Interface of MySQL and MariaDB now is unstable
You can use wal-g as a tool for encrypting, compressing MySQL backups and push/fetch them to/from storage without saving it on your filesystem.
WALG_MYSQL_DATASOURCE_NAME
To configure the connection string for MySQL. Required. Format user:password@host/dbname
WALG_MYSQL_SSL_CA
To use SSL, a path to file with certificates should be set to this variable.
WALG_STREAM_CREATE_COMMAND
Command to create MySQL backup, should return backup as single stream to STDOUT. Requried.
WALG_STREAM_RESTORE_COMMAND
Command to unpack MySQL backup, should take backup (created by WALG_STREAM_CREATE_COMMAND
)
to STDIN and unpack it to MySQL datadir. Required.
WALG_MYSQL_BACKUP_PREPARE_COMMAND
Command to prepare MySQL backup after restoring. Optional. Needed for xtrabackup case.
WALG_DELTA_MAX_STEPS
Delta-backup is the difference between previously taken backup and present state. WALG_DELTA_MAX_STEPS
determines how many delta backups can be between full backups. Defaults to 0.
Restoration process will automatically fetch all necessary deltas and base backup and compose valid restored backup (you still need WALs after start of last backup to restore consistent cluster).
Delta computation is based on ModTime of file system and LSN number of pages in datafiles.
Note: Incremental backups only supported in wal-g xtrabackup-push
command.
WALG_DELTA_ORIGIN
To configure base for next delta backup (only if WALG_DELTA_MAX_STEPS
is not exceeded). WALG_DELTA_ORIGIN
can be LATEST (chaining increments), LATEST_FULL (for bases where volatile part is compact and chaining has no meaning - deltas overwrite each other). Defaults to LATEST.
Note: Incremental backups only supported in wal-g xtrabackup-push
command.
WALG_MYSQL_BINLOG_REPLAY_COMMAND
Command to replay binlog on running MySQL. Required for binlog-replay command.
WALG_MYSQL_BINLOG_DST
To place binlogs in the specified directory during binlog-fetch, binlog-replay or binlog-server
WALG_MYSQL_BINLOG_SERVER_HOST
WALG_MYSQL_BINLOG_SERVER_PORT
WALG_MYSQL_BINLOG_SERVER_USER
WALG_MYSQL_BINLOG_SERVER_PASSWORD
To configure the data to connect the replica to for binlog server.
WALG_MYSQL_BINLOG_SERVER_ID
To configure the server id of the binlog server. Should be unique for each replica.
WALG_MYSQL_BINLOG_SERVER_REPLICA_SOURCE
To configure the connection string that will be used by binlog-server
to connect to your MySQL. DSN format: user:password@host/dbname
Operations with binlogs: If you'd like to do binlog operations with wal-g don't forget to activate the binary log by starting mysql/mariadb with --log-bin and --log-basename=[name].
WALG_STREAM_SPLITTER_PARTITIONS
To configure split backup stream into several parts and upload them in parallel.
Backup file names have a suffix _0000.bz
.
WALG_STREAM_SPLITTER_BLOCK_SIZE
To configure block size (bytes) into which backup stream split. Blocks of this length size is put into each partition in turn.
WALG_STREAM_SPLITTER_MAX_FILE_SIZE
To configure max file size (bytes) before compressing. If partition size become more than max file size, it split on several files.
Backup file names have a suffix _0000_0000.bz
.
WALG_BACKUP_DOWNLOAD_MAX_RETRIES
Configure max attempts to download backup file. Default value 1
.
WAL-G mysql extension currently supports these commands:
Creates new backup and send it to storage. Runs WALG_STREAM_CREATE_COMMAND
to create backup.
wal-g backup-push
Creates new backup with xtrabackup
tool and send it to storage. Runs WALG_STREAM_CREATE_COMMAND
to create backup.
WAL-G levereages knowledge of xtrabackup format to support additional feature (e.g. incremental backups)
wal-g xtrabackup-push
WALG_MYSQL_INCREMENTAL_BACKUP_DST
To place incremental backip in the specified directory during backup-fetch
Lists currently available backups in storage
wal-g backup-list
Fetches backup from storage and restores it to datadir.
Runs WALG_STREAM_RESTORE_COMMAND
to restore backup.
