Remote Copy snapshot CPG...

[T16]

Ive setup bi-directional remote copy between two storserves, I am new to this so please be gentle.

I have a CPG for normal snapshots, a CPG for remote copy replicas, and some other AO CPGs as well as the default ones on the array. So far so good. Say I snapshot a VV, I see the snap appear in the Copy CPG.

However, when starting Remote Copy, the snaps it creates of the source volumes DO NOT have any CPG defined on them at all. Its like RC has a mind of its own and creates its own special temporary snap outside of any CPG I can define. Array is an up to date 8200 btw.

Is this normal/expected behaviour?

THanks! its all a bit strange for me to see these temporary snaps, and the effect they have on the original volumes with (1) etc suffixed to the original voiume name. All this is hidden from view with vendors like DellCompellent, and you get pretty graphs and human readable names instead of all the rcpy.x.y.z stuff that I am seeing.

I can only imagine with say 100 volumes all replicating, there will be hundreds of snapshots at a time.

Thanks if anyone can clarify the above question.

[MammaGutt]

Yes, this is normal.

3PAR uses snapshots for replication. For sync you will never see a snapshop before the RC group is stopped (as it never needs to keep a list of unsynced changes). Async will create one new snapshot per volume for each sync period. Once the snapshot is synced and acked at the target, the source will delete it.

RC snapshots are stored in copy/snap CPG (see showvvcpg in CLI).

[T16]

Got you, thanks for this.

For some reason our RC snaps had a dash showing instead of a CPG.

What do the pros do with Async replication? Have a Copy Group for say 30mins RPO, 1hr RPO, then another for 4hrs perhaps? Or do they group them as HPE suggest with application/similar functions per copy group?

One thing I cannot get used to is that with our old Compellents, the replication would replicate all the snapshots for the volumes from the source array too, more like a virtual copy of each VV over the wire, which is more useful to us than just the latest version of the VV being copied over.

I had setup VVols with vSphere and came to the conclusion it was still not what I would call production ready. (IMO)

Really appreciate your replies, its tough setting it all up from scratch on your own!..

I liked the sound of Async streaming, but on an 8200 it seems it is not an option, plus we have 500mbps between our SAN locations, even though the latency is nice and low at 2/3ms.

[MammaGutt]

Async doesn't impact write latency.

Sync allows Peer Persistence, RPO =0, no bursts when suddently syncing a large snapshot.

Just to mention a few things that comes to mind.

Remember that sync period isn't the same as RPO. Sync interval at 10 minutes would mean RPO <20 minutes.

As for syncing snapshots, you could just create snaps on both systems.

As for async streaming it only has one advantage over sync, not impacting write latency as it doesn't have to wait for write ack from target, but there are multiple cons (no PP, RPO not 0, requires more resources to manage to mention a few).