"vinita bansal" <sagivini@hotmail.com> writes:
> Do these settings seem fine or I am making some mistake. These settings when
> used with Postgres 7.4 gave me good results but they don't seem to work with
> Postgres 8.0. Am I missing out on something??
Define "don't seem to work", please.
Offhand the only thing that comes to mind is that the GEQO parameters
changed meaning a bit in 8.0; you shouldn't blindly set geqo_effort the
same as you used to. But with geqo_threshold so high it may not matter
... do you even have any queries with more than 25 tables?
Also, it's just plain bizarre to have work_mem larger than
maintenance_work_mem; I cannot imagine any sane reason to do that.
Are you sure that work_mem = 128M is really a safe setting on your
hardware? If you have a good number of clients all concurrently
doing complicated queries, you could easily find yourself pushed
into swapping.
I concur with Mike's thought that shared_buffers = 100000 is on the high
side. wal_buffers = 1000 strikes me as a waste of RAM too; do you have
any proof that it helps to set it above 10?
Do you really think it's a good idea to disable the bgwriter?
Especially with such a long intercheckpoint time?
regards, tom lane