User should specify the name of the backup to fetch.
wal-g backup-fetch example_backup
WAL-G can also fetch the latest backup using:
wal-g backup-fetch LATEST
Sends (not yet archived) binlogs to storage. Typically run in CRON.
wal-g binlog-push
When WALG_MYSQL_CHECK_GTIDS
is set wal-g will try to be upload only binlogs which GTID sets contains events that
wasn't seen before. This is done by parsing binlogs and peeking first PREVIOUS_GTIDS_EVENT that holds GTID set of
all executed transactions at the moment this particular binlog file created.
This feature may be useful when you are uploading binlogs from different hosts (e.g. after master switchower)
Note: Don't use WALG_MYSQL_CHECK_GTIDS
when GTIDs are not used - it will slow down binlog upload.
Fetches binlogs from storage and saves them to WALG_MYSQL_BINLOG_DST
folder.
User should specify the name of the backup starting with which to fetch an binlog.
User may also specify time in RFC3339 format until which should be fetched (used for PITR).
User have to replay binlogs manually in that case.
wal-g binlog-fetch --since "backupname"
or
wal-g binlog-fetch --since "backupname" --until "2006-01-02T15:04:05Z07:00"
or
wal-g binlog-fetch --since LATEST --until "2006-01-02T15:04:05Z07:00"
You can stop wal-g from fetching newly created/modified binlogs by specifying --until-binlog-last-modified-time
option.
This may be useful to achieve exact clones of the same database in scenarios when new binlogs are uploaded concurrently whith your restore process.
wal-g binlog-replay --since LATEST --until "2006-01-02T15:04:05Z07:00" --until-binlog-last-modified-time "2006-01-02T15:04:05Z07:00"
Fetches binlogs from storage and passes them to WALG_MYSQL_BINLOG_REPLAY_COMMAND
to replay on running MySQL server.
User should specify the name of the backup starting with which to fetch an binlog.
User may also specify time in RFC3339 format until which should be fetched (used for PITR).
If until
timestamp is in the future, wal-g will search for newly uploaded binlogs until no new found.
Binlogs are temporarily save in WALG_MYSQL_BINLOG_DST
folder.
Replay command gets name of binlog to replay via environment variable WALG_MYSQL_CURRENT_BINLOG
and stop-date via WALG_MYSQL_BINLOG_END_TS
, which are set for each invocation.
wal-g binlog-replay --since "backupname"
or
wal-g binlog-replay --since "backupname" --until "2006-01-02T15:04:05Z07:00"
or
wal-g binlog-replay --since LATEST --until "2006-01-02T15:04:05Z07:00"
You can stop wal-g from applying newly created/modified binlogs by specifying --until-binlog-last-modified-time
option.
This may be useful to achieve exact clones of the same database in scenarios when new binlogs are uploaded concurrently whith your restore process.
wal-g binlog-replay --since LATEST --until "2006-01-02T15:04:05Z07:00" --until-binlog-last-modified-time "2006-01-02T15:04:05Z07:00"
Runs mysql server implementation which can be used to fetch binlogs from storage and send them to MySQL slave by replication protocol.
wal-g binlog-server
Backups can be marked as permanent to prevent them from being removed when running delete
. To mark backup as permanent call wal-g backup-mark -b backup_name
. To remove permanent flag - call wal-g backup-mark -b backup_name -i
When incremental backup is marked as permanent - all parent backups also marked as permanent.
It's recommended to use wal-g with xtrabackup tool in case of MySQL for creating lock-less backups. Here's typical wal-g configuration for that case:
WALG_MYSQL_DATASOURCE_NAME=user:pass@tcp(localhost:3306)/mysql
WALG_STREAM_CREATE_COMMAND="xtrabackup --backup --stream=xbstream --datadir=/var/lib/mysql"
WALG_STREAM_RESTORE_COMMAND="xbstream -x -C /var/lib/mysql"
WALG_MYSQL_BACKUP_PREPARE_COMMAND="xtrabackup --prepare --target-dir=/var/lib/mysql"
WALG_MYSQL_BINLOG_REPLAY_COMMAND='mysqlbinlog --stop-datetime="$WALG_MYSQL_BINLOG_END_TS" "$WALG_MYSQL_CURRENT_BINLOG" | mysql'
Restore procedure is a bit tricky:
- stop mysql
- clean a datadir (typically
/var/lib/mysql
) - fetch and prepare desired backup using
wal-g backup-fetch "backup_name"
- start mysql
- in case of you have replication and GTID enabled: set mysql GTID_PURGED variable to value from
/var/lib/mysql/xtrabackup_binlog_info
, using
gtids=$(tr -d '\n' < /var/lib/mysql/xtrabackup_binlog_info | awk '{print $3}')
mysql -e "RESET MASTER; SET @@GLOBAL.GTID_PURGED='$gtids';"
- for PITR, replay binlogs with
wal-g binlog-replay --since "backup_name" --until "2006-01-02T15:04:05Z"
It's possible to use wal-g with standard mysqldump/mysql tools. In that case MySQL backup is a plain SQL script. Here's typical wal-g configuration for that case:
WALG_MYSQL_DATASOURCE_NAME=user:pass@localhost/mysql
WALG_STREAM_CREATE_COMMAND="mysqldump --all-databases --single-transaction --set-gtid-purged=ON"
WALG_STREAM_RESTORE_COMMAND="mysql"
WALG_MYSQL_BINLOG_REPLAY_COMMAND='mysqlbinlog --stop-datetime="$WALG_MYSQL_BINLOG_END_TS" "$WALG_MYSQL_CURRENT_BINLOG" | mysql'
Restore procedure is straightforward:
- start mysql (it's recommended to create new mysql instance)
- fetch and apply desired backup using
wal-g backup-fetch "backup_name"
wal-g can work as replication source to do fast PiTR. In this case it will serve binlogs from storage directly to MySQL. It is expected that PiTR in this mode will be much faster than in classic approach because there won't be additional transformations from binlog to SQL, also you can benefit from multi-threaded replication.
