SnapMirror destination volumes and DR

[tvmo_tvmo]

Hello,

ONTAP: 7.3.2

My SnapMirror destination volumes (NFS datastores) are located on a different site and set with a volume guranatee of 'none'. What's best practice for volume guarantee on SnapMirror destinations? In a DR event I'm obviously going to need to bring these volumes online so should I set them to fully guaranteed?

Thanks

[radek_kubka]

Hi,

SnapMirror target volumes inherit all properties of source volumes in a case of breaking SM relationship, so these settings at target filer are pretty much meaningless.

Regards,

Radek

[radek_kubka]

On a second thought:

I can imagine a situation where on a source filer volumes are thickly provisioned & aggregate is simply bigger, whilst on a target filer destination volumes are thinly provisioned & aggregate is smaller (possible only if actual data is not filling volumes). Then I am not sure what happens in a fail-over scenario & whether the fact that target volumes guarantees (changed from none to volume) cannot be fulfilled can break anything...

[aborzenkov]

Yes, I was about to mention it ☺ We got into such situation at least once. In this case destination filer fails to establish guarantees and your volume remains with guarantee none (or, better with volme(disabled)). This by itself does not cause any harm, but of course one has to be aware of it.

For this reason I prefer to set guarantee to volume now, when 7.3 made this possible.

[tvmo_tvmo]

I want to guarantee that in a DR situation my SnapMirror destination volumes will come online.

So are we saying I should set the SM destination volume guarantee to 'volume'? I'm not interested in saving space at the destination, the whole point of the destination is DR.

Thanks

[radek_kubka]

So are we saying I should set the SM destination volume guarantee to 'volume'?

Well, IMHO it only makes a difference *if* your source volumes guarantees are set to 'volume' and destiantion aggregate is smaller than the source one.

One thing though re target volumes size: things can go pear-shaped if source volume(s) have autosize set to on. They may outgrow destination volumes at one point & SnapMirror updates will start failing (unless you manually grow target volumes). One way to avoid this is to make destination, say, twice as big as the source volumes & (probably) set guarantees to none.

So back to square one

[jondtabar]

Your decision on what to set the destination volumes' guarantee to may depend on where you're replicating from.  If you have an environment with many filers at various remote sites replicating back to a single filer, you may be OK with a guarantee of "none" on the destination volumes.  The chances of all of your remote sites' filers failing simultaneously and requiring you to bring up all the volumes on the destination filer at once are nearly 0.  Though if you're only replicating from one source filer to one destination filer, I can see where you may want to choose a guarantee of "volume".  Also, don't forget about the "file" guarantee option that's available....that may or may not be helpful to you.

[tvmo_tvmo]

Here's my thinking.

I have two sites, both have live data. Both sites VSM to the opposite site. If one site goes down, the opposite site will become live for all data.

This is why I'm thinking about setting the SnapMirror volume guarantees to 'volume', so that I know exactly where I am from a space perspective and guarantee that if a site fails I can bring all volumes online.

Does that make practical sense?