[Pgpool-general] Newbie questions

[Doyon, Jean-Francois]

Very well ... So what happens to SQL statements right after the cable was
unplugged, but before the health check ran?  It fails on the secondary
server, but ran properly on the first?  More importantly, does the client
get an error?

Pgpool says nothing, even in -d mode.  I suspect it's because, without
health checking,  the first connect attempt fails because my secondary
server doesn't exist yet.  It enters degeneration mode the first time a
client tries to connect, but the client gets an error when this happens.  On
the second try, all works well.

That read-only when degenerated feature would be nice, no need to worry
about syncing ;)

J.F.

-----Original Message-----
From: Tatsuo Ishii [mailto:ishii at sraoss.co.jp] 
Sent: March 23, 2006 10:03 PM
To: Jean-Francois.Doyon at CCRS.NRCan.gc.ca
Cc: pgpool-general at pgfoundry.org
Subject: Re: [Pgpool-general] Newbie questions

> Hello,
> 
> I am working on setting up a pgpool + heartbeat cluster for pgsql 
> replication and fail over, and have a few questions:
> 
> - Do I need the health check running for proper fail-over? What is a 
> good value?

Without health checking some case such as plugging out ethernet cables may
not be detected for a while since in this case the kernel TCP/IP keep on
retrying for hours.

I recommend health_check_period to be larger than health_check_timeout. For
example, health_check_period=30 and health_check_timeout=20 would be your
starter.

> - Does degeneration mode imply "read-only"? Is there any way to make 
> things read-only while the cluster isn't fully available?

Currently no.

> - Sometimes when I connect with the secondary server down, I get an error.
> I can connect the second time around.  Is this normal?

Can you be more specific? What pgpool log says? (hint: -d option will be
helpfull).
--
Tatsuo Ishii
SRA OSS, Inc. Japan