SRM and MirrorView: Avoiding a Full Sync after Failover

Question

Hi, all, We are working with a customer's MirrorView/A + SRM + VNX deployment.  To simplify operations, we want to avoid a full synchronization after failover.  However, every time we do a failover the entire data set at the protected site is copied back to the original production site. We have tried inserting a pause in the recovery plan after all VMs are shut down.  We have confirmed that the consistency groups change to 'synchronized' state, which MV docs say should avoid the need for a full reverse synchronization.  But, the sync is still happening. We even did a test without SRM and verified that after going to synchronized state we can promote the consistency groups at the protected site and no full sync is needed.  But with the SRM/SRA we cannot get the solution to do an incremental re-sync (not a full sync).  It almost looks like the SRA with MV/A it is not possible to avoid a full sync. Can anyone tell me if SRM with MV/A is capable of a failover without a full sync?  If it is, what do we need to do to make this happen? Thanks, Scott

reseach · Answer

Scott,

Here is some my thought for your reference.

Unlike MirrorView/S, MirrorView/A uses scheduled or manual updates to replicate the changes between primary and secondary images. Recovery operation (Consistency Group or Mirror Promote) is not allowed while a group or mirror update is in progress.

You can insert a message step in the recovery plan (before "Prepare Storage") to check the mirror state or to manually synchronize the mirrors using Navisphere. You must wait until the mirror synchronization completes before resuming the execution of the recovery plan.

Eddy

GearoidG · Answer

Just an FYI Mirrorview Cluster Enabler works in a similar fashion doing force promote - so maybe its a developmental thing that I am not privy too. Gearoid

GearoidG · Answer

So its doing a MV/a Force Promote (ie a full sync) I dont know at this stage is this by design of SRM - but if you were looking at a Mirrorview, we would normally recommend the luns were in a consistent state before doing a promote, no clue why they are looking to do a force. See p 759 of the VNX CLI guide

https://support1.emc.com/docu31547_VNX-for-Block-Command-Line-Interface-(CLI)-Reference-1.0.pdf

Gearoid

Andy261983 · Answer

Thanks Scott for starting this discussion on the forum to get a few more eyes working on it.

Appending to your description I would like to add an excerpt from the SRA logs

2012-04-11 08:52:30,069 [com.emc.mirrorview.platform.naviseccli.NaviseccliConnection]: 10.10.8.117 Executing command: mirror -async -promoteimage -o -type force -name "LUN 167 TGTEST_PRAAPP01" -imageuid 50:06:01:60:C4:60:26:68

Does it because of the "-type force" is it doing a complete sync at the later stages?

Regards

Anand.

VMware

Was this post helpful?