is a failover target necessary for node-mgmt LIF in a 6-node cluster

[heightsnj]

We have used e0M for node-mgmt LIF on FAS8200, and the failover-policy is "local-only", a broadcast domain as the failover group. 

In this case, is an addidtional LIF for the node-mgmt necessary? If e0M is failed, my understanding is that the LIF will have no place to fail over to, because it can only fail over to a local LIF, or if this is a conern at all?

Can somebody please shed the light for me?

[jcbettinelli]

Hello,

The node management doesn't need to failover because it is designed to manage the node, and if the node has failed, you won't be able to manage it.

What is really important to have a failover setup, is the cluster lif management, which is necessary to failover in case of updates or failover.

I hope this helps.

[heightsnj]

Correct, if the node got failed, then I am not able to manage. However, what if the node itself is fine, only the mgmt LIF for any reasons got failed, would that justify the need to have a failover? I understand, I could use cluster-mgmt LIF to manage the node, or even SP connection, would that be the answer of no need for the failver? In our old setups, we do have failovers for node-mgmt LIF. It is different from our new setups. I am just trying to find out if this is really unnecessary. 

Thanks!

[jcbettinelli]

There are some commands which should directly be run from the node, like a cluster lif migrate, or a network port down up and down. But IMHO that shouldn't be so critical that you need to failover the node lif in case there is a link or hardware failure. 

[francisjkim]

It's always a good idea to set up redundant links for even the local-only LIFs.  There are certain node management operations that can only be done via a node-management LIF.  The FAS8200 has two more 1Gb ports e0c and e0d, in addition to your e0M port.  If your upstream switch has spare ports, I recommend you use them.

[heightsnj]

francisjkim,

 Good to hear that you think a failover for node-mgmt LIF is needed. However, the argument here is that all node mgmt operations can be achived by the cluser lif, or SP connection, so, the redundant is not needed.

Can you list a few examples here that such redundant is needed?

[wareer]

Generally, we do not configure redundancy for nodes. To avoid non-redundancy alarms in node management, we turn off these alarms，A LIF that provides a dedicated IP address for managing a particular node in a cluster. Node management LIFs are created at the time of creating or joining the cluster. These LIFs are used for system maintenance, for example, when a node becomes inaccessible from the cluster.