DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma;
	h=mime-version:in-reply-to:references:date:message-id:subject:from:to
	:cc:content-type:content-transfer-encoding;
	b=rZlpvnBA1BawoGVjo9MFfHmdbinZ1qM3Dq9LjycR1CbZqosXEaGbbm3s6+GWfJLtZY
	ambYGAAaG99IEiMQAeIpM6XUEN5LZaKz4KHYeONRKTrREChtmtVbR49c1RLnXK483BYS
	QsU/2fJzJ7chO6nA89SyfdofZebVOxBISwvDc=
MIME-Version: 1.0
In-Reply-To: <1482.71.192.236.30.1235793714.squirrel@mail.autorevenue.com>
References: <1482.71.192.236.30.1235793714.squirrel@mail.autorevenue.com>
Date: Mon, 2 Mar 2009 12:29:49 +0900
Message-ID: <3f0b79eb0903011929x3e774491m20b5d8ffa46bd57c@mail.gmail.com>
Subject: Re: Recovery issue
From: Fujii Masao <masao.fujii@gmail.com>
To: msteben@autorevenue.com
Cc: pgsql-admin@postgresql.org
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

Hi,

On Sat, Feb 28, 2009 at 1:01 PM,  <msteben@autorevenue.com> wrote:
> Hi L:sters,
>
> We're at postgresql 8.2.5. =A0I'm running PITR backups on our production =
server
> and copying the resulting tarball and the WAL logs to our reporting serve=
r.
> I restart the reporting server and begin re-playing logs. =A0Everything
> works fine for about 3 days until a network outage creates a gap in the
> WAL log copies. =A0The reporting server interprets this as the end of the
> logs and comes up in ready mode.
>
> I need to begin recovery at that point and continue to replay logs.
> I asked this question last week with no clue how to proceed.
> Now I throw an idea out to the group. =A0Can I take a full file copy on
> the reporting server and continue to play logs at the point the last
> recovery terminated?
>
> Procedure in summary:
>
> =A01. =A0Start archive logging on the reporting server.
> =A0 =A0 (Since no updates are occurring on this server there
> =A0 =A0 =A0should be no logs archived. but I need to start archive
> =A0 =A0 =A0logging to enable pg_start_backup)
> =A02. =A0select pg_start_backup('new_backup')
> =A03 =A0 tar -czvf (the data cluster)
> =A04. =A0select pg_stop_backup();
> =A05. =A0stop the server. Remove the data cluster
> =A06. =A0tar xvf the backup -replenish the data cluster
> =A07 =A0 remove postmaster.pid, (pg_xlogs already removed from backup)
> =A08. =A0 restart the server using recovery.conf to start copying the
> =A0 =A0 accumulated archive logs from the production server
>
> Possible? =A0Or =A0pipe dream?
>
> Any ideas welcome. =A0Thank you,

I don't think that your procedure works fine because timeline of
new backup is greater than that of the remaining archive logs.
I recommend that you use the previous backup, and re-execute
recovery from the beginning.

Regards,

--=20
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center