Net switching back in a SG cluster

[Jdamian]

Hi everyone.

My cluster has net redundancy, i.e, two NICs configured -- one active (lan1), one stand-by (lan2).

I'm sure that when active NIC fails, SG will switch all IP addresses to the stand-by NIC.

My doubts lay in the net failback: when lan1 works back fine again, does SG will move the IP addresses back to lan0 ? Is this the default behaviour? How can I configure?

Thanx a lot.

[Stephen Doud]

Every NETWORK_POLLING_INTERVAL (see your cluster configuration file), Serviceguard tests the transmission ability of each NIC it monitors.
If after a primary NIC failure which caused Serviceguard to move the network stack to the standby NIC, Serviceguard detects the primary NIC is healthy again, it will automatically move the network stack back to the primary NIC.

[Ralph Grothe]

Hi Oscar,

the internal network monitoring algorithm of SG should detect this on its own.
It should also failback automatically if primary NIC gets reconnected, if I recall correctly.
Anyway, this failover and failback tests is something you should devise in any case whenever you deploy a cluster.
Simply unplug the NIC wires, giving enough time for discovery and failover.
Then run netstat -in to verify.
Finally, reconnect.
Do this with all your nodes.

[Jdamian]

Thank you

I'm rather interested in finding out if this behaviour can be changed by means of a cluster parameter in the style of FAILBACK_POLICY for packages (http://docs.hp.com/en/B3936-90065/ch03s03.html#d0e2335)

[Stephen Doud · Accepted solution]

Standby -> Primary NIC failback cannot be disabled through a configuration parameter. It is not a feature of Serviceguard.

FAILBACK_POLICY in the package configuration file does not refer to NIC failback, but to package failback.

[Jdamian]

fine