About Binary Logs
MySQL implements replication by logging queries into a series of binary log files. The files are written to the server's local file system. Slaves connect to the MySQL instance and request these files starting at a specific file and offset.
MySQL's binary logging has the following drawbacks:
- It maintains a single binary log for the entire system.
- Fault tolerance is provided by the underlying filesystem, if at all.
- Static configuration is read only when the server starts, so changes to /etc/my.cnf require server restart.
ClustrixDB surpasses MySQL's capabilities by providing the following additional features:
- ClustrixDB can maintain multiple independent binary logs (for example, binlogs per table, per database, or for a set of tables and databases).
- The binary logs are fault-tolerant, with the same guarantees as the rest of the ClustrixDB system.
- Writes to the binlog are transactional, consistent, and durable (full ACID guarantee).
- Binary logs can be created and dropped online.
To configure a ClustrixDB system with a single row-based binary log, issue the following commands:
master> CREATE BINLOG 'clustrix-bin' FORMAT='ROW'; master> SHOW MASTER STATUS;
To disable binary logging and drop (permanently) an existing binlog:
master> DROP BINLOG 'clustrix-bin';
When running MySQL database as a slave to a ClustrixDB master, ClustrixDB does not support the variable binlog_checksum, which causes the master to write checksums for events written to the binary log.
MySQL 5.7 Replication and GTIDs
Beginning optionally with MySQL 5.6 and required in MySQL 5.7, MySQL produces global transaction identifiers (GTIDs). ClustrixDB does not implement nor support GTIDs. To enable replication between ClustrixDB and a MySQL instance that produces GTIDs, the following settings must be configured, dependent on replication direction.
For ClustrixDB (Master) to MySQL 5.7 (Slave) Replication
- The MySQL slave must have gtid_mode set to OFF, OFF_PERMISSIVE, or ON_PERMISSIVE.
- The MySQL slave should set enforce-gtid-consistency to OFF as described in Startup Options Used with GTID Replication.
For MySQL 5.7 (Master) to ClustrixDB (Slave) Replication
- Run ClustrixDB 7.6 or higher.
- The MySQL master configuration is inconsequential and does not require special changes.
- Further information is available at Using ClustrixDB as a Replication Slave.
ClustrixDB will not pass GTID events to its binlogs. This is similar to the behavior of a MySQL 5.6 slave with gtid_mode set to OFF.
Create Replication User
The user name and password used for replication are stored as plain text within the binlogs. As such, Clustrix recommends establishing a separate account for exclusive use with replication to prevent compromising the security of regularly used accounts.
Follow this sample to create an account that will be used when setting up a slave. You must have privileges to CREATE USER and GRANT to perform this step.
master> CREATE USER 'replication'@'%' IDENTIFIED BY 'clustrix'; master> GRANT REPLICATION SLAVE ON *.* TO 'replication'@'%';
Creating a Binary Log File
To create a binary log, issue the following command:
|CREATE BINLOG Syntax|
master> CREATE BINLOG 'binlog_name' [LOG (target1, target2, ...),] [IGNORE (target3, target4, ...),] [FORMAT='STATEMENT'|'ROW']
master> CREATE BINLOG 'mybinlog' FORMAT='ROW';
Optional attributes are:
- LOG: A list of specific tables or databases to log
- IGNORE: A list of specific tables or databases to ignore
- FORMAT: Format specifier (STATEMENT or ROW).
By default, this command creates a binary log file for the entire cluster in STATEMENT format, which is the most common format in MySQL environments. Alternatively, you can create binlogs that scope a database or a list of tables. For more information, please see the section on Binlog scope.
For most workloads, row-based replication (FORMAT ='ROW') provides better performance than statement-based replication. If you are unsure what is most appropriate for your environment, contact Clustrix Support.
If a database is set to both LOG and IGNORE, ClustrixDB will IGNORE. This deviates from MySQL, which will log and not ignore.
Setting Binary Logging Options
To configure binary logging options, issue the ALTER BINLOG command. Options for the ALTER BINLOG logfile command are as follows.
|LOG (db1, db2)||Only log updates to databases db1 and db2|
|IGNORE (db3)||Ignore updates to db3|
|ADD LOG (db4)||Log updates to db4, in addition to others|
|ADD IGNORE (db5)||Ignore updates to db5, in addition to others|
|DROP LOG (db6)||Stop logging to db6|
|LOG ALL||Log updates to all databases, as opposed to specific databases. Does not reset the IGNORE list.|
|DISABLE||Disable logging to this binlog|
|ENABLE||Enable logging to this binlog|
|RENAME bar||Rename specified binlog to "bar"|
|FORMAT='row' or 'statement'||Configure log format (row-based or statement-based)|
Displaying Binary Log Information
If only one binary log exists, you can display its filename, segment number and position by issuing the following command:
master> SHOW MASTER STATUS;
If more than one binary log exists, the log configured by the global variable master_status_binlog is displayed. If master_status_binlog is unset, an error is returned. This behavior is compatible with behavior of the MySQL mysqldump --master-data command.
