From: Tatsuo Ishii <t-ishii@sra.co.jp>
To: pathat@comcast.net
Cc: pgsql-performance@postgresql.org
Subject: Re: determining max_fsm_pages
Date: Fri, 29 Oct 2004 23:31:51 +0900 (JST)
Message-ID: <20041029.233151.78703659.t-ishii@sra.co.jp> (raw)
In-Reply-To: <41824E05.8000903@comcast.net>
References: <41824E05.8000903@comcast.net>
> Pg: 7.4.5
> 8G ram
> 200G RAID5
>
> I have my fsm set as such:
> max_fsm_pages = 300000 # min max_fsm_relations*16, 6 bytes each
> max_fsm_relations = 500 # min 100, ~50 bytes each
>
>
> I just did a vacuum full on one table and saw this result:
> INFO: analyzing "cdm.cdm_fed_agg_purch"
> INFO: "cdm_fed_agg_purch": 667815 pages, 3000 rows sampled, 52089570
> estimated total rows
>
>
> My question is this: I have about 8 databases running on this server.
> When I do a vacuum full on each of these databases, there is a INFO
> section that I assume is the total pages used for that database. Should
> add ALL these individual pages together and pad the total and use this
> as my new max_fsm_pages? Should I do the same thing with max_fsm_relations?
I think that's too much and too big FSM affects performance in my
opinion. The easiest way to calculate appropreate FSM size is doing
vacuumdb -a -v and watching the message. At the very end, you would
see something like:
INFO: free space map: 13 relations, 1447 pages stored; 1808 total pages needed
DETAIL: Allocated FSM size: 100 relations + 1600 pages = 19 kB shared memory.
In this case 1808 is the minimum FSM size. Of course this number would
change depending on the frequency of VACUUM. Therefore you need some
room for the FSM size.
--
Tatsuo Ishii
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: t-ishii@sra.co.jp, pathat@comcast.net
Subject: Re: determining max_fsm_pages
In-Reply-To: <20041029.233151.78703659.t-ishii@sra.co.jp>
* 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