Idiotically I rebooted a machine that is used as a MySQL slave without running STOP SLAVE, FLUSH TABLES first.

I had thought MySQL automatically handles all of this during a machine reboot, but apparently it doesn't, at least not in the configuration I was using, because the SLAVE doesn't start back up.

mysqld does start but there is an error message in the logs indicating the slave portion stops due to duplicate primary key issues. Meaning it is trying to insert data that has already been added.

Here are the errors generated in the mysql log:

120104 11:07:54 [Warning] Slave: Duplicate entry '94459' for key 'PRIMARY' Error_code: 1062
120104 11:07:54 [ERROR] Error running query, slave SQL thread aborted.
    Fix the problem, and restart the slave SQL thread with "SLAVE START".
    We stopped at log 'mysql1-bin.000362' position 3384732

How do I determine where in the master binary log the process needs to start from to run a CHANGE MASTER statement? I know I could potentially, skip log entries using the sql_slave_skip_counter but without knowing how many to skip I would need to go one by one, and that could take all day.


1 Answer 1


The Primary thing you need to do was left in a comment:

Follow the suggestion in Duplicate entry in MySQL slave:

SET GLOBAL sql_slave_skip_counter=1;

However, here is how you can start off at the better position with the CHANGE MASTER TO command.

Let's take a look at a sample SHOW SLAVE STATUS\G

               Slave_IO_State: Waiting for master to send event
                  Master_User: replicant
                  Master_Port: 3306
                Connect_Retry: 60
              Master_Log_File: mysql-bin.001527
          Read_Master_Log_Pos: 554619670
               Relay_Log_File: relay-bin.004561
                Relay_Log_Pos: 554619815
        Relay_Master_Log_File: mysql-bin.001527
             Slave_IO_Running: Yes
            Slave_SQL_Running: Yes
          Replicate_Ignore_DB: phpmyadmin
                   Last_Errno: 0
                 Skip_Counter: 0
          Exec_Master_Log_Pos: 554619670
              Relay_Log_Space: 554620007
              Until_Condition: None
                Until_Log_Pos: 0
           Master_SSL_Allowed: No
        Seconds_Behind_Master: 0
Master_SSL_Verify_Server_Cert: No
                Last_IO_Errno: 0
               Last_SQL_Errno: 0

Please take note of Relay_Master_Log_File and Exec_Master_Log_Pos. These represent the log file and the position of the last SQL statement that was completed on the Master that is next in line to be executed on the Slave.

So, the CHANGE MASTER TO in this case would be:

CHANGE MASTER TO master_log_file='mysql-bin.001527',master_log_pos=554619670;

For your particular case, here is what you need to do on the Slave Server:

Step 01) Start mysql up with replication disabled at startup

$ service mysql restart --skip-slave-start

Step 02) Login to mysql and show the slave status:


Step 03) Get Relay_Master_Log_File and Exec_Master_Log_Pos from SHOW SLAVE STATUS\G

Step 04) Run CHANGE MASTER TO command using Relay_Master_Log_File and Exec_Master_Log_Pos

This will erase any relay logs collected and start collecting with a fresh, empty relay log.

Step 05) mysql> START SLAVE;

Step 06) mysql> SHOW SLAVE STATUS\G repeatedly to watch Seconds_Behind_Master go to 0

Step 07) If replication breaks due to Error 1062 (Duplicate Key), now you can implement @DTest's suggestion. Then, goto Step 06. Repeat this until Seconds_Behind_Master go to 0.

We stopped at log 'mysql1-bin.000362' position 3384732.

That's what you use:

CHANGE MASTER TO master_log_file='mysql1-bin.000362'. master_log_pos=3384732;

If you ran RESET SLAVE on the Slave, do not despair. Just run the full syntax version of the CHANGE MASTER TO command:

MASTER_HOST='IP Address of Master',
MASTER_USER='Replication Username',
MASTER_PASSWORD='Replication Password',
  • Thanks for all the help. I ended up using the Change Master command using a MASTER_LOG_POS that I knew would work, which is okay for now, but I will probably need to sync up the databases in the future.
    – Aglystas
    Jan 4, 2012 at 19:04
  • If you want to sync the databases without a reload the full dataset, look into the tools pt-table-checksum and pt-table-sync. I wrote a script to automate that process using MAATKIT's mk-table-checksum and mk-table-sync : dba.stackexchange.com/a/8808/877 Jan 4, 2012 at 19:17
  • @RolandoMySQLDBA: As per my understanding, we can't start replication from the pointer where error log is showing as mysql has executed last point due to this error is related with duplicate key...here we have to skip this position as it has been executed....yes here Aglystas's concern is how can skip blindly so first we have to read binary logs and make sure that statement before this position and of this position has been executed on slave and the just next position sql is not updated till now, so we can skip 1 pointer here or use that position in change master to... Aug 18, 2015 at 11:52
  • Also there should not any need of mysql service restart. Aug 18, 2015 at 11:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.