On tor, 2012-01-05 at 10:04 +0000, Benedikt Grundmann wrote:
> We have recently upgrade two of our biggest postgres databases
> to new hardware and minor version number bump (8.4.5 -> 8.4.9).
>
> We are experiencing a big performance regression in some queries.
> In those cases the planner seems to choose a nested loop index
> scan instead of hashing the index once and then joining.
There was a planner regression introduced in version 8.4.8, which was
thought to be fixed in 8.4.9. Maybe you got caught by that. See
Message-Id: <760C0206-B5F4-4DC6-9296-B7A730B7F403@silentmedia.com>
for some information. Check if your queries match that pattern.