On Wed, 3 Dec 2008, hubert depesz lubaczewski wrote:
Check bug report from 2008-11-28, by Grzegorz Jaskiewicz: query failed, not enough memory on 8.3.5 http://archives.postgresql.org/pgsql-bugs/2008-11/msg00180.php
Thanks, that does explain everything. So workmem is not a hard limit on the amount of memory used per hash. Once the planner has committed to using a hash (and it can only use estimates to work out whether it will fit in workmem), then the execution will blindly go ahead and try and fit absolutely everything in a hash in memory even if it doesn't fit in workmem.
I agree it would be nice to fix this, but I'm not sure how at the moment. Matthew -- To most people, solutions mean finding the answers. But to chemists, solutions are things that are still all mixed up. -- Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
