I've read the Admin Guide for vSphere Replication but it's not super-detailed regarding this (multiple disk considerations). I did, however, note a couple of caveats regarding virtual disks spread across 2 or more data stores which I (mostly) understand but I'm seeking more feedback from those of you who have dealt with multi-disk (and big ones at that) VMs and vSphere Replication between a primary and secondary/recovery site.
Let's say I have a source VM named 'FRED' with 5 virtual disks. 2 of the 3 VMDKs are located in the same datastore ('DS1'), and are named FRED and FRED_1. The 3rd and 4th disks reside in datastore 'DS2', and the 5th disk is in 'DS3'. So, my VM's disks look like the below, and note the file sizes of each:
\DS1\FRED\FRED.vmdk 50GB
\DS1\FRED\FRED_1.vmdk 700GB
\DS2\FRED\FRED.vmdk 700GB
\DS2\FRED\FRED_1.vmdk 1TB
\DS3\FRED\FRED.vmdk 800GB
Now let's say I want to pre-seed in the future 'recovery' site without shutting down the source VM or running Converter. In order to pre-seed, I will be running restores from Commvault and Data Domain directly from within the recovery site across 3 datastores (we only provision 2TB max datastores as our standard), spreading out the VMDKs as with the source. I'm just not entirely sure what I'm up against in this scenario when pointing to the destination datastores for each disk within the VM while configuring for first-time replication.
KB 2012610's Method 1 won't work for us because of the 2TB limit (our standard) won't 'fit' all the disks. Method 2 isn't desirable either because we don't want source system downtime until the cutover/recovery date is set (not before for renaming the disks to be unique). So, that would leave the WORKAROUND, perhaps? When configuring for replication, point to VMDKs of the same names, but distributed across the 3 like-datastores on the replica/recovery vCenter end? Thoughts?
In vSphere Repliation,if the source virtual machine has disks with the same names but located in different datastores, then you can choose to place each disk in different datastore or different folder on the recovery datacenter so that the disk path is unique.
Okay, so when pre-seeding, I just need to make sure that the destination disks are in separate directories and have unique names in each. Also, it would seem that for pre-seeding to prompt me, the disk names must MATCH EXACTLY on both ends... correct?
Yes, name of the source vmdk and the target vmdk that will be used as seed must match exactly. Also ddb.uuid values of the source and target vmdk must match.
The UUID factor is worthy of note... thanks for that. I searched further and found this: vXpress: "Target disk UUID validation failed" Error while configuring vSphere Replication ...
However, when I restore the VM's VMDKs from backup, into the target site datastores, I won't be registering those VMs and starting them up and answering the question incorrectly such that the UUIDs would be regenerated---so in short I won't need to worry about that.