Upgrade mysql from 5.0 to 5.5 in master-master configuration leads to broken replication

Question

We created two new mysql servers running 5.5. Our old server were 5.0.

The two servers are set up with master-master replication - this was the case for the two 5.0 servers which had no issues.

We have some long running select queries (40-50 sec), and when they are run replication breaks - we have to skip a couple of records to get replication restarted. This is consistent and repeatable. I suspect that some tables are getting updated on each server, but replication is paused trying to update a locked table - and when it is done, each server is trying to send data to the other and replication is broken.

I am seeing a lot of "waiting for table level lock" on select statements when this long query is running.

Is there a setting I've missed to prevent selects from locking tables? Is there some other cause, or any way to fix this?

Answer 1

Since all the data is MyISAM, you should expect table locks rather frequently since INSERT , UPDATE, and DELETE command issues a full table lock.

If you have a lot of tables that use auto increment keys, you should add the following

Server1

[mysqld]
auto_increment_increment = 5
auto_increment_offset = 1

Server2

[mysqld]
auto_increment_increment = 5
auto_increment_offset = 2

This will prevent PRIMARY KEY collisions with autoincrement columns

I have other suggestions in my post I have been tasked with Mysql Master-Master replication?