Mysql master-master replication auto flush old logs

Question

I have mysql-5.1.41 master-master setup running on ubuntu machines and below is the configuration of a master:

skip-host-cache
skip-name-resolve
event_scheduler = ON
max_connections = 500
max_connect_errors = 1000

server-id = 10
replicate-same-server-id = 0
auto-increment-increment = 10
auto-increment-offset = 1

master-host = 192.168.1.106
master-user = repli
master-password = secret
master-connect-retry = 60

binlog-format = MIXED
binlog-ignore-db = information_schema
binlog-ignore-db = lb1

log-bin = /var/log/mysql/mysql-bin
log-bin-index = /var/log/mysql/bin-log.index

log-slave-updates
report-host = 192.168.1.105
replicate-ignore-db = information_schema
replicate-ignore-db = lb2

relay-log = /var/log/mysql/relay.log
relay-log-index = /var/log/mysql/relay-log.index

The log files seems to be growing everyday. Does anybody know what parameter I should include in above configuration to automatically remove/flush old unnecessasry logs with a decent value causing no damage to sync?.

Answer 1

To rotate out logs more than 7 days old, add this to my.cnf

[mysqld]
expire-logs-days=7

then restart mysql.

To perform this manually, run this command:

mysql> PURGE BINARY LOGS BEFORE DATE(NOW() - INTERVAL 7 DAY) + INTERVAL 0 SECOND;

This will erase binary logs back to Midnight 7 days ago.

WARNING !!!

Do not erase binary logs from the OS level because it disturbs mysqld's internal functionality for rotating binary logs.

UPDATE 2012-01-24 13:20 EDT

Make sure you do not purge binary logs that replication is using.

Here is an example how you can safely erase binary logs manually...

For a Master-Master setup M1 and M2

Go to M2 and run this:

mysql> SHOW SLAVE STATUS\G

You will see two lines that show binary log names

• Master_Log_File
• Relay_Master_Log_File

Master_Log_File represents the current binary log on M1 that was last read on M2.

Relay_Master_Log_File represents the current binary log on M1 that was last executed on M2.

Let's suppose that M2 has this:

mysql> show slave status\G
*************************** 1. row ***************************
               Slave_IO_State: Waiting for master to send event
                  Master_Host: 10.64.176.205
                  Master_User: replicant
                  Master_Port: 3306
                Connect_Retry: 60
              Master_Log_File: mysql-bin.000131
          Read_Master_Log_Pos: 570079419
               Relay_Log_File: relay-bin.003496
                Relay_Log_Pos: 251
        Relay_Master_Log_File: mysql-bin.000131
             Slave_IO_Running: Yes
            Slave_SQL_Running: Yes
              Replicate_Do_DB:
          Replicate_Ignore_DB:
           Replicate_Do_Table:
       Replicate_Ignore_Table:
      Replicate_Wild_Do_Table:
  Replicate_Wild_Ignore_Table:
                   Last_Errno: 0
                   Last_Error:
                 Skip_Counter: 0
          Exec_Master_Log_Pos: 570079419
              Relay_Log_Space: 545
              Until_Condition: None
               Until_Log_File:
                Until_Log_Pos: 0
           Master_SSL_Allowed: No
           Master_SSL_CA_File:
           Master_SSL_CA_Path:
              Master_SSL_Cert:
            Master_SSL_Cipher:
               Master_SSL_Key:
        Seconds_Behind_Master: 0
Master_SSL_Verify_Server_Cert: No
                Last_IO_Errno: 0
                Last_IO_Error:
               Last_SQL_Errno: 0
               Last_SQL_Error:
1 row in set (0.00 sec)

mysql>

ON M2, Relay_Master_Log_File says mysql-bin.000131. So on M1, you can run this:

mysql> PURGE BINARY LOGS TO 'mysql-bin.000131';

Since Master_Log_File and Relay_Master_Log_File are the same, why not use Master_Log_File?

REASON #1 : SQL Error in the Slave

When an SQL Error happens in M2 on the SQL Thread, Replication stops processing SQL from the Relay Logs. The IO Thread continues downloading new SQL entries from M1 and appending to M2's Relay Logs.

REASON #2 : Long Running SQL Query in the Slave

If a query comes along, such as UPDATE that updates every row in a table and it takes 5 minutes. During those 5 minutes, the IO Thread continues downloading new SQL entries from M1 and appending to M2's Relay Logs.

Given both reasons, there are some occassion where Relay_Master_Log_File and Master_Log_File are different. In that instance, always use Relay_Master_Log_File. In light of this, there is another reason:

REASON #3 : Log Corruption in Slave or Master

There are rare times when bad network transmission will produce corrupt relay logs. When you do SHOW SLAVE STATUS\G, an error message will explain the the binary log may be corrupt. You should start firsdt with the relay logs.

To clear them out and start with fresh set, do SHOW SLAVE STATUS\G, get the Relay_Master_Log_File (RMLF), the Exec_Master_Log_Pos (EMLP), and run this on M2

STOP SLAVE;
CHANGE MASTER TO master_log_file='RMLF',master_log_pos=EMLP;
START SLAVE;

The CHANGE MASTER TO command erases all relay logs and starts downloading from scratch.

Once you START SLAVE;, if you get the corrupt error log message again, then network transmission was not the issue. The binary log on M1 may just be corrupt after all.

You must go to M1, make copy of the the suspected log, run mysqlbinlog against that copied binary log and redirect to a text file. Read the text file. If the text files contains gibberish or indiscernable characters, then you must perform a full sync of the slave.