P4000 Move Failover Manager from Cluster

[BrettYoung]

Hi Folks,

We recently had a clustered P4000 San (2 nodes) installed in our network complete with Failover Manager.

The individual who installed the San installed the Hyper V edition of Fail Over Manager on a Hyper V server that is part of a CSV HyperV cluster on the same P4000 San Cluster. (if that make sense).

When the HyperV server that hosts the F.O.M. reboots, we lose quorum on the P4000 San Cluster and everything slams shut.

My question is how do I go about moving the Fail Over Manager to another server?

Do I remove the Fail Over Manager from the P4000 Cluster,  Move F.O.M. to another Hyperv box, and then re-add it to the P4000 Cluster?

or

Can I just power off F.O.M and move it to another HyperV machine?

or

Am I best to remove the F.O.M from the P4000 cluster and start fresh with a new F.O.M. on another machine?

[oikjn · Accepted solution]

I had this big long speech posted, but I'll keep it short...

either way works.  remove the old and recreate a new FOM on a local disk or export the old and import it back to a local disk... doesn't matter... either way getting that FOM to a local independent disk is correct.

edit #2>   You weren't clear if your VSA system is hardware dedicated or the software VSA... if its software, make sure you don't put one of the VSA nodes on the same system as the FOM since if you lose that hardware you still lose quorum.  IE:  you must have three physical host computers to have a two node SAN and two node hyper-v cluster.  You can make a two-node hyper-v cluster and put one VSA instance on the local storage of each of the hyper-v nodes and you will maintain N+1 redundancy as long as you have a FOM installed on a third hyper-v server (that is using local storage for the FOM as well)

[BrettYoung]

That was the answer i was looking for.  Thanks!

[Urimuqi]

to be honest, i do not understand very much about failover manager and virtual manager.

[1957]

We need to move the FOM in an VM to another ESX server.  Is process to shut down FOM, VMotion VM to other ESX server, restart FOM?

I assume if you shut down FOM you don't have FO capabilities but arrays will run fine.

Thanks,

Dean

[HPstorageTom]

The cluster will work as long as you do have the quorum (majority) of manager. The FOM is only one of the manager running in a StoreVirtual cluster. You should see additional manager running on your StoreVirtual nodes in your cluster.

I guess that you do have a multi-site setup (hence your comment on FO capabilities). Correct?

If yes, then for a 2-site configuration you should have the following manager configuration according to HP best practices:

                                                                               Site 1               Site 2                 Site 3

2 node StoreVirtual cluster:                       1 manager     1 manager      FOM

4 node StoreVirtual cluster:                       2 manager     2 manager      FOM

6+ nodes SV cluster:                                     2  manager     2 manager      FOM

You can see the manager setup in the management group overview in the CMC.

[1957]

Thanks,

See attachment in previous post.  SHows FOM and two sites.  All are normal manager and FOM is special manager.

This is a configuration I inherited.

[oikjn]

The FOM acts as the arbitrator for quorum in the management group.  It holds the state of the management group as well so it does a little more than just provide a vote, but basically it acts as a tie breaker for what node should take over to prevent the possability of having a split-brain situation when each node is operating but the connection between the two is lost.  

Assuming you don't have common connection issues between your sites, you should be safe to shut down the FOM and VMotion it to its new home and turn it back on and everything will be happy (assuming your IP layer will still allow for communication at the new site).  The only caution would be if you have to change the IP at the same time as the site move.  In this situation, I would talk through with support for some extra guidance in the move, but they may say to just install a new FOM at the new location and then when its ready and patched to the same level to remove the existing FOM from the management group and add the new one.

Your config looks fine and is setup correctly.  It will handle a single node failure without downtime, but the loss of a 2nd node will cause an outage (including FOM).  Basically thats about as good as you can get with just two storage nodes and it won't get any better until you hit 4+ storage nodes.

[1957]

Thank you for the confirmation!