To: "Reuven M. Lerner" <reuven@lerner.co.il>
cc: pgsql-performance@postgresql.org
Subject: Re: Very long deletion time on a 200 GB database 
In-reply-to: <4F45FB55.1020600@lerner.co.il> 
References: <4F45FB55.1020600@lerner.co.il>
Comments: In-reply-to "Reuven M. Lerner" <reuven@lerner.co.il>
	message dated "Thu, 23 Feb 2012 10:39:49 +0200"
Date: Thu, 23 Feb 2012 14:04:34 -0500
Message-ID: <3806.1330023874@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

I just reread the original post and noted this:

"Reuven M. Lerner" <reuven@lerner.co.il> writes:
> (1) I tried to write this as a join, rather than a subselect.  But B has 
> an oid column that points to large objects, and on which we have a rule 
> that removes the associated large object when a row in B is removed. 

A rule?  Really?  That's probably bad enough in itself, but when you
write an overcomplicated join delete query, I bet the resulting plan
is spectacularly bad.  Have you looked at the EXPLAIN output for this?

I'd strongly recommend getting rid of the rule in favor of a trigger.
Also, as already noted, the extra join inside the IN sub-select is
probably hurting far more than it helps.

> (3) There are some foreign-key constraints on the B table.

If those are FK references *to* the B table, make sure the other end
(the referencing column) is indexed.  Postgres doesn't require an index
on a referencing column, but deletes in the referenced table will suck
if you haven't got one.

I don't think any of the fancy stuff being discussed in the thread is
worth worrying about until you've got these basic issues dealt with.

			regards, tom lane