Re: PITR Base Backup on an idle 8.1 server - Mailing list pgsql-general

From Marco Colombo
Subject Re: PITR Base Backup on an idle 8.1 server
Date
Msg-id 4666B454.6030904@esiway.net
Whole thread Raw
In response to Re: PITR Base Backup on an idle 8.1 server  ("Simon Riggs" <simon@2ndquadrant.com>)
List pgsql-general
Simon Riggs wrote:
> On Tue, 2007-06-05 at 18:39 +0200, Marco Colombo wrote:
>> I'm asking: what _exactly_ can go wrong?
>
> If a checkpoint occurs while taking the backup then the contents of the
> files will be overwritten
   ^^^^^
Data files or WAL segments? My archive command prevents WAL segments
from being recycled during the backup.

  and you will be unable to rollforward from
> before the backup until after the backup. This will give you the FATAL
> error message "WAL ends before end time of backup dump". You won't know
> this until you have attempted recovery using those files, even if the
> scripts give rc=0.

Well, my procedure currently produces two tar achives. One is the 'base
backup' (a copy of the datafiles). One is a copy of wal segments, right
after the backup. Which one do you expect to be corrupted if a
checkpoint happens during the backup?

.TM.

pgsql-general by date:

Previous
From: "Bart Degryse"
Date:
Subject: Re: plperl and/or insert trigger problem
Next
From: Frank Wittig
Date:
Subject: Re: wal files restoration