Re: Fatal Error during PITR Recovery

Hi Tom,

Thanks for your comment.

After posting this question here, I did a quick search in Google, and I'm
finding other kind of reasons also, like a hardware problem, here:
http://www.mail-archive.com/pgsql-general(at)postgresql(dot)org/msg69085.html

But in my case, I don't think that this could be a hardware problem, because
my database server is running in Amazon EC2 cloud. So the chances of
hardware failure is least expected. As you said, this is more likely of a
bug in v8.2.3.

I'm finding another solution also here, to do 'pg_resetxlog', even though
this problem is reported on during database startup after an abrupt
shutdown. I believe that 'pg_resetxlog' will work only when the database is
in normal mode and not in recovery mode. Is my understanding correct?
Please comment on this.
http://stackoverflow.com/questions/598200/how-do-i-fix-postgres-so-it-will-s
tart-after-an-abrupt-shutdown

If I would like to upgrade to the latest minor version in 8.2.x series, that
is v8.2.17, how do I upgrade this without doing dump/restore?
Links/documentation on this are appreciated.

Regards,
Gnanam

-----Original Message-----
From: Tom Lane [mailto:tgl(at)sss(dot)pgh(dot)pa(dot)us]
Sent: Wednesday, June 09, 2010 9:27 PM
To: gnanam(at)zoniac(dot)com
Cc: pgsql-admin(at)postgresql(dot)org
Subject: Re: [ADMIN] Fatal Error during PITR Recovery

"Gnanakumar" <gnanam(at)zoniac(dot)com> writes:
> My production is running PostgreSQL v8.2.3 on CentOS release 5.2 (Final).

> As part of our routine, we just wanted to make sure and practice once in a
> while, whether PITR recovery process is performed without fail. When I
> started the recovery process, after sometime, I see the following error in
> the serverlog.

> []:2010-06-09 04:44:59 EDTLOG: could not fsync segment 0 of relation
> 1663/169462/252056: No such file or directory
> []:2010-06-09 04:44:59 EDTCONTEXT: xlog redo checkpoint: redo
13/3FE33EB0;
> undo 0/0; tli 1; xid 0/331456443; oid 256783; multi 337; offset 676;
online
> []:2010-06-09 04:44:59 EDTFATAL: storage sync failed on magnetic disk: No
> such file or directory

This is probably a bug we fixed in 8.2.4:

2007-04-12 11:04 tgl

* src/backend/commands/dbcommands.c (REL8_2_STABLE): Cancel pending
fsync requests during WAL replay of DROP DATABASE, per bug report
from David Darville. Back-patch as far as 8.1, which may or may
not have the problem but it seems a safe change anyway.

You really ought to be running a less ancient minor release of PG --- 8.2.x
is up to 8.2.17. Your OS sounds a bit long in the tooth as well.

regards, tom lane