This method,altho interesting, fails to properly account for the new MAC address of the "live" toaster and also fails to account for differences that exist in the NVRAM states of the two machines..both of these issues are addressed in our new Cluster-Failover product that we have announced and is due out in the October time-frame...
-----Original Message----- From: ian.gardner@vf.vodafone.co.uk [mailto:ian.gardner@vf.vodafone.co.uk] Sent: Thursday, September 10, 1998 10:22 AM To: toasters@mathworks.com Cc: richard.hillier@vf.vodafone.co.uk; mark.jones@vf.vodafone.co.uk; julie.stephen@vf.vodafone.co.uk Subject: A netapp failover possibility...???
Hi,
We have two netapp filers serving mainly static data (updated overnight) to several hp-ux clients.
If one server fails we can swap the clients onto the other server by mounting and switching symbollic links.
eg. on a client:
/discs/data -> /netapp1/data
If netapp1 fails we mount data from netapp2 and relink:
/discs/data -> /netapp2/data
Now this leaves processes hanging on the dead server which can be killed using fuser -k netapp1:/data. This is only necessary if memory/swap on the clients is tight, otherwise the "hung" processes will sleep until the server comes back.
This is not ideal so I had a thought, since we have Quad 100BaseT cards, in the event of a failure, we could configure up a spare port to be the same name and IP address as the dead server and therefore take over the nfs requests for it.
Since we are LIVE I am unable to test this, but subject to the routers and arp tables becoming updated I would have thought this would work. I'm not sure about file locking, but I didn't think this works very well on nfs anyway - does it ? NFS is stateless so it should be ok.
Anyone out there tried this or able to on a couple of test boxes ?
Cheers
Ian Gardner Vodafone Ltd. Newbury UK. ian.gardner@vf.vodafone.co.uk