Re: Split-brain among HACMP cluster and Oracle9RAC

ArneS wrote:
> Thanks for answere.
> But even if the IP network is down, I have still NON-IP network that is up
> and running (over diskhb). Should I still got global network down in such a
> case?

imho - yes - Search for IY50046 on IBM website - its an old apar but it describe the same behaviour as you do .
The non-ip network is just the last resort against a split brain situation.

> (I did not get any global network down event in my cluster.log file(s) when
> I did the test)

>From the Certification Study Guide - pSeries HACMP for AIX
...
Network events
network_down This event occurs when the Cluster Manager determines a network has failed. A network_down event can take one of two forms:
Local network_down, where only a particular node has lost contact with a network.
Global network_down, where all of the nodes connected to a network have lost contact with a network. It is assumed in this case that a network-related failure has occurred rather than a node-related failure.

The network_down event mails a notification to the system administrator, but takes no further action since appropriate actions depend on the local network configuration.
...

Even in case you have had a global network down nothing will happen except sending an email to the sysadmin - might worth to check mail.

So check your network setup again , read the HACMP logs carefully again or get somebody who has a very good understanding of hacmp and the logs.

In case the all of the nodes network adapter really lost their link i would ask IBM why you did not saw a global network down event or what should happen and whether this must be configured or not.

regards
Hajo

P.S please keep in mind that my hacmp knowledge is a little bit outdated.
P:S:S Would be nice if you could post the final result Received on Fri Sep 22 2006 - 15:29:33 CDT