Convert MetroCluster in ordinary HA Pair

[SJOERDHOOFT]

Hey everyone,

we have a metrocluster which seems to be setup normally. Aggr0 is mirrored between the filers, and has only one raidgroup on each head. The original reason for creating a metrocluster is not valid anymore since the two heads are only a few meter away from each other. The filers are almost end of life, and almost out of diskspace. For budget reasons, we've decided to not buy anymore disks/shelves, but instead break the current aggregate mirrors so we free up the disks holding the copies.

Looking for the correct way to do this, I found two methods:

Splitting the aggregate is one option, as explained here: https://library.netapp.com/ecmdocs/ECMM1278318/html/onlinebk/protecting/task/t_oc_prot_sym-split-miror-aggr-or-trad-vol.html#t_oc_prot_sym-split-miror...

If I understand correctly I would have to do:

aggr split aggr0/plex2 aggr1

This will split off the second plex and create a new aggregate called aggr1. I could then remove the aggregate aggr1 and add the disks to aggr0.

The other option is to remove the plex from the aggregate, as explained here: https://library.netapp.com/ecmdocs/ECMM1278318/html/onlinebk/protecting/task/t_oc_prot_sym-split-miror-aggr-or-trad-vol.html#t_oc_prot_sym-split-miror...

That would require these commands:

aggr offline plex2

aggr destroy plex2

This will destroy the plex, and the disks will be turned into hot spares which can then  be added to aggr0.

So the second option would save the steps to remove the newly created aggregate.

Is any of these ideas wrong, am I missing something? Is there something else to consider? Both steps do not need any downtime as fas as I can see.

Any comments/ideas etc please let me know!

Kind regards

Sjoerd

[thomas_glodde · Accepted solution]

offline/destroy plex is fine, requires no downtime, just do a disk zero spares afterwards and you are good to go and can create new aggregates.

[aborzenkov]

Is it fabric or stretch MetroCluster?

[SJOERDHOOFT]

Does that change anything? Because I'm not sure what that means? I'm planning to do the offline/destroy option, does it make a difference in my case?

[radek_kubka]

Stretched MC - short distance, where you just use longer FC cables for cluster interconnect & disk shelves cabling

Fabric MC - longer distance, where back-end FC switches (& long-haul / dark fibre) are used for cluster interconnect & disk shelves cabling

In the latter scenario you'd probably wish to remove FC switches from the setup, which would be disruptive.

Regards,
Radek

[SJOERDHOOFT]

That would be a stretched metro cluster then, so no problem for me. Thanks for the responses and help!