master> SET GLOBAL master_status_binlog = 'foo'; master> SHOW MASTER STATUS;
To display status for all binary logs, issue the following command:
master> SHOW ALL MASTER STATUS;
To display detailed information about binary logs, issue the following command:
master> SHOW BINLOGS;
Most of this information is not directly useful, though log size can help you decide whether to trim the log.
Trimming a Binary Log
You can trim a binary log using either of the following methods:
- TRIM BINLOG command
- trim-binlog script
Trimming using the TRIM BINLOG Command
Back up your database regularly using the mysqldump --master-data command, which records the binary log filename at the start of the dump. To keep the size of the binary log under control, use this value to trim older data after it is backed up. The extent to which you trim is a matter of policy: you can choose to retain a week's history, or you might prefer to minimize disk consumption as much as possible by trimming all but the current file. To minimize the amount of space being used by your binary log, trim according to the Slave that is farthest behind in replication.
To list the files that compose the binary log, issue the following:
master> SHOW BINLOG FILES; +-----------------+-----------+-----------------------+ | File | Size | First Event Timestamp | +-----------------+-----------+-----------------------+ | eukanuba.000001 | 104857600 | 2016-01-09 19:51:08 | | eukanuba.000002 | 104857600 | 2016-01-09 20:02:09 | | eukanuba.000003 | 104857600 | 2016-01-09 22:22:27 | | eukanuba.000004 | 104857600 | 2016-01-09 22:30:37 | | eukanuba.000005 | 104857600 | 2016-01-09 22:38:11 | | eukanuba.000006 | 104857600 | 2016-01-09 22:45:44 | | eukanuba.000007 | 104857600 | 2016-01-09 22:53:03 | | eukanuba.000008 | 104857600 | 2016-01-09 23:00:44 | | eukanuba.000009 | 104857600 | 2016-01-09 23:07:46 | | eukanuba.000010 | 104857600 | 2016-01-09 23:15:00 | ...
To display current Slave locations, issue the SHOW SLAVE STATUS command, which displays status as follows:
master> SHOW SLAVE STATUS\G *************************** 1. row *************************** Slave_Name: default Slave_Status: Running Master_Host: alpo Master_Port: 3306 Master_User: root Master_Log_File: alpo Slave_Enabled: Enabled Log_File_Seq: 3383 Log_File_Pos: 58790712 Last_Error: no error Connection_Status: Connected Relay_Log_Bytes_Read: 0 Relay_Log_Current_Size: 0 Seconds_Behind_Master: 0 1 row in set (0.00 sec)
The filename argument is the filename returned by the SHOW MASTER STATUS command. The sequence number (Log_File_Seq) indicates the binary log file currently in use (it's the numeric portion of the file name). To delete old binary data without dropping the entire log, issue the following command (nnnnn represents the sequence number):
master> TRIM BINLOG 'binlog_name' BEFORE FILE 'binlog_name.nnnnn';
For example, if the clx001.000283 file is timestamped at 7:15 PM on September 28, 2016, the following sample would delete all log files before that time.
master> TRIM BINLOG 'clx001' BEFORE FILE 'clx001.000283';
Trimming Using the binlog-trim Script
You can find the binlog-trim script in /opt/clustrix/bin/binlog-trim of your system.
|-h, --help||Show this help message and exit|
|-H HOST, --host=HOST||Specifies the host|
|-P PORT, --port=PORT||SQL port, default is mysql port: 3306|
|-u USER, --user=USER||Username, default is root|
|-p PASSWD, --passwd=PASSWD||Specifies the password|
|-n NUM_FILES, --num-files=NUM_FILES||Number of files to trim a time|
|-d, --dry||Do not perform any actual trims|
|-k KEEP_DAYS, --days=KEEP_DAYS||Keep this many days of binlogs|
|-i INTERVAL, --interval=INTERVAL||Seconds between trims|
|-b BINLOG_NAME, --binlog_name=BINLOG_NAME||Name of binlog to trim; must specify if multiple binlogs exist|
|-M MAX_RUN_TIME_MINS, --max-run-time-mins=MAX_RUN_TIME_MINS||Maximum time (minutes) script may run|
|-V, --version||Indicates the version|
binlog-trim is generally deployed as a cron job on one of the nodes. An example entry to run once a day at 5:35UTC, with a retention policy of 7 days, trimming no more than 50 files at a time, with a minimum 60-second pause between trims, and operating on the binlog called clustrix-bin:
35 5 * * * root /bin/binlog-trim -H localhost -i 60 -k 7 -n 50 -b clustrix-bin 2>&1 >> /var/log/binlog-trim.log
The INTERVAL is a minimum wait between trims; there is additional logic in the script to prevent the trims from building up too much cleanup work (the logs will indicate this with 'waiting for bigc to pass trim').
