[Pgcluster-general] Dying replicators

[Benjamin Vogt]

Hi,

We have a setup running one load balancer, two clusters and two
replicators using pgcluster-1.7.0rc7 on CentOS 5 as follows:


		LB
		 |
	 ----------------
	 |		|
	Cluster1	Cluster2
	Replicator1	Replicator2

The cluster dbs and replicators run on the same machine. The setup is
fine, I've checked it about a hundred times. However, we have a problem
where the replicators simply die after a while. The replicator daemon
process keeps running on both machines, but no updates can be made to
the db as below:

billing=# update channel set chan_name = 'Rolf' where chan_id = 4;
UPDATE 1
billing=# update channel set chan_name = 'Rolf' where chan_id = 4;
UPDATE 1
billing=# update channel set chan_name = 'Rolf' where chan_id = 4;
UPDATE 1
billing=# update channel set chan_name = 'Rolf' where chan_id = 4;
ERROR:  This query is not permitted when all replication servers fell down 
billing=# update channel set chan_name = 'Rolf' where chan_id = 4;
ERROR:  This query is not permitted when all replication servers fell down 

Another thing I noticed is that when I start the replicators before the
cluster daemons, the replicators will immediately die after logging out
from the server. I first have to start the cluster daemons, then the
replicator daemons, but even then, after a certain amount of time the
replicator service will cease working.

We would like to implement pgcluster into a production environmnet, but
the habit of the replicators does not make me feel secure. Before
looking at pgcluster, we had a look at Slony and I must say I feel sick
even speaking about it. I hope we can get pgcluster running in an
optimal manner.

Any advice?

Sincerely,
Benjamin Vogt