[pgpool-general: 3144] [pgpool 3.3.3-4] - defunct processes while failover

[Attila Heidrich]

Hi!

I have seen basicallythe same issues here, but currently my problem is a
bit different.

Similar configuration (I guess):

5 clients, all running pgpool, no virtual IP, hosts use the local pool.
2 backends (actually 2 of the 5 clients, since backends run pgpool2 as
well), streaming replication, online_recovery, pool config is streaming
replication as well, with load balancing.

I still don't get the exact idea of how this should work.

Pools don't seem to share the status information:
 - Stopping one of the backends removes the backend only on the pool hosts
where there is open database link.
f.e.:

qa-openfire-1:
    Node: 0
    Host: qa-postgres-1
    Port: 5433
    Weight: 0.250000
    Status: Up, in pool and connected (2)

    Node: 1
    Host: qa-postgres-2
    Port: 5433
    Weight: 0.750000
    Status: Up, detached from pool (3)
qa-postgres-1:
    Node: 0
    Host: qa-postgres-1
    Port: 5433
    Weight: 0.250000
    Status: Up, in pool (1)

    Node: 1
    Host: qa-postgres-2
    Port: 5433
    Weight: 0.750000
    Status: Up, in pool (1)

"postgres*" hosts are the backends, pool instance is almost always idle. I
have restarted postgres-2 recently. Only those pool instalces detached the
stopped server, where it was in Connected status. Is it OK this way?

If other_pgpool* and heartbeat_destination* are setup correctly, the nodes
are supposed to share consistent status aren't they?

regards,
 Attila
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.pgpool.net/pipermail/pgpool-general/attachments/20140904/ac1dc0c9/attachment.htm>