Backing Up Binary Logs
Because ClustrixDB binary logs (binlogs) aren't stored as plain files, they cannot be backed up as MySQL binlogs can. For backup purposes, ClustrixDB provides the repclient utility, which copies binlogs from a ClustrixDB or MySQL system as if it were a replication slave. The repclient utility can be run on any ClustrixDB node.
To copy all of the binlogs off a ClustrixDB cluster, perform the following steps:
- To list the most recent binlog, issue the SHOW MASTER STATUS command. The command returns a filename such as clustrix-bin.001903.
- Create a directory in the /clustrix mount on a node and cd to it.
To retrieve all binlog files up to the most recent, issue the following command:
shell> node# repclient -addr 10.52.2.20 -dumpbinlog -logname clustrix-bin.000001 -end_logname clustrix-bin.001903
By default, the tool outputs decoded binlog messages to stdout. To specify an output file, specify the -dumpbinlog option. If you intend to archive the binlogs, omit -logpos, which can create gaps in the resulting binlog. By default, the utility stays connected to the master. To specify when it is to disconnect, include the -end_logname or -end_logpos option.
Valid options for the repclient command are as follows:
|-addr hostname||Database host (default: 127.0.0.1)|
|-count n||Number of messages to dump|
|-end_logname path||Ending replication log name|
|-end_logpos offset||Ending replication log position (default: EOF)|
|-help||List command options|
|-help-debug||List command options plus debugging output options|
|-logname path||Starting replication log name|
|-logpos offset||Starting replication log position (default: 4)|
|-max-packet-size bytes||Maximum packet size (default: 16777216)|
|-max-retries n||Maximum retries after an error (default: 3)|
|-no-decode-rows||Don't decode row values|
|-pass password||Database password (default: #undef)|
|-perf||Dump performance statistics|
|-perf-interval seconds||Dump performance statistics interval (default: 30)|
|-port port||Database port (default: 3306)|
|-retry-timeout seconds||Timeout in seconds for retries (default: 10)|
|-set-variable NAME= VALUE||Set a variable to the given value|
|-slave-id n||Slave ID (default: 1)|
|-testconnect||Test database connection and display status|
|-truncate||Truncate any existing files|
|-user username||Database username (default: root)|
|-verbose||Display debugging messages|
Excluding A Session from Binary Logs
To prevent a session's statements from being inserted into any binary log, set sql_log_bin to false by issuing the following command:
master> SET sql_log_bin=false;
This variable inherits the value of the identically-named global variable at the start of each session. To replicate from a ClustrixDB instance, set sql_log_bin to true.
Be careful using sql_log_bin in production. Improper use can lead to data skew between the master and the slave(s).
Dropping a Binary Log File
To stop logging to the specified binary log and drop it from the system, issue the following command:
master> DROP BINLOG binlog_name;
You cannot recover a binary log after dropping it.
The following global and session variables control binary log behavior:
|Name||Description||Default Value||Session Variable|
|binlog_checksum||Always NONE. Clustrix masters do not support generating event checksums.||NONE|
|binlog_format||Force all binlogs to log in this format, unless set to 'DEFAULT'.||DEFAULT|
|gtid_mode||Always OFF. Clustrix masters do not support generating GTID events.||OFF|
|gtid_purged||Dummy variable for compatibility. (Clustrix does not support replication with Global Transaction Identifiers.)|
|master_status_binlog||Binlog used in SHOW MASTER STATUS when used without specifying a binlog.|
|sql_log_bin||Log statements to binary logs. This variable can be set to FALSE on a per-session basis.||true|
|sync_binlog||Dummy variable for compatibility.||0|
Exercise extreme care when changing these settings. The defaults may not be ideal for your system, but they should be reasonable. The product will not warn you if you configure inadvisable settings.
The following pages describe areas that should be understood when using ClustrixDB as a Replication Master
- Online Schema Changes
- Replicating User Account Management Statements
- Using ClustrixDB with Multiple Slaves