Reply-To: <gnanam@zoniac.com>
From: "Gnanakumar" <gnanam@zoniac.com>
To: "'Scott Marlowe'" <scott.marlowe@gmail.com>
Cc: <pgsql-admin@postgresql.org>
References: <002301cb306f$c8967dc0$59c37940$@com>
	<AANLkTin6oBtd9gS79mRmNiW7iVa7Hfv63UYvXWvYTUWr@mail.gmail.com>
In-Reply-To: <AANLkTin6oBtd9gS79mRmNiW7iVa7Hfv63UYvXWvYTUWr@mail.gmail.com>
Subject: Re: Unexpected disk space growth controlling measures
Date: Sat, 31 Jul 2010 12:30:44 +0530
Organization: Zoniac Software Private Limited
Message-ID: <005101cb307e$1b12f1b0$5138d510$@com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: 7bit
Thread-Index: Acswem1qGT+FN8yCQeeW0RcbnlJ0gwAAYONg
Content-Language: en-us

Hi Scott,

Thanks for your valuable suggestion.  Our production server is running
PostgreSQL 8.2.3 on Red Hat 4.1.1-52.

>  Esp in 8.3 and before where blowing out the free space map is quite easy
to do and it can go unnoticed for some time.

As you pointed out rightly, recently in our Production server there is a
warning showing up in VACUUM ANALYZE VERBOSE.
WARNING:  relation "public.oneofmytable" contains more than "max_fsm_pages"
pages with useful free space
HINT:  Consider compacting this relation or increasing the configuration
parameter "max_fsm_pages".

Currently, there are 439 tables and 743 indexes, adding up 1182 relations.
What would you recommend me to set the value for "max_fsm_pages" and
"max_fsm_relations" parameters?

> Where I work we use about 2.5M entries but have our fsm set to 10M so we
don't have to worry about blowing it out overnight or anything.
If you don't mind, can you make me clear here.  Is 2.5M entries in a single
table or is it something else?

> Lastly, make sure your IO subsystem can keep up.  If you're on the hairy
edge, then vacuum may never be able to keep up.
How do I confirm/make sure that IO subsystem can keep up?