From: | Greg Smith <greg(at)2ndQuadrant(dot)com> |
---|---|
To: | pgsql-performance(at)postgresql(dot)org |
Subject: | Re: How to configure a read-only database server? |
Date: | 2011-04-19 14:30:48 |
Message-ID: | 4DAD9C98.5090707@2ndQuadrant.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
On 04/18/2011 06:08 PM, Stefan Keller wrote:
> * What about wal_level and archive_mode?
>
Presumably you don't care about either of these. wal_level=minimal,
archive_mode=off.
The other non-obvious thing you should do in this situation is do all
the database maintenance in one big run after the data is loaded,
something like:
VACUUM FREEZE ANALYZE;
Otherwise you will still have some trickle of write-activity going on,
not always efficiently, despite being in read-only mode. It's because
of what's referred to as Hint Bits:
http://wiki.postgresql.org/wiki/Hint_Bits
VACUUMing everything will clean those us, and freezing everything makes
sure there's no old transactions to concerned about that might kick off
anti-wraparound autovacuum.
The only other thing you probably want to do is set checkpoint_segments
to a big number. Shouldn't matter normally, but when doing this freeze
operation it will help that execute quickly. You want a lower
maintenance_work_mem on a read-only system than the master too, possibly
a higher shared_buffers as well. It's all pretty subtle beyond the big
parameters you already identified.
--
Greg Smith 2ndQuadrant US greg(at)2ndQuadrant(dot)com Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.us
"PostgreSQL 9.0 High Performance": http://www.2ndQuadrant.com/books
From | Date | Subject | |
---|---|---|---|
Next Message | Josh Berkus | 2011-04-20 01:50:40 | Re: Shouldn't we have a way to avoid "risky" plans? |
Previous Message | Robert Haas | 2011-04-19 14:29:13 | Re: Shouldn't we have a way to avoid "risky" plans? |