To: Greg Smith <greg@2ndquadrant.com>
cc: Tomas Vondra <tv@fuzzy.cz>, pgsql-performance@postgresql.org
Subject: Re: since when has pg_stat_user_indexes.idx_scan been counting? 
In-reply-to: <4DCC3DD9.4000402@2ndquadrant.com> 
References: <BANLkTikhNLPhCjE819=RFG-+CB6UV4GbUA@mail.gmail.com>
	<4DCC24BD.9040900@fuzzy.cz> <4DCC3DD9.4000402@2ndquadrant.com>
Comments: In-reply-to Greg Smith <greg@2ndquadrant.com>
	message dated "Thu, 12 May 2011 16:06:49 -0400"
Date: Thu, 12 May 2011 16:16:18 -0400
Message-ID: <8490.1305231378@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Greg Smith <greg@2ndquadrant.com> writes:
> You're saying to watch out for (3); I think that's not usually the case, 
> but that's a fair thing to warn about.  Even in that case, though, it 
> may still be worth dropping the index.  Year-end processes are not 
> usually very sensitive to whether they take a little or a long time to 
> execute.  But you will be paying to maintain the index every day while 
> it is there.

Yeah.  Another idea worth considering is to have the year-end processing
build the index it wants, use it, drop it.  It seems unlikely that it's
worth maintaining an index year-round for such infrequent usage.

			regards, tom lane