Re: Failover failed

--On 18 August 2006 15:07:29 +0100 Ian Eiloart <[email protected]> wrote:

> Hi,
>
> I have a cluster with two nodes, lynndie and sivits. Sivits has been
> falling over a lot recently - require reboots every other day. It's a new
> machine, so I expect to get it repaired. I'm not in a great hurry,
> because failover with wackamole seems to be working fine....
>
> Until now. Lynndie just fell over, and no failover happened. Spread and
> wackamole configurations are identical on both hosts, as are hardware
> configurations.
>
> As Lynndie came up, Sivits got all the IP addresses for the cluster, but
> has since dropped some. Lynndie hasn't aquired any. Spread logs show that
> Lynndie did join the spread group on boot, and Sivits logs show that
> *something* happened at the time Lynndie broke, but not that any member
> left the group.
>
> Wackamole logged nothing on either host until I rebooted Lynndie.
>
> Does anyone know what could cause this kind of asymmetry.

It turned out to be two named DNS server processes running on the host.


-- 
Ian Eiloart
IT Services, University of Sussex