Janning Vygen <vygen@gmx.de> writes:
> I recently reported this problem and i would like to help solving it. But how
> can i build a self-contained test-case? It just happens sometimes under load.
I didn't say you need to make it 100% reproducible; you just have to
make a case that someone else can run that will eventually produce the
error. The sort of poking and prying that will need to happen to debug
it will involve things you do not want done to your production database,
therefore we need to be able to make the error happen in a test setup.
You probably need to create a client script that will issue multiple
parallel queries that are similar to what your regular application does.
See for instance this discussion:
http://archives.postgresql.org/pgsql-hackers/2005-05/msg00613.php
If you're handy with C, pgbench might be a useful starting point.
But a script in perl python or tcl will be fine too.
regards, tom lane