[Pacemaker] Postgresql replication problems

Eloy Coto Pereiro eloy.coto at gmail.com
Mon Dec 1 11:44:41 UTC 2014


Hi all.

After one year using Postgresql Cluster like this configuration[0], start
to fail, and I don't know why :-(

In the last days, my database move to  the second machine, and always had
pgsql-status: HS:SYNC, but I don't know why second server didn't wake up
correctly and pacemaker moved againg to the first server.

The problem was: when come back database in the first server was down
(Readonly) and data folder had recovery.conf. I don't know why, but in this
moment both postgresql convert to slaves and I can't promote the server.

It's a bit hard after the problem wake up the service. I need to delete
recovery.conf in server A, and start to make pg_basebackup in server 2. Now
I'm trying to wake up the cluster and I always got "PSQL is down" in the
syslog messages. And crm_mon get always msPostgresql slaves both nodes.

Does anyone has this problems? How can I setup Postgresql in severA as
master (to restore the cluster) without problems? Any special attribute?

I got always this messages in syslog:
Nov 30 21:42:34 serverA pgsql(pgsql)[10431]: INFO: PostgreSQL start command
sent.
Nov 30 21:42:34 serverA pgsql(pgsql)[10431]: INFO: PostgreSQL is down
Nov 30 21:42:35 serverA pgsql(pgsql)[10431]: INFO: PostgreSQL is started.
Nov 30 21:42:35 serverA pgsql(pgsql)[10431]: INFO: Changing pgsql-status on
serverA : STOP->HS:alone.

This is my config nowadays:
https://gist.github.com/eloycoto/c07177c63ee814c0397c

Can someone help me?
[0] http://clusterlabs.org/wiki/PgSQL_Replicated_Cluster

Many thanks.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20141201/3202ce50/attachment-0003.html>


More information about the Pacemaker mailing list