Subject: Re: PITR wrm-standby startup fails
From: Simon Riggs <simon@2ndQuadrant.com>
To: bob_lunney@yahoo.com
Cc: pgsql-admin@postgresql.org
In-Reply-To: <553812.74940.qm@web39702.mail.mud.yahoo.com>
References: <553812.74940.qm@web39702.mail.mud.yahoo.com>
Content-Type: text/plain
Date: Tue, 26 Aug 2008 11:24:21 +0100
Message-Id: <1219746261.5343.1292.camel@ebony.2ndQuadrant>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit


On Sun, 2008-08-24 at 22:17 -0700, Bob Lunney wrote:
> I'm trying to vet the PITR/warm-standby process so set up a primary
> and secondary server on two different Windows machines.  (Yeah, I
> know, but I don't have a choice in the matter.)
> 
> The primary works fine and copies its WAL files to the archive
> directory.  As long as the secondary is in recovery mode life is good.
> When I introduce the trigger file, however, to kick the secondary out
> of recovery mode it bombs, complaining that its looking for the next
> WAL file (which is never going to arrive, especially not in the local
> pg_xlog directory), then gives me an "invalid parameter" error.  The
> relevent part of the secondary's log file is here:
> 
> 2008-08-24 23:02:56 CDT LOG:  restored log file
> "000000010000000400000088" from archive
> 2008-08-24 23:03:01 CDT LOG:  restored log file
> "000000010000000400000089" from archive
> 2008-08-24 23:03:06 CDT LOG:  restored log file
> "00000001000000040000008A" from archive
> 2008-08-24 23:07:02 CDT LOG:  could not open file
> "pg_xlog/00000001000000040000008B" (log file 4, segment 139): No such
> file or directory
> 2008-08-24 23:07:05 CDT LOG:  startup process (PID 1468) was
> terminated by exception 0xC000000D
> 2008-08-24 23:07:05 CDT HINT:  See C include file "ntstatus.h" for a
> description of the hexadecimal value.
> 2008-08-24 23:07:05 CDT LOG:  aborting startup due to startup process
> failure
> 
> My recovery command is:
> 
> restore_command = '..\bin\pg_standby -d -s 5
> -t .\pgsql.trigger.5442 ..\data\archive %f %p %r 2>>pg_log
> \standby.log'
> 
> 
> I found this on the forums.  Could it be the source of the problem I'm
> experiencing?

The way you are using this looks correct.

The startup process would not/should not fail in this way and there
should be additional messages between the log entries from 23:03 to
23:07. I don't see any reason for a 4 minute wait at that point,
especially since previous files took 5 seconds. Are you showing us the
whole log?

Please re-run this at DEBUG3 so we can see what else is happening.

-- 
 Simon Riggs           www.2ndQuadrant.com
 PostgreSQL Training, Services and Support