Nimble dHCI Peer Persistence question

[jlangmead]

Hi

I would like to confirm if it is supported to configure a Nimble dHCI environment without using a quorum witness for Automatic SwichOver (ASO). In a scenario where deploying a QW isn't feasible will that prevent the use of a Nimble dHCI peer persistence to benefit from the synchronous replication features? All pre-reqs associated with Nimble model, Array OS versions and network latency etc will be met.

There is a blog here from from Fred Gagne HPE Nimble Storage dHCI and Peer Persistence that infers a QW witness is an option to provide ASO if desired - he says:

"In a few steps, you have two HPE Nimble Storage dHCI deployed with Peer Persistence on top of it. Now you can decide if you would like to use Automatic Switch Over (a.k.a ASO) or only use Synchronous replication. If you plan to use ASO, you need a third site where you can deploy a witness vm. In case you would so, please refer to the deployment guide on InfoSight available here."

I read the above to mean the QW isn't mandatory with a PP solution it just allows for automatic failover to be configured - that's what I would like confirming please.

many thanks in advance

[Sheldon Smith · Accepted solution]

If you want Synchronous Replication, (no automatic failover) you do not need the Quorum Witness.
If you want Peer Persistence, (automatic failover) you do need the Quorum Witness.

SR + QW = PP
Synchronous Replication plus Quorum Witness equals Peer Persistence.
There is no Automatic SwichOver (ASO) without having the Quorum Witness somewhere.

So: Do you want Peer Persistence?

Hope this helps

[jlangmead]

Hi Sheldon

Perfect - thanks for clarifying that. So it's just acase of getting the terminology correct. So I don't need PP just synchronous replication.... I think I read some previous literature saying Nimble synchronous replication was only supported for PP solutions so I had synchronous replication as synonmous with PP when it doesn't need to be.

Fred's wording is a little ambiguous though - he says

"In a few steps, you have two HPE Nimble Storage dHCI deployed with Peer Persistence on top of it. Now you can decide if you would like to use Automatic Switch Over (a.k.a ASO)..... if you plan to use ASO, you need a third site where you can deploy a witness vm"

But actually we don't have "dHCI with Peer Persistence on top of it" until we have deployed the QW (at this point we just have Nimble dHCI with synchronous replication) so there isn't really any decision to whether we want to deploy the QW or not as he suggests - it's a must.  Perhaps I'm just not reading this in the correct context.

[Sheldon Smith]

With synchronous replication if something happens you have to manually switch over and start things back up. The step from synchronous replication to Peer Persistence is deploying a Synchronous Replication Witness, Which seem to me like a very small step. The Witness responds quickly enough that things won't time out and stop. Either a VM or a RPM installed in a CentOS system.  Searching with Google, I see even a Raspberry Pi can run CentOS. 8^)

[Thomas_Lam_HK]

We always recommend having Witness setup and  with ASO enable to minimize the effort for manual recovery as the process is in fact quite complex.

https://infosight.hpe.com/InfoSight/media/kb/active/sup_KB-000372_Manual_Failover_for_Synchronous_Replication_Volumes.pdf

Plese review the document and it might help strengthen your mind.

[Sheldon Smith]

Goodness. KB-372 Manual Failover for Synchronous Replication Volumes shows all the effort the system handles having the Witness and ASO enabled.

@jlangmead, you really want to have Peer Persistence set up!