WALG_MYSQL_BINLOG_SERVER_HOST="127.0.0.1"
WALG_MYSQL_BINLOG_SERVER_PORT=9306
WALG_MYSQL_BINLOG_SERVER_USER="walg"
WALG_MYSQL_BINLOG_SERVER_PASSWORD="walgpwd"
WALG_MYSQL_BINLOG_SERVER_ID=99
WALG_MYSQL_BINLOG_SERVER_REPLICA_SOURCE="user:[email protected]:3306/db"
Restore procedure is straightforward:
- restore backup
- disable replication threads in MySQL:
skip-slave-start
- start MySQL, purge GTIDs (see above)
- in second terminal start binlog-server:
wal-g binlog-server --until "1985-10-26T01:21:00Z"
- in MySQL:
SET GLOBAL SERVER_ID=999 CHANGE MASTER TO MASTER_HOST="127.0.0.1", MASTER_PORT=9306, MASTER_USER="walg", MASTER_PASSWORD="walgpwd", MASTER_AUTO_POSITION=1; SHOW SLAVE STATUS \G START SLAVE;
- wait until wal-g exit (it will wait until binlogs will be applied)
- in case of errors use classic approach
It's recommended to use wal-g with mariabackup
tool in case of MariaDB for creating lock-less backups.
Here's typical wal-g configuration for that case:
WALG_MYSQL_DATASOURCE_NAME=user:pass@tcp(localhost:3305)/mysql
WALG_STREAM_CREATE_COMMAND="mariabackup --backup --stream=xbstream --datadir=/var/lib/mysql"
WALG_STREAM_RESTORE_COMMAND="mbstream -x -C /var/lib/mysql"
WALG_MYSQL_BACKUP_PREPARE_COMMAND="mariabackup --prepare --target-dir=/var/lib/mysql"
WALG_MYSQL_BINLOG_REPLAY_COMMAND='mysqlbinlog --stop-datetime="$WALG_MYSQL_BINLOG_END_TS" "$WALG_MYSQL_CURRENT_BINLOG" | mysql'
For the restore procedure you have to do similar things to what the offical docs says about full backup and restore:
- stop mariadb
- clean a datadir (typically
/var/lib/mysql
) - fetch and prepare desired backup using
wal-g backup-fetch "backup_name"
- after the previous step you might have to fix file permissions:
chown -R mysql:mysql /var/lib/mysql
- start mariadb
- WAL-G doesn't support automatic PITR for MariaDB. There are 2 possible workarounds:
- You can configure restored database to replicate from your master, so it will be able to catch up (follow the official docs)
- You can manually replay events with binlog and position:
wal-g binlog-fetch --since [backup name | LATEST]
# Get binlog-name, position and GTIDs:
tail -n 1 < /var/lib/mysql/xtrabackup_binlog_info
# eg 'mysql-bin.000005 385 0-1-5763'
# then replay it manually:
mysql -e "STOP ALL SLAVES; SET GLOBAL gtid_slave_pos='$gtids';"
mysqlbinlog --stop-datetime="some point in time" --start-position [position above] [all binlogs starting from thouse we seen above] | mysql --user XXX --host YYY [other options]
The procedure is same as in case of MySQL. You can follow the instructions from the previous section.