From: "Ed L." <pgsql@bluepolka.net>
To: "George Pavlov" <gpavlov@mynewplace.com>
Subject: Re: query log corrupted-looking entries
Date: Fri, 1 Jun 2007 15:20:19 -0600
User-Agent: KMail/1.9.5
Cc: pgsql-general@postgresql.org
References: <29860.1161202644@sss.pgh.pa.us>
	<200705291119.55567.pgsql@bluepolka.net>
	<8C5B026B51B6854CBE88121DBF097A86CDD9A4@ehost010-33.exch010.intermedia.net>
In-Reply-To: 
 <8C5B026B51B6854CBE88121DBF097A86CDD9A4@ehost010-33.exch010.intermedia.net>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200706011520.19227.pgsql@bluepolka.net>

On Friday 01 June 2007 3:09 pm, George Pavlov wrote:
> On 5/29/2007 10:19 AM, Ed L. wrote:
> > On Wednesday 23 May 2007 1:04 pm, George Pavlov wrote:
> > FWIW, I've also been seeing this sort of query log
> > corruption for as long as I can remember, 7.1 through 8.2,
> > HPUX (parisc, ia64), Linux on intel, amd...
>
> Do you have any tricks for dealing with the problem from a
> query analyzer perspective? That is, if you use something like
> pgfouine do you have any quick and easy way to remove those
> lines (and the affected lines around them)? Or someway to
> "fix" the corrupted lines? I'd say that lately ~40% of my
> daily query logs are suffering from this problem making query
> analysis very hard.

Not really.  I filter by perl regular expressions, but I wouldn't 
say that is reliable.  However, our query volumes are high 
enough that losing a few here and there is no big deal in 
analysis.  I long ago realized that reliably replaying query 
logs was infeasible due in part to this issue.  Regex filtering 
is problematic and error-prone (for example, how do you reliably 
tell the difference between log timestamp and DB data?).

Perhaps others have better answers.

Ed