[ClusterLabs] Postgres Cluster - Two Slaves, No Master

Robert Pyzalski pyzals at gmail.com
Wed May 17 00:48:33 EDT 2017

Hello Cluster Experts,

I followed this guide for establishing a replicated Postgres cluster
managed by Pacemaker:

The only difference being that I implemented replication using a
replication slot.

When I attempted to start the cluster, I wound up with two slaves. After
much teeth gnashing and hair pulling I noticed that both nodes were in the
DISCONNECT state. I manually set node1 to LATEST like so:

crm_attribute -l forever -N node1 -n pg_cluster-data-status -v LATEST

and Pacemaker immediately, successfully started the Master/Slave cluster.

This seems like something I shouldn't have needed to do. Does anyone know
where I might've gone wrong?

Thank you,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.clusterlabs.org/pipermail/users/attachments/20170516/8f4ddd8e/attachment-0002.html>

More information about the Users mailing list