On Tue, 2007-04-03 at 15:02 -0400, Tom Lane wrote:
> Tim Goodaire <tgoodair(at)ca(dot)afilias(dot)info> writes:
> > While going through some log files, we noticed that some of the log entries
> > are "garbled". For example:
> > 2007-03-27 01:19:44.139 UTC  oxrsa aepp xx.xx.xx.xx LOG:
> > duratio2007-03-n: 3751.27 01:19801 ms :44.139 statemenUTC 
> > oxrt: EXECUsor
> > g aTE <unnaepp 10.4med> [P0.136.10REPARE: 8 LOG: select
> > durationname fro: 3866.1m epp_do88 ms smain_dz_tatementnames wh:
> > EXECUTere nameE <unnam = $1]
> > 2007-03-27 01:19:49.213 UTC  oxrsa aepp xx.xx.xx.xx LOG:
> > duration: 4510.129 ms statement: EXECUTE <unnam2007-03-ed> [PR27
> > 01:19EPARE: :49
> > .213 select nUTC [294ame from930] oxr b_domsa aain_dz_nb
> > 10.4ames whe0.136.10re name 7 LOG: = $1]
> > We're seeing this in postgresql 8.1.8 on AIX 5.3. Also, even though we're
> > using the same binaries for a bunch of clusters, we've only seen this issue
> > on one of them.
> It looks like you've got messages from different backends being
> interleaved, which is surely not impossible but normally it only happens
> line-by-line at worst. This must have something to do with the
> buffering behavior on stderr. Perhaps the difference is caused
> by a difference in where the postmaster's stderr originally pointed
> --- was this postmaster started in a different fashion than the others?
No. Standard init script was used to start all clusters.
Brad Nicholson 416-673-4106
Database Administrator, Afilias Canada Corp.
In response to
pgsql-hackers by date
|Next:||From: Tom Lane||Date: 2007-04-03 20:51:33|
|Subject: Re: "Garbled" postgres logs |
|Previous:||From: Andrew Dunstan||Date: 2007-04-03 19:35:25|
|Subject: Re: "Garbled" postgres logs|