go/vt/mysqlctl: add configurable read buffer to builtin backups (#12073)

* go/vt/mysqlctl: add configurable read buffer to builtin backups Signed-off-by: Max Englander <[email protected]> * pr review: use uint Signed-off-by: Max Englander <[email protected]> * doc/releasenotes: v16 --builtinbackup-file-read-buffer-size and --builtinbackup-file-write-buffer-size Signed-off-by: Max Englander <[email protected]> * update flags txt Signed-off-by: Max Englander <[email protected]> * make part of v17 Signed-off-by: Max Englander <[email protected]> * Update doc/releasenotes/17_0_0_summary.md Co-authored-by: Deepthi Sigireddi <[email protected]> Signed-off-by: Max Englander <[email protected]> * improve wording -block Signed-off-by: Max Englander <[email protected]> --------- Signed-off-by: Max Englander <[email protected]> Signed-off-by: Max Englander <[email protected]> Co-authored-by: Deepthi Sigireddi <[email protected]>
vitessio · Feb 9, 2023 · b22b419 · b22b419
1 parent 07c85db
commit b22b419
Show file tree

Hide file tree

Showing 6 changed files with 56 additions and 5 deletions.
diff --git a/doc/releasenotes/17_0_0_summary.md b/doc/releasenotes/17_0_0_summary.md
@@ -0,0 +1,27 @@
+## Summary
+
+### Table of Contents
+
+- **[Major Changes](#major-changes)**
+  - **[New command line flags and behavior](#new-flag)**
+    - [Builtin backup: read buffering flags](#builtin-backup-read-buffering-flags)
+
+## <a id="major-changes"/> Major Changes
+
+### <a id="new-flag"/> New command line flags and behavior
+
+#### <a id="builtin-backup-read-buffering-flags" /> Backup --builtinbackup-file-read-buffer-size and --builtinbackup-file-write-buffer-size
+
+Prior to v17 the builtin Backup Engine does not use read buffering for restores, and for backups uses a hardcoded write buffer size of 2097152 bytes.
+
+In v17 these defaults may be tuned with, respectively `--builtinbackup-file-read-buffer-size` and `--builtinbackup-file-write-buffer-size`.
+
+ - `--builtinbackup-file-read-buffer-size`:  read files using an IO buffer of this many bytes. Golang defaults are used when set to 0.
+ - `--builtinbackup-file-write-buffer-size`: write files using an IO buffer of this many bytes. Golang defaults are used when set to 0. (default 2097152)
+
+These flags are applicable to the following programs:
+
+ - `vtbackup`
+ - `vtctld`
+ - `vttablet`
+ - `vttestserver`
diff --git a/go/flags/endtoend/vtctld.txt b/go/flags/endtoend/vtctld.txt
@@ -11,6 +11,8 @@ Usage of vtctld:
       --backup_storage_compress                                          if set, the backup files will be compressed. (default true)
       --backup_storage_implementation string                             Which backup storage implementation to use for creating and restoring backups.
       --backup_storage_number_blocks int                                 if backup_storage_compress is true, backup_storage_number_blocks sets the number of blocks that can be processed, at once, before the writer blocks, during compression (default is 2). It should be equal to the number of CPUs available for compression. (default 2)
+      --builtinbackup-file-read-buffer-size uint                         read files using an IO buffer of this many bytes. Golang defaults are used when set to 0.
+      --builtinbackup-file-write-buffer-size uint                        write files using an IO buffer of this many bytes. Golang defaults are used when set to 0. (default 2097152)
       --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)
       --catch-sigpipe                                                    catch and ignore SIGPIPE on stdout and stderr if specified

diff --git a/go/flags/endtoend/vttablet.txt b/go/flags/endtoend/vttablet.txt
@@ -26,6 +26,8 @@ Usage of vttablet:
       --binlog_ssl_key string                                            PITR restore parameter: Filename containing mTLS client private key for use in binlog server authentication.
       --binlog_ssl_server_name string                                    PITR restore parameter: TLS server name (common name) to verify against for the binlog server we are connecting to (If not set: use the hostname or IP supplied in --binlog_host).
       --binlog_user string                                               PITR restore parameter: username of binlog server.
+      --builtinbackup-file-read-buffer-size uint                         read files using an IO buffer of this many bytes. Golang defaults are used when set to 0.
+      --builtinbackup-file-write-buffer-size uint                        write files using an IO buffer of this many bytes. Golang defaults are used when set to 0. (default 2097152)
       --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)
       --catch-sigpipe                                                    catch and ignore SIGPIPE on stdout and stderr if specified

diff --git a/go/flags/endtoend/vttestserver.txt b/go/flags/endtoend/vttestserver.txt
@@ -6,6 +6,8 @@ Usage of vttestserver:
       --backup_storage_block_size int                                    if backup_storage_compress is true, backup_storage_block_size sets the byte size for each block while compressing (default is 250000). (default 250000)
       --backup_storage_compress                                          if set, the backup files will be compressed. (default true)
       --backup_storage_number_blocks int                                 if backup_storage_compress is true, backup_storage_number_blocks sets the number of blocks that can be processed, at once, before the writer blocks, during compression (default is 2). It should be equal to the number of CPUs available for compression. (default 2)
+      --builtinbackup-file-read-buffer-size uint                         read files using an IO buffer of this many bytes. Golang defaults are used when set to 0.
+      --builtinbackup-file-write-buffer-size uint                        write files using an IO buffer of this many bytes. Golang defaults are used when set to 0. (default 2097152)
       --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)
       --catch-sigpipe                                                    catch and ignore SIGPIPE on stdout and stderr if specified

diff --git a/go/vt/mysqlctl/builtinbackupengine.go b/go/vt/mysqlctl/builtinbackupengine.go
@@ -51,7 +51,6 @@ import (
 const (
 	builtinBackupEngineName = "builtin"
 	autoIncrementalFromPos  = "auto"
-	writerBufferSize        = 2 * 1024 * 1024
 	dataDictionaryFile      = "mysql.ibd"
 )
 
