From: Gavin Sherry <swm@alcove.com.au>
To: Kari Lavikka <tuner@bdb.fi>
Cc: pgsql-performance@postgresql.org
Subject: Re: Finding bottleneck
Date: Thu, 28 Jul 2005 19:34:50 +1000 (EST)
Message-ID: <Pine.LNX.4.58.0507281931220.2769@linuxworld.com.au> (raw)
In-Reply-To: <Pine.HPX.4.62.0507281215420.2164@purple.bdb.fi>
References: <Pine.HPX.4.62.0507281215420.2164@purple.bdb.fi>
Hi,
On Thu, 28 Jul 2005, Kari Lavikka wrote:
> ----------------->8 Relevant rows from postgresql.conf 8<-----------------
>
> shared_buffers = 15000 # min 16, at least max_connections*2, 8KB each
> work_mem = 1536 # min 64, size in KB
As an aside, I'd increase work_mem -- but it doesn't sound like that is
your problem.
> maintenance_work_mem = 32768 # min 1024, size in KB
>
> max_fsm_pages = 1000000 # min max_fsm_relations*16, 6 bytes each
> max_fsm_relations = 5000 # min 100, ~50 bytes each
>
> vacuum_cost_delay = 15 # 0-1000 milliseconds
> vacuum_cost_limit = 120 # 0-10000 credits
>
> bgwriter_percent = 2 # 0-100% of dirty buffers in each round
>
> fsync = true # turns forced synchronization on or off
> # fsync, fdatasync, open_sync, or open_datasync
> wal_buffers = 128 # min 4, 8KB each
Some benchmarking results out today suggest that wal_buffers = 1024 or
even 2048 could greatly assist you.
> commit_delay = 80000 # range 0-100000, in microseconds
> commit_siblings = 10 # range 1-1000
This may explain the fact that you've got backed up queries and idle CPU
-- I'm not certain though. What does disabling commit_delay do to your
situation?
Gavin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: pgsql-performance@postgresql.org
Cc: swm@alcove.com.au, tuner@bdb.fi
Subject: Re: Finding bottleneck
In-Reply-To: <Pine.LNX.4.58.0507281931220.2769@linuxworld.com.au>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox