VMware datastore recovery with 3par remote copy

[dennischen]

Hi,

I have a question about 3par remote with vmware.

The environment is, I have a two 3par storage(A->B), configured with synced remote copy. and with no peer persitence license. 3par A presents the VVs to vmware esxi hosts.  The VVs in 3par B keeps the synced status, and not present to the host. 

If 3par A unluckly crashed, and cannot be brought up. Can I failover the remote copy, and present the VVs in B the same ESXi hosts? Any configurations should be done in the ESXi hosts to bring up the Virtual Machine?

I know in the hpux host. I can do the vgimport the recover the vg. Not sure if it fits in the VMware environment. Thanks in advance! 

[Sheldon Smith]

The use of the auto_failover policy and the Quorum Witness require the Peer Persistence license. As far as I can see, the use of the path_management policy to support ALUA does not.
ESX ALUA support is provided by persona 11. Any older ESX hosts configured with the ESX legacy persona, persona 6, would need to be changed.

As long as the secondary volumes have identical WWNs as their primary counterparts, the secondary volumes can be presented to the ESX hosts prior to any failure. Once presented, the ESX hosts do a rescan and the additional secondary paths show as being in a "standby" state.
Once a failure occurs, all you would need to do is failover the Remote Copy Group(s).

And start the VMs back up which I would assume stopped after losing their access to the primary storage.

All of this sounds like an excellent reason to get the Peer Persistence licenses for both storage A and B. See the Implementing vSphere Metro Storage Cluster using HPE 3PAR Peer Persistence techncial paper, #4AA4-7734ENW.

[dennischen]

Yes, I know the PP is the perfect solution to VMware ESXi hosts. But the customer has only the remote license now. So the mannual failovered VV can work well with the ESXi hosts? 

[Sheldon Smith]

---

@dennischen wrote:

So the mannual failovered VV can work well with the ESXi hosts? 

---

As far as I am aware, yes. I would suggest the customer take a test server and set up a test RCG with a volume for datastore and see. A manual switchover might also work.