[Pgpool-general] Switchover question

[Joachim Wieland]

Tatsuo,

On Mon, Nov 27, 2006 at 10:33:33PM +0900, Tatsuo Ishii wrote:
> > Is pgpool supposed to switch back to the primary backend automatically at
> > all?

> Not it isn't. pgpool-II can do that, however. Do you want to do the
> samething for pgpool too?

I don't know how exactly how pgool does its healthchecks but that would be a
reasonable feature I think.

Is the switch to the master supposed to work from the command line? It
doesn't work here, I always have to restart pgpool to make it connect to the
master again.


> > Another thing: I'd also appreciate an intermediate log level. Right now
> > there is either no logging or too much of it. What I'd be interested in is
> > some logging that just tells me essential warning messages like "master
> > down", "no free pool slots available" and so on... Is something similar
> > available or in preparation?

> I thought important events such as failover or master going down are
> logged without turning on the debugging mode. Am I missing something?

Maybe I am missing something...

Okay, you are right, there were log entries... However this time it took
pgpool very long to find out that one server is down (about 15 minutes).

I just put a firewall rule on the machine that runs pgpool.

iptables -I OUTPUT -p tcp -d master --dport 5432 -j REJECT

First I could still connect to pgpool but not execute queries and then psql
just hang trying to connect to pgpool (I imagine that all slots were just
In my case, the master and the secondary server are more or less
close to the clients and all are on a fast network, so I'd like to tune the
failover process. Can I somehow lessen this failover interval?


Thanks,
Joachim