ServiceGuard failover when IP signal is lost

[S.Rider]

From past experience,we know our MC/SG packages will fail over if we loose IP signal to a primary server. Next month, the entire local network is being shut down for a rebuild/reconfiguration. Is there a way to tell MC/SG to turn off it's failover and ignore such things as loss-of-ip-signal (which is be happening on both the primary server and the failover server - so there actually won't be a place to fail to) ?

Any better alternative than just bringing down all the packages before the network outage and bringing them back up manually after the network inage ? 8-)

[Rita C Workman]

cmmodpkg -d

When your done.....re-enable the package with cmmodpkg -e

Rgrds,
Rita

[Rita C Workman]

I need to add something....cause cmmodpkg will do what you asked for........but.....

Why bother.

If your network is going to be down (IP not available), then are you planning on bringing down your database & apps. If you run oracle, the listener will get out of sync. For any third party apps, you can bet it will create some kind of problem if they broadcast and can't hit the database.
Another thing - your heartbeat. If your using a simple hub with nodes connected, you may continue to have the HB. But if your HB is through a switch - your down.

So....sorry for my ranting, but as Clay would say "....that dog won't hunt...". And your packages may avoid failover, but they wouldn't fail over if there was no network anyway. So if "your entire network is being shutd down" Your down!

Your better off to just cmhaltcl and come down clean.

Rgrds,
Rita

[Anka]

If the heartbeat is lost then you can not avoid the Serviceguard TOC as you have 2 node cluster.

a workarround would be to switch the packages on one of the nodes and then to halt the other one: #cmhaltnode...

If all the packages are currently running on one of the nodes then it is no problem, you can just halt the other node.

[melvyn burnard]

Well if you are shutting down ALL the network, what is the point in keep[ing any node or package up?
I would advise halt the cluster, and restart it once all the networks are back up and tested

[John Bigg]

And don't forget, that if the configuration of the network which affects the cluster chaanges in any way, e.g. bridging, you will need to run a cmapplyconf to re-build your cluster configuration afterwards before re-starting the cluster.

[Stephen Doud]

If the package is configured to failover if the subnet fails, then you have 2 choices...
If you wish to keep the package running in spite of a network outage, you must reconfigured the package to no longer monitor the subnet. See SUBNET reference in the package configuration file.
If, however, you are okay with the package halting when the network outage occurs, you can disable package failover using,
# cmmodpkg -d -n
or
# cmmodpkg -d (if several adoptive nodes are available)

If the network also supports heartbeat traffic, use cmviewconf to determine which networks sustain heartbeat, and insure a complete heartbeat transmission loss will not occur due to the network outage, or a node in the cluster will reboot itself.