serviceguard node shutdown on multiple lan failure

[Robert Meredith]

Is there a way to stop serviceguard shutting down a node if all the lan cards (or networks) fail. ie we had a situation where all the network conneections went down and therefore the active node went down. We want the situation where it will just sit still and wait for the networks to come back.

Cheers,

ROB

[Vincenzo Restuccia]

One characteristic of SG is that without communication lan the package is down not creed is ways for delaying it.

[melvyn burnard]

This is not what SG was designed to do. What if the lans never come up again?
The idea is that a package will switch to a node that hopefully has networking still ok :-}
Why would you want a node with no comms to stay up??

[A. Clay Stephenson]

There is no way to do what you want; the best you can do is increase the NODE_TIMEOUT value
in the cluster configuration file. You will probably also have to increase the HEARTBEAT_INTERVAL also unless this is a 2 node cluster in which case you might look into a serial heartbeat link. The downside is that increasing the NODE_TIMEOUT value will also increase failover times. It looks as though you should look into a more robust network.

[Peggy Fong]

Robert,
The previous post is the one that should help you. We used to have a poor network and would get 8 or so different serviceguard nodes failing because of lost networks. We increased the NODE_TIMEOUT parameter (sorry I don't have the value - I think it was changed to 20000000 (ms) but not sure. It did solve our problem. We also had to increase the value per HP for some other reason. I know I'm not much help but wanted to confirm that Stephenson's response should help you out.
Pfong