warm standby server configuration

From: "George Wilk" <gwilk(at)ellacoya(dot)com>
To: <pgsql-admin(at)postgresql(dot)org>
Subject: warm standby server configuration
Date: 2007-06-26 11:47:24
Message-ID: 016601c7b7e7$c337f060$0d7ba8c0@ellacoya.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

Hi there,

A bit of background information:

I am running Postgres 8.2.4 on Solaris 10 in a warm standby configuration.
My archive command scp's WAL logs from the live server to the standby
server. The standby server runs in a recovery mode waiting for the WAL logs
and processing them as they become available (at least that's the intent).

The standby server was originally seeded with a copy of the live server
PGDATA directory. The live server is configured with the checkpoint_timeout
= 5min, resulting in WAL logs being created every 5 minutes. Because this
is a test environment, I am testing the backup/recovery scenario w/o any
database updates during the test cycle. The warm standby server produces
the following error log entry on recovery startup:

LOG: starting archive recovery

LOG: restore_command = "exec /var/ipsc/pgsql/data/recovery_cmd %f %p"

cp: cannot access /var/ipsc/pgsql/WAL/00000001.history

ll /var/ipsc/pgsql

file: 00000001.history

path: pg_xlog/RECOVERYHISTORY

ll /var/ipsc/pgsql

file: 0000000100000000000000B7

path: pg_xlog/RECOVERYXLOG

LOG: restored log file "0000000100000000000000B7" from archive

LOG: record with zero length at 0/B7000340

LOG: invalid primary checkpoint record

ll /var/ipsc/pgsql

file: 0000000100000000000000B7

path: pg_xlog/RECOVERYXLOG

LOG: restored log file "0000000100000000000000B7" from archive

LOG: record with zero length at 0/B70002F0

LOG: invalid secondary checkpoint record

PANIC: could not locate a valid checkpoint record

LOG: startup process (PID 1222) was terminated by signal 6

LOG: aborting startup due to startup process failure

The WAL logs are copied from the live server to the /var/ipsc/pgsqlWAL
directory on the standby server. It seems as though the standby server is
looking for a history file. Where is this file being produced? I couldn't
find it anywhere on the live server.

The real problem, however, is caused by the "record with zero length" and
prevents the standby server from booting up. Since my WAL logs may have no
records in them, is this going to be a problem?

As always, your help will be greatly appreciated!

Cheers,

~george

Browse pgsql-admin by date

  From Date Subject
Next Message Michael Monnerie 2007-06-26 15:11:06 How to set a constraint on non-unique fields?
Previous Message Adam Radlowski 2007-06-26 06:43:06 Re: replication between linxu postgresql and Windows postgresql