Re: Split-brain among HACMP cluster and Oracle9RAC

ArneS wrote:
> "Hajo Ehlers" <service_at_metamodul.com> wrote in message
...
> > Is the cluster node directly connected to these switches ?
> >
> Yes.
> 2 server rooms with 2 switches in each server room (and 2 AIX boxes in each
> room as well)
> The test was based on turing off the two switches in one server room, we did
> not touch anything on the AIX boxes, just sat down and observed a lot of
> logs.

In case all physical network connection are lost you should get finaly a:
 'global network down event '

So check if you got one and how the system shell behave in this case. If you did NOT got a 'global network down event ' then either the hacmp is buggy - at least v5.1 had problems - or your network is not as you assume so the cluster will only have a 'network down event.

hth
Hajo Received on Fri Sep 22 2006 - 11:43:02 CDT