On 31 March 2016 at 10:43, Slava Bendersky <volga629@xxxxxxxxxxxxxx> wrote:
Hello Craig,The current setup is two server which run libvirt and for storage which run glusterfs (storage server feed two virtual servers). Right now is no fencing in place. Each of the nodes have one PostgreSQL vm with bdr.
That's a disaster waiting to happen. You can't just share storage like that in PostgreSQL, BDR or otherwise. I'm amazed it didn't fail earlier.
Try to dump whatever data you can recover, initdb a new normal PostgreSQL instance, restore, and set up normal replication. Look into repmgr, pgbarman, etc. Read the manual on replication, backup and failover. Do _not_ use shared storage.
