Page 1 of 1

Stupid Question - 3PAR + VMware SRM

Posted: Wed Oct 30, 2013 2:38 pm
by Schmoog
Stupid question (Coming from EVA). On EVA, when used with SRM, the vdisks would be presented to the hosts at the destination data center on the destination (secondary, target, standby, pick your terminology) array. On our new 3PAR's, I notice that the hosts are allowed RO access to the destination (secondary, target, standby...) VV's. Is it the case that these should be exported to the host set at the destination site? Is it good practice for the destination hosts to have RO access to the luns (It doesn't seem like a good idea to me...)

Re: Stupid Question - 3PAR + VMware SRM

Posted: Thu Oct 31, 2013 6:29 am
by hdtvguy
You don't export the luns on the destination side. When SRM brings up a datastore it will leverage the 3par SRA and take a snapshot of the destination datastore and present that to the hosts as RW. We have 800 VMs on 167 replicated datastores using about 100TB on 3par using SRM and it works fairly well. A few caveats. First if a lun is replicating when you run a test or recovery the SRM recovery plan will fail. I have been pushing 3par to address this issue, but I do not believe they have any plans to do so. Also if you stop the replication while it is in progress you need to make sure you check the detestation 3par that it has rolled back the remote copy snapshot before you run SRM. If you run SRM right after stopping a replication that is in flight the SRA will take a snapshot of a volume that is in the middle of having its replication snapshot rolled back giving you extremely inconsistent and often corrupt results. I have really pushed 3par on this and again have gotten no indication they will be fixing that.

Part of the problem is how they do replication in 3.x - 3.1.2. They replicate direct into the base volume. Replication changes are coming in 3.1.3, but they will not address these issues with periodic replication.

Overall the 3par rocks in many areas, including SRM, with the exception that its replication technology is its biggest Achilles heel. I know much is improving in 3.1.3, but I still feel they have a long road ahead to really be Enterprise ready when it comes to replication. I am under NDA, but will say that in the next 8-12 months if they follow through on what they have shared with me, I believe their replication will be far more robust.