Community

Subscribe
Highlighted
Accepted Solution

How do we non-disruptively recreate an if_grp?

We are currently running cDOT 8.2P6 on a 2-node cluster with mirroring to a similar system at a remote site.

During implementation, we created an if_grp (a0m) on each node that is made up of ports e0a and e0b.  The a0m if_grp hosts the admin LIF's for our FCoE SVM's, the cluster interconnect LIF, and the cluster_mgmt LIF.

Everything is working fine, but we just realized that the if_grp was created as a multimode type with mac-based sourcing.  The switches that the ports are connected to are not setup in an ether-channel and do not support it.  (We will be moving to newer switches in the future that do support ether-channel, but we need to address this situation now.)

I know that we cannot change the if_grp type from multimode to single.  From what i've gathered, we need to delete and recreate the if_grp.

My question is what is the best way to delete/recreate the if_grp without impacting the system?  And what are the steps?

Is it as simple as:

1) Migrate those LIFs on node1 to node2

2) Delete the if_grp

3) Create a new if_grp

4) Migrate those LIFs back to node1

5) Repeat 1 - 4 for node2

Re: How do we non-disruptively recreate an if_grp?

I just worked with TAC on this same issue with slightly different variables.  The short answer is that you can not change the Distribution Function or Create Policy without deleting the ifgrp.  Doing so means migrating the LIFs, removing VLANs (if required), deleting the ifgrp, creating the ifgrp, creating the VLANs, then migrating back.  Then repeat on the other ifgrps just as you mentioned. 

-rh

Re: How do we non-disruptively recreate an if_grp?

After working with Tech Support, we identified the following steps to remove the if_grp.  (We decided not the recreate the if_grp and just use failover groups for redundancy of the ports, since trunking was not required)

  1. Pause any replication occuring on the if_grp
  2. On Node 2
    1. Migrate any node2 vserver/cluster admin LIF's from the if_grp to a valid port on node1
    2. Migrate any node2 intercluster LIF's from the if_grp to a valid port on node2
    3. Set home port on any node2 vserver/cluster admin LIF's to a valid port on node1
    4. Set home port on any node2 intercluster LIF's to a valid port on node2
    5. Remove any failover group references to the if_grp
    6. Verify nothing is using the if_grp (or the system won't let you delete it)
    7. Delete the if_grp
    8. Set home port on any node2 vserver/cluster admin LIF's back to the correct port on node2
    9. Set home port on any node2 intercluster LIF's back to the correct port on node2
    10. Revert the interfaces to their home ports
    11. Check cluster peer status
    12. Check network interface status
      1. Make sure all interfaces are up
      2. Make sure all interfaces are home
    13. Put the vserver/cluster admin LIF's into a new failover-group
  3. Repeat the steps on Node1 (send the LIF's to node2, then back to Node1)
  4. Resume intercluster replication

Hope that helps someone