Full recovery plan without reprotect (severed WAN)

[pstover]

I would like to sever the WAN connection between my 2 datacenters, run recovery, test basic functionality of all VMs and then shut down the recovered datacenter.  Since there was no communication between SRM to SRM and VC to VC, what will I need to do after shutdown of the recovered datacenter?  I was going to do the below:

- delete the Recovery Plan and protection group in DR

- power down all DR VMs and remove from inventory

- unmounts datastore from DR hosts

- get replication reset (Nutanix) in primary > DR direction as needed from DR

- reconnect the WAN

- get replication reset (Nutanix) in primary > DR direction as needed from primary

- ??? with SRM

What state will SRM be in after reconnect of the datacenters?  Will it see no recovery plan, no protection groups (deleted at DR) and I will just recreate the PG and Recovery Plan anew and go on my way?

[vbrowncoat]

Please read these two blog posts:

https://blogs.vmware.com/virtualblocks/2016/11/04/recovery-testing-srm/

https://blogs.vmware.com/virtualblocks/2017/03/02/srm_testing_pt2/

I would not recommend doing what you outline. More than likely SRM will be out of sync and it will quite likely require you to rebuild it. Regardless it isn't supported.

I highly recommend using the test functionality built into SRM. If the SRA supports it, it can be done with the sites disconnected.

[pstover]

I am surprised there isn't something that can be done to keep SRM in sync after a site disconnect test like this, based on numbers of forum posts on this topic and a customer request to fully test DR but leave all production datacenter systems in exact state.  The array replication can be put back in place easily enough with vendors like NetApp.  But I think there needs to be something in SRM like the test RP cleanup task where you could clean up a real RP and have SRM understand that a recovery plan was run but it was rolled back when the 2 SRM servers re-communicate.

[vbrowncoat]

It is something we're looking into doing. I can't provide a timeframe but want to be clear that it is being evaluated/investigated for a future release