I've seen a few questions recently with regards to setting up shared-recovery with vSphere Replication and thought it'd be a good time to hit on one or two fairly straightforward ones.
When setting up a many-to-one scenario with SRM and vSphere Replication, how many VRMS servers need to be set up to support the N:1 model?
The question really is: In order to have a shared recovery model where multiple sites fail over to a single recovery site, do we need to set up a unique VRMS management framework pairing for *each* site, with the recovery site having a unique and dedicated VRMS for every site that is failing over to it?
The answer is you need a single VRMS server paired with *each vCenter Server*. So if you have 3 sites, each with a VC and each failing over to a 4th, a shared recovery site, you will need a single VRMS at each site including the shared recovery site for a total of 4 VRMS instances.
The VRMS pairs with a single vCenter and in the N:1 model must be jointly seen and shared by each site's vCenter/VRMS pair.
You can have a number of separate VRS (vSphere Replication Server – the target for the changed blocks) for each site if you wish, but only a single VRMS. Keep in mind there can be a maximum of 10 VRS targets deployed per VC/VRMS instance.
This leads to another question I've been asked a number of times:
If I've deployed a number of VRS target appliances at my recovery site, how does the "auto-assignment" work when configuring vSphere Replication for a virtual machine?
When configuring vSphere Replication for a VM you may have noticed that you have the option of specifying a VRS target, or allowing it to be auto-assigned.
How does auto-assign work?
Very simply – if there are multiple VRS destinations then vSphere Replication will automatically choose the VRS target that has the fewest replications configured at the time in which configuration of the VM is being done.
