Wrong SYSID in WAL segment

From: "Latrous, Youssef" <YLatrous(at)BroadViewNet(dot)com>
To: <pgsql-general(at)postgresql(dot)org>
Subject: Wrong SYSID in WAL segment
Date: 2010-10-25 18:20:58
Message-ID: 4B393627EE3CE74A9F189B639F857D2E02182B60@BVNYMAIL.bvn.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

Hi there,

In one of our systems we've noticed the following (strange?) behavior.
In a Master/Slave configuration, we run wal-mgr to achieve data
replication from the master to the slave node. Both nodes run PostgreSQL
8.4.1 (on Solaris 10). The slave starts in archive mode and initiates a
restore from the log files. After few segments, it complains with the
following error message:

...

LOG: WAL file is from different system

DETAIL: WAL file SYSID is 5466170076771909117, pg_control SYSID is
5516922116183112703

LOG: redo done at 0/8FFE920

LOG: last completed transaction was at log time 2010-09-17
10:19:49.545025-04

2010-09-17 10:20:15,621 20400 INFO 000000010000000000000008: Found

2010-09-17 10:20:16,776 20400 INFO {count: 1}

LOG: restored log file "000000010000000000000008" from archive

2010-09-17 10:20:17,118 20417 INFO 00000002.history: not found, ignoring

2010-09-17 10:20:17,119 20417 INFO got SystemExit(1), exiting

LOG: selected new timeline ID: 2

2010-09-17 10:20:17,458 20433 INFO 00000001.history: not found, ignoring

2010-09-17 10:20:17,459 20433 INFO got SystemExit(1), exiting

LOG: archive recovery complete

LOG: database system is ready to accept connections

LOG: autovacuum launcher started

I'm trying to understand few things here:

1) Why one of the segments has a different SYSID (knowing that all WAL
segments are originating from the same node)? In other words, under
which circumstances the SYSID could be different from one segment to
another one on the same node? CRC is ok, which means that none of these
segments is corrupted.

2) Once the postmaster encounters this issue, it stops the recovery and
switches to master mode, hence breaking Master/Slave mode. What's the
rationale behind changing underneath the system the database mode?

3) How would one know that the replication is progressing correctly?

I'm not sure if it's the right mailing list. If not, please let me know
which one is more appropriate and I'll post there.

Thank you in advance for your time and help,

Regards,

Youssef

Responses

Browse pgsql-general by date

  From Date Subject
Next Message Alexey Luchko 2010-10-25 18:27:24 replication on win32/64
Previous Message Fredric Fredricson 2010-10-25 17:58:32 Re: What is "return code" for WAL send command