[Deploying Sakai] FW: 2.6 Production Issues

Charles Hedrick hedrick at rutgers.edu
Thu Sep 3 06:38:16 PDT 2009


I looked around a bit more. It doesn't look like a connection leak in  
the normal sense, which is code taking a connection out of the pool  
and not returning it. dbcp has been told to allocate 300 connections.  
It thinks 25 are in use, but netstat shows only 275 TCP connections.  
So it looks like some connections have closed on the server side. I  
don't think normal Sakai code would be closing a connection, since  
connection.close() for a pooled connection just returns it to the  
pool. So far I don't see a cause. However the nice thing about it is  
that if we get dangerously low I can just increase the number of  
connections without needing a restart.

On Sep 2, 2009, at 9:42 PM, Charles Hedrick wrote:

> There does seem to be a database connection leak, which looks new to  
> 2.6 (and/or new to something we've added since last semester). We  
> currently have 25-50 open connections from each front end. They are  
> almost certainly not in current use. If we can't find the problem  
> that will force periodic JVM restarts.



More information about the production mailing list