Josh Berkus <josh@agliodbs.com> writes:
>> I'm working on cleaning up postgresql.conf and pg_settings for the
>> ...
>> seq_scan_cost: this is independant of all of the other _costs.
So? All the other costs are independent of it, too. I don't understand
what problem you have with it.
>> (change requires restart): this phrase appears over 20 times in the
>> notes. This is enough times to be really repetitive and take up a lot
>> of scrolling space, while not actually covering all startup-time
>> parameters. We should either (a) remove all such notes and rely on
>> docs, or (b) make an annotation symbol (e.g. *R) and mark 100% of them.
That was put in deliberately not long ago, so I doubt (a) will pass.
(b) seems fine to me.
>> transaction_isolation and transaction_read_only appear more than once in
>> the pg_settings pseudo_table.
Not for me.
> # work_mem = ( RAM * 0.5 ) / max_connections, or less
That seems guaranteed to drive people into swap hell, unless they
execute only trivial queries.
> # wal_buffers = 1MB
Is there really evidence in favor of such a high setting for this,
either? (I expect the walwriter in the async-commit patch will change
the landscape here, btw.)
> # max_fsm_pages = expected database size * 0.1
This might be too small.
> # checkpoint_segments = 8 to 16 if you have the disk space (0.3 to 0.6 GB)
This seems definitely too small --- for write-intensive databases I like
to set it to 30 or so, which should eat about a GB if I did the
arithmetic right.
> #explain_pretty_print = on
Putting this under "planner options" is wrong and illogical.
The file seems to be missing the effects of some recently committed
patches, eg, bgwriter_all_percent shouldn't be there anymore.
regards, tom lane
