Red Hat clustering on HP blades iLO fencing and OA failover

[chuckk281]

Richard had a customer question on Red Hat clustering:

**************

Customer asks the question regarding RHEL Fencing with iLO on HP BladeSystem and OA failover.  Please advise.  Might their also be a whitepaper or best practice on this topic?

"When fencing via iLO with blades and the primary OA goes down, will the OA/iLO route the fence through the standby OA?"

*************

Reply from Dan:

****************

I have no idea what fencing is, but it goes like this:

The path to the iLOs goes through the Active OA.

If the Primary OA FAILS, the secondary takes over automatically.

If the primary loses network, a failover will not occur (and thus iLOs will be unreachable) unless you turn on Link Loss Failover.

***********

Input from Joseph:

***********

Fencing is a term in Red Hat lingo to prevent data corruption in case of cluster connectivity loss, which can be implemented in different ways like shut down the non-responsive node, or/and on HP through ilo. Perhaps the RHCS manuals have more info on the supported use cases.

**************

Input from Jani:

************

If you are using ipmi-lan or ilo-fencing, you will use iLo ip-address for fencing. And yes, traffic is routed to another path if OA make failover. Fenced doesn´t care, where target is located, behind oa1 or oa2. Fencing only using ip-address and correct username and password.

**************

Any other help on this?