On Jan 25, 2006, at 5:12 PM, Oliver Jowett wrote:
> David Hustace wrote:
>
>> Conclusion java is doing a read on this socket, and postmaster is
>> doing a recv.
>
> Well, that's pretty weird.
>
> Can you reproduce this with a recent driver and loglevel=2 (append
> it as a URL parameter to your JDBC URL) and send me the resulting
> debug output offlist? It sounds like it might be hard to trigger
> though :(
Sure. I already have that enabled. Finding the output is tricky.
We're currently using:
url="jdbc:postgresql://localhost:5432/opennms?
protocolVersion=2&loglevel=2"
Version=2 is the latest attempt in debugging.
> I doubt that using DataSource will make any difference. From the
> stack traces you showed, this does not look like a Java-level
> deadlock at all.
You are right, we've been trying to eliminate every possible
solution. Also, for future reference, can you tell me if the call to
PGSimpleDataSource.getConnection() is thread safe?
> (I noticed that your two stack traces seem to be talking to
> different server versions, though.. what's up with that?)
Not quite sure what you mean by "different server versions". If you
are referring to PostgreSQL server versions, yes, we've been
upgrading and trying different versions to try and fix the problem.
-David
David Hustace
The OpenNMS Group, Inc.
Main : +1 919 545 2553 Fax: +1 503-961-7746
Key Fingerprint: 8EC1 F319 2C1D 3197 AAB9 580F FACD 1D19 175E D903