[pgpool-general: 1239] Clear different states between two PGPool nodes

[Maik Kulbe]

Hello List,

I have two PGPool 3.2.1 nodes(both having PGPoolAdmin) running with the
watchdog and working with four Postgres 9.2 nodes as backend servers in
Load-balanced replication mode and I am evaluating weaknesses of this
cluster at the moment. Now I have a problem I could not really find how to
solve:

I create an inconsistent state on one of the machines, recover, repeat.
After a while PGPoolAdmin on the primary node shows all 4 nodes as okay,
but PGPoolAdmin on the secondary node(not having the shared IP) reports all
but one node as failed.

Is there a way to sync both PGPool nodes or at least 'reset' the PGPool on
the second node(and will it resync the states from master or do I have to
attach the nodes via a pcp command)?

Also it seems I am having problems with the recovery process when it
involves sequences and more than two nodes. I read that the recovery script
in 2nd stage has to flush all sequences on all nodes that are online but
I'm not sure how to do this when dealing with a large amount of nodes(the
basic assumption is, that the cluster I am testing will be used for a large
production environment and might push towards 100 nodes) of which some
nodes may be completely offline or have gone through fail over.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.sraoss.jp/pipermail/pgpool-general/attachments/20121206/eecfd322/attachment.html>