Received: from localhost (unknown [200.46.204.191]) by postgresql.org (Postfix) with ESMTP id 6775E2E00E5 for ; Mon, 28 Jan 2008 21:47:16 -0400 (AST) Received: from postgresql.org ([200.46.204.71]) by localhost (mx1.hub.org [200.46.204.191]) (amavisd-maia, port 10024) with ESMTP id 23892-01 for ; Mon, 28 Jan 2008 21:47:05 -0400 (AST) Received: from bay0-omc3-s2.bay0.hotmail.com (bay0-omc3-s2.bay0.hotmail.com [65.54.246.202]) by postgresql.org (Postfix) with ESMTP id A5AC42E0048 for ; Mon, 28 Jan 2008 21:47:04 -0400 (AST) Received: from BAY116-W2 ([64.4.38.102]) by bay0-omc3-s2.bay0.hotmail.com with Microsoft SMTPSVC(6.0.3790.3959); Mon, 28 Jan 2008 17:47:03 -0800 Message-ID: Content-Type: multipart/alternative; boundary="_0e10eba0-4f01-491f-beeb-7d5c80be5893_" X-Originating-IP: [204.132.58.105] From: dx k9 To: Tom Lane CC: posgres support Subject: Re: Zen and art if WAL and dump file mainteance Date: Mon, 28 Jan 2008 20:47:02 -0500 Importance: Normal In-Reply-To: <7479.1201555625@sss.pgh.pa.us> References: <7479.1201555625@sss.pgh.pa.us> MIME-Version: 1.0 X-OriginalArrivalTime: 29 Jan 2008 01:47:03.0214 (UTC) FILETIME=[D8906CE0:01C86218] X-Virus-Scanned: Maia Mailguard 1.0.1 X-Archive-Number: 200801/273 X-Sequence-Number: 28047 --_0e10eba0-4f01-491f-beeb-7d5c80be5893_ Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable I notice in 'pg_data' that a file will stay in pg_xlog until it passes 16 M= B, then it goes to the archive pretty much right away. So, when it's busy = you get a lot of files swapping into the archive , but when it's not you ca= n get one file taking an hour or more to get up to 16 MB before it swaps ov= er to the archive. In reading the documentation "every 5 minutes or 36 log= file segments which ever is first" - - 74 files in my case. But I notice= d when it's slow it's more it can be longer than 5 minutes. I'm thinking t= he "5 minutes or 36 log segments" rule applies to just after files hit 16 M= B in size, until then, they are building up. =20 Does that all sound right?> To: bitsandbytes88@hotmail.com> CC: pgsql-admin= @postgresql.org> Subject: Re: [ADMIN] Zen and art if WAL and dump file main= teance > Date: Mon, 28 Jan 2008 16:27:05 -0500> From: tgl@sss.pgh.pa.us> > = dx k9 writes:> > We are trying to figure out w= hy maintenance on A (zipping up & removing files) would have an affect on p= roduction processing (killed it) on pgdata - B.> > Sounds like an OOM kill = problem --- do you have memory overcommit> disabled on that machine?> > reg= ards, tom lane _________________________________________________________________ Climb to the top of the charts!=A0Play the word scramble challenge with sta= r power. http://club.live.com/star_shuffle.aspx?icid=3Dstarshuffle_wlmailtextlink_ja= n= --_0e10eba0-4f01-491f-beeb-7d5c80be5893_ Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable I notice in 'pg_data' that a file will stay in pg_xlog until it passes 16 M= B, then it goes to the archive pretty much right away.  So, when it's = busy you get a lot of files swapping into the archive , but when it's not y= ou can get one file taking an hour or more to get up to 16 MB before it swa= ps over to the archive.  In reading the documentation "every 5 minutes= or 36 log file segments which ever is first" - -  74 files in my case= .  But I noticed when it's slow it's more it can be longer than 5 minu= tes.  I'm thinking the "5 minutes or 36 log segments" rule applies to = just after files hit 16 MB in size, until then, they are building up.
 
Does that all sound right?

> To: bitsandbytes88@hotmail.com
&g= t; CC: pgsql-admin@postgresql.org
> Subject: Re: [ADMIN] Zen and art = if WAL and dump file mainteance
> Date: Mon, 28 Jan 2008 16:27:05 -0= 500
> From: tgl@sss.pgh.pa.us
>
> dx k9 <bitsandbytes= 88@hotmail.com> writes:
> > We are trying to figure out why mai= ntenance on A (zipping up & removing files) would have an affect on pro= duction processing (killed it) on pgdata - B.
>
> Sounds like = an OOM kill problem --- do you have memory overcommit
> disabled on t= hat machine?
>
> regards, tom lane



Climb to= the top of the charts!=A0Play the word scramble challenge with star power.= Play now! = --_0e10eba0-4f01-491f-beeb-7d5c80be5893_--