Hi, Thanks for the replies. I don't think there's a leak unless it manifests itself only very occasionally. The stats now are Active 6, Idle 133, Created 0.02% (757 / 3.3M), Failed 0, max-connections 256, idle time 30000. The active count is not increasing.
The driver type in the pool definition is com.mysql.jdbc.jdbc2.optional.MysqlConnectionPoolDataSource. The connector/j version is mysql-connector-java-5.0.8-bin.jar. I'm interested that you mention a known issue with mysql - could we be seeing that possibly? Is there a fix, or somewhere I can learn about the issue? The puzzling thing is that as far as I can see from the thread lists, when these freezes happen, there is no runnable thread in any of our code at all - just the threads waiting on the lock. MySQL show processlist also claims its idle. Thanks again. Steven On 6 July 2012 14:59, Scott Ferguson <f...@caucho.com> wrote: > On 07/06/2012 11:47 AM, Knut Forkalsrud wrote: > > You may want to keep track of the DB pool statistics. Active versus maximum > allocation counts. If this is a connection leak you would notice the number > of active connection creeping upwards until it reaches max and your freezing > symptoms appear. > > > Also, which driver is this with? There's a specific synchronization problem > with the mysql ConnectionPool driver that might be a problem. > > And, for jstack, the thread that owns the lock/counter is more important > than the threads waiting. If you can find the threads that are opening the > connection (in the actual driver code), that would help track this down. > > An upgrade might help, but it would depend on what's causing the problem. > > -- Scott > > > Knut Forkalsrud > > > > On Fri, Jul 6, 2012 at 5:51 AM, Steven Cassidy <steven.cass...@gmail.com> > wrote: >> >> Hi, I'm using resin-3.1.3 on a busy site. Periodically the site >> freezes. A jstack on the server shows most threads in this state: >> >> java.lang.Thread.State: TIMED_WAITING (on object monitor) >> at java.lang.Object.wait(Native Method) >> at com.caucho.jca.ConnectionPool.create(ConnectionPool.java:897) >> - locked <0x00002aaaba60be48> (a java.util.ArrayList) >> at com.caucho.jca.ConnectionPool.allocatePool(ConnectionPool.java:779) >> at com.caucho.jca.ConnectionPool.allocate(ConnectionPool.java:740) >> at >> com.caucho.jca.ConnectionPool.allocateConnection(ConnectionPool.java:551) >> at com.caucho.sql.DataSourceImpl.getConnection(DataSourceImpl.java:65) >> at com.caucho.sql.DBPool.getConnection(DBPool.java:669) >> >> It looks like no thread ever gets notified and all wait indefinitely. >> Resin generally restarts itself about 15 minutes later if left alone. >> The database itself apppears idle during this time. >> >> I wondered if this might be a known problem and if upgrading to 3.1.12 >> would help? I don't think the pool is full. I noticed a pool timing >> issue was fixed in 3.1.10 (bug 3837) and the implementation of the >> allocate method changed, but that wasn't directly related to freezing. >> Perhaps we should just try it and see, but I figured I'd ask if this >> was noted / known behavior. >> >> Many thanks, >> Steven >> >> _______________________________________________ >> resin-interest mailing list >> resin-interest@caucho.com >> http://maillist.caucho.com/mailman/listinfo/resin-interest > > > > > _______________________________________________ > resin-interest mailing list > resin-interest@caucho.com > http://maillist.caucho.com/mailman/listinfo/resin-interest > > > > > _______________________________________________ > resin-interest mailing list > resin-interest@caucho.com > http://maillist.caucho.com/mailman/listinfo/resin-interest > _______________________________________________ resin-interest mailing list resin-interest@caucho.com http://maillist.caucho.com/mailman/listinfo/resin-interest
