mysql persistence performance - high load + deadlocks

[jduty]

ENV:

JBoss 4.2.2
JBM 1.4 SP1
MySQL 5.0.x

From within an EJB3 slsb we pump about 600,000K messages into a clustered topic daily using XA transactions, which is consumed by 4 separate clustered durable subscriptions (this occurs in a cluster of 3 jboss nodes). When we use persistent messaging we see lots of deadlocks in mysql and a high unix load. When we don't use persistent messaging the deadlocks and high load goes away.

My question is have you all come across this also and is there any configuration work around to lower the load and deadlocks when using persistent messaging?

Thanks,
~Jonathan

[clebert.suconic]

You might find a way to decrease those deadlocks using some MySQL configuration, maybe increasing locks or some other configure.

We just retry those statements, and that will not be causing you any problem.

When you don't using persistent messages, obviously MySQL is not used.

[timfox]

Could you try upgrading to 1.4.0.SP3?

Also, on mysql, if you go to a mysql console and type "show innodb status"

http://dev.mysql.com/doc/refman/5.0/en/innodb-monitor.html

It will tell you what's causing the deadlocks.

Based on that you may be able to add some indexes to prevent the deadlock.