Re: 10g clusterware causes reboots

From: <johnbhurley_at_sbcglobal.net>
Date: Tue, 14 Apr 2009 05:12:58 -0700 (PDT)
Message-ID: <715bc121-00ce-44f3-8d09-41f7f3c2d5a7_at_w40g2000yqd.googlegroups.com>

On Apr 13, 4:01 pm, The Derfer <derf..._at_gmail.com> wrote:
> I have Linux (OEL5.2) running 10g clusterware on 2 nodes.
> OCFS2 is running as well, with a SAN-based OCFS2 filesystem
> mounted on both hosts (written to by only one until the other fails
> over).
>
> When I reboot one host, the ethernet interface eth0:2 appears on the
> other, assuming the VIP of the downed host.
> But then when that host comes back up, the 2nd host reboots
> itself outright.   Has  anyone seen this behavior in either
> clusterware or OCFS2 before?  Oracle engineers claim it's
> OCFS2 doing this but I'm not sure.   Why would wither host
> reboot at all without me doing the rebooting?

Yeah I saw it a while back when testing in a similar config. Lots of varied opinions on OCFS2 and stability/useability ... etc. There are some timeout config thingie's one can play around with ( check out Jeffrey Hunter's site www.idevelopment.info ).

One of the formerly frequent posters on cdos when he stayed on topic recommended only deploying RAC in situations like this in a 4 node configuration. I don't think that is an official recommendation from Oracle but I could be wrong. Received on Tue Apr 14 2009 - 07:12:58 CDT