@@ -62,6 +61,17 @@ var (
 	BuiltinBackupMysqldTimeout = 10 * time.Minute
 
 	builtinBackupProgress = 5 * time.Second
+
+	// Controls the size of the IO buffer used when reading files during backups.
+	builtinBackupFileReadBufferSize uint
+
+	// Controls the size of the IO buffer used when writing files during restores.
+	builtinBackupFileWriteBufferSize uint = 2 * 1024 * 1024 /* 2 MiB */
+
+	// Controls the size of the IO buffer used when writing to backupstorage
+	// engines during backups.  The backupstorage may be a physical file,
+	// network, or something else.
+	builtinBackupStorageWriteBufferSize = 2 * 1024 * 1024 /* 2 MiB */
 )
 
 // BuiltinBackupEngine encapsulates the logic of the builtin engine
@@ -123,6 +133,8 @@ func init() {
 func registerBuiltinBackupEngineFlags(fs *pflag.FlagSet) {
 	fs.DurationVar(&BuiltinBackupMysqldTimeout, "builtinbackup_mysqld_timeout", BuiltinBackupMysqldTimeout, "how long to wait for mysqld to shutdown at the start of the backup.")
 	fs.DurationVar(&builtinBackupProgress, "builtinbackup_progress", builtinBackupProgress, "how often to send progress updates when backing up large files.")
+	fs.UintVar(&builtinBackupFileReadBufferSize, "builtinbackup-file-read-buffer-size", builtinBackupFileReadBufferSize, "read files using an IO buffer of this many bytes. Golang defaults are used when set to 0.")
+	fs.UintVar(&builtinBackupFileWriteBufferSize, "builtinbackup-file-write-buffer-size", builtinBackupFileWriteBufferSize, "write files using an IO buffer of this many bytes. Golang defaults are used when set to 0.")
 }
 
 // isIncrementalBackup is a convenience function to check whether the params indicate an incremental backup request
@@ -569,7 +581,7 @@ type backupPipe struct {
 	closed int32
 }
 
-func newBackupWriter(filename string, maxSize int64, w io.Writer) *backupPipe {
+func newBackupWriter(filename string, writerBufferSize int, maxSize int64, w io.Writer) *backupPipe {
 	return &backupPipe{
 		crc32:    crc32.NewIEEE(),
 		w:        bufio.NewWriterSize(w, writerBufferSize),
@@ -670,10 +682,11 @@ func (be *BuiltinBackupEngine) backupFile(ctx context.Context, params BackupPara
 		}
 	}(name, fe.Name)
 
-	bw := newBackupWriter(fe.Name, fi.Size(), wc)
+	bw := newBackupWriter(fe.Name, builtinBackupStorageWriteBufferSize, fi.Size(), wc)
 	br := newBackupReader(fe.Name, fi.Size(), source)
 	go br.ReportProgress(builtinBackupProgress, params.Logger)
 
+	var reader io.Reader = br
 	var writer io.Writer = bw
 
 	// Create the gzip compression pipe, if necessary.
@@ -690,9 +703,13 @@ func (be *BuiltinBackupEngine) backupFile(ctx context.Context, params BackupPara
 		writer = compressor
 	}
 
+	if builtinBackupFileReadBufferSize > 0 {
+		reader = bufio.NewReaderSize(br, int(builtinBackupFileReadBufferSize))
+	}
+
 	// Copy from the source file to writer (optional gzip,
 	// optional pipe, tee, output file and hasher).
-	_, err = io.Copy(writer, br)
+	_, err = io.Copy(writer, reader)
 	if err != nil {
 		return vterrors.Wrap(err, "cannot copy data")
 	}
@@ -875,7 +892,7 @@ func (be *BuiltinBackupEngine) restoreFile(ctx context.Context, params RestorePa
 	bp := newBackupReader(name, 0, source)
 	go bp.ReportProgress(builtinBackupProgress, params.Logger)
 
-	dst := bufio.NewWriterSize(dstFile, writerBufferSize)
+	dst := bufio.NewWriterSize(dstFile, int(builtinBackupFileWriteBufferSize))
 	var reader io.Reader = bp
 
 	// Create the uncompresser if needed.

diff --git a/go/vt/mysqlctl/xtrabackupengine.go b/go/vt/mysqlctl/xtrabackupengine.go
@@ -66,6 +66,7 @@ var (
 
 const (
 	streamModeTar        = "tar"
+	writerBufferSize     = 2 * 1024 * 1024 /*2 MiB*/
 	xtrabackupBinaryName = "xtrabackup"
 	xtrabackupEngineName = "xtrabackup"
 	xbstream             = "xbstream"