purging it (default 24h0m0s)
       --s3-backup-aws-endpoint string                                    endpoint of the S3 backend (region must be provided).
       --s3-backup-aws-min-partsize int                                   Minimum part size to use, defaults to 5MiB but can be increased due to the dataset size. (default 5242880)

go/flags/endtoend/vttestserver.txt

@@ -19,6 +19,9 @@ Flags:
       --catch-sigpipe                                                    catch and ignore SIGPIPE on stdout and stderr if specified
       --cells strings                                                    Comma separated list of cells (default [test])
       --charset string                                                   MySQL charset (default "utf8mb4")
+      --clone-from-primary                                               Clone data from the primary tablet in the shard using MySQL CLONE REMOTE instead of restoring from backup. Requires MySQL 8.0.17+. Mutually exclusive with --clone-from-tablet.
+      --clone-from-tablet string                                         Clone data from this tablet using MySQL CLONE REMOTE instead of restoring from backup (tablet alias, e.g., zone1-123). Requires MySQL 8.0.17+. Mutually exclusive with --clone-from-primary.
+      --clone-restart-wait-timeout duration                              Timeout for waiting for MySQL to restart after CLONE REMOTE. (default 5m0s)
       --compression-engine-name string                                   compressor engine used for compression. (default "pargzip")
       --compression-level int                                            what level to pass to the compressor. (default 1)
       --config-file string                                               Full path of the config file (with extension) to use. If set, --config-path, --config-type, and --config-name are ignored.