[Pgpool-general] Old master can not behaviour as slave. CURRVAL error after recovery.

[Armin Nesiren]

Hi there,

I'm running pgpool cluster with stream-replication on postgres 9.0.2
and pgpool-II-3.1.0alpha1.
There is one master and two slave nodes.

After failure simulation of my master node (node 0), my secondary node
took master role.
After I have recovered my old master node, i have problems.. and here it is
what I have in my log (this is just one part that is repeated many times):

10468 2011-05-23 13:57:17 CEST ERROR:  currval of sequence
"action_logs_id_seq" is not yet defined in this session

10468 2011-05-23 13:57:17 CEST STATEMENT:  SELECT
CURRVAL(pg_get_serial_sequence('public.action_logs','id'))

I must say that I have tried to recover my other slave node with new master
(with data from my new master, node 1),
and everything is working great with that node.

Problem is old master that can not behaviour normally as a slave.

Here is my log from old master (node 0) after recovery:

9362 2011-05-23 13:29:53 CEST LOG:  shutting down


9362 2011-05-23 13:29:53 CEST LOG:  database system is shut down


10428 2011-05-23 13:56:56 CEST LOG:  database system was interrupted; last
known up at 2011-05-23 13:52:41 CEST

10428 2011-05-23 13:56:56 CEST LOG:  creating missing WAL directory
"pg_xlog/archive_status"

10428 2011-05-23 13:56:56 CEST LOG:  entering standby mode


10429 2011-05-23 13:56:56 CEST LOG:  streaming replication successfully
connected to primary

10430 2011-05-23 13:56:56 CEST FATAL:  the database system is starting up


10428 2011-05-23 13:56:56 CEST LOG:  redo starts at 51/56000020


10433 2011-05-23 13:56:57 CEST FATAL:  the database system is starting up


10434 2011-05-23 13:56:58 CEST FATAL:  the database system is starting up


10435 2011-05-23 13:56:58 CEST FATAL:  the database system is starting up

10447 2011-05-23 13:57:05 CEST FATAL:  the database system is starting up


10448 2011-05-23 13:57:06 CEST FATAL:  the database system is starting up


10449 2011-05-23 13:57:07 CEST FATAL:  the database system is starting up


10450 2011-05-23 13:57:07 CEST FATAL:  the database system is starting up


10428 2011-05-23 13:57:07 CEST LOG:  consistent recovery state reached at
51/5D000000

10426 2011-05-23 13:57:07 CEST LOG:  database system is ready to accept read
only connections

10454 2011-05-23 13:57:10 CEST LOG:  unexpected EOF on client connection


10468 2011-05-23 13:57:17 CEST ERROR:  currval of sequence
"action_logs_id_seq" is not yet defined in this session

10468 2011-05-23 13:57:17 CEST STATEMENT:  SELECT
CURRVAL(pg_get_serial_sequence('public.action_logs','id'))

10468 2011-05-23 13:57:17 CEST ERROR:  currval of sequence
"action_logs_id_seq" is not yet defined in this session

10468 2011-05-23 13:57:17 CEST STATEMENT:  SELECT
CURRVAL(pg_get_serial_sequence('public.action_logs','id'))
..
..
..
..

Is it known problem that node 0 can not work as a slave normally, or I have
misconfigured  something?

Thanks for answers.

-- 
Armin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://pgfoundry.org/pipermail/pgpool-general/attachments/20110523/02571707/attachment.html>