Breaking a metrocluster

We're trying to break a metrocluster and it's not a particularly well documented thing. We want one to stay as primary storage (we have offsite recovery now) and the other to become a reference system. All systems will retain half the disks (3 SATA sheleves and 9 FC)

Two 3160s in the MetroCluster. Everythin is syncmirrored AYWE.

How do we go about this? Can it be done live and without (or with minimal) disruption?

I've built MCs but never had a need to destroy them.

Re: Breaking a metrocluster

It can’t be done online – you need to shut down filers (for extended period of time) which obviously will be disruptive without cluster failover.

Basic steps are described here:

Make sure filers own whole shelves, otherwise you may need to reshuffle disks between shelves as non-HA-pair cannot have access to the same storage. You may need to change NVRAM slot, check system configuration guide for your model. You will also need to change HA cabling from MP-HA (or recable if you have FMC). You will also need to remove cluster_remote license to avoid warnings about non-mirrored root aggregate.

Otherwise it is relatively straightforward ☺

Re: Breaking a metrocluster

I would also run it by your rep and make sure support will be ok after doing this... since licenses are cluster based (discount on the 2cnd controller) there could be some questions if you need support and the system is no longer clustered... make sure you are covered under two single controllers unclustered...and are you going to meet your SLAs with a single controller in production?  It might be worth bringing the controllers into the same data center and add a new single node for the reference system then you keep HA for primary.. the budget may not have that in the cards though.

Re: Breaking a metrocluster


Both nodes own whole shelves and these are divided equally.

Here's a little of my thinking.

I am removing a node. I don't want it any more and it will not be used in the MC again. I have syncmirrors or original plex of all the data on the primary node. Can I not make use of the cluster failover somehow to "grab" all the data on the primary controller and then make use of snapmirrors to sync to a known set of disks (aggregate). Then I simply remove the extra disk shelves (both logically and physically). Then I can build the other standalone system.

I'm not asking for a totally non-disruptive solution as I know I'll have to remove VIs, shelves, cabling etc but can I not make it so that at least I have the primary system presenting the correct stuff before I do any of that.

Re: Breaking a metrocluster

Then you probably will be stuck with cluster system in takeover state and degraded mirror forever without any clear path how to get out of this situation.