Heads-up on fail-over plan when multiple replicas exist

[pkelly_sts]

[Originally logged as case 01861305]

This is just a courtesy heads-up on what I would describe as unexpected behaviour, even though technically the logic is kind of sound (in an unexpected way in my opinion) so as a result the documentation is apparently going to be updated.

I have a small site with very little local storage but with, effectively, an active/passive pair of ESXi hosts (due to lack of vMotion licensing etc.) hosting around 5 VMs.

My DR plan for that site has been to send backups directly off-site (due to not enough local storage) plus a LOCAL replica from host 1 (active) to host 2 (passive) keeping only 3 RPs.
I also have a SECOND replica job going off-site to the main office, created from the backup that's sent to that site, so that's two replica jobs for these VMs.

I had a complete host1 failure on that site recently so we chose, fairly obviously, to fail-over to the passive host.

To do this I decided to create a quick Failover plan so I could just place the machines in startup order & let 'em churn away.

Whilst creating the plan, I very specifically selected/added the replicas in question, on the passive host, but when I started the failover plan, I found that it was actually starting the other replicas back at the main site which was obviously not what I wanted.

I repeated the process twice more, the last time with a colleague watching over me to ensure I wasn't crazy but the behavior persisted so I ended up just failing them over manually & logged a case later.

After some head-scratching & internal testing today, I got the following feedback from support:

"We elaborated this internally and our documentation team did the corrections to the UG.

It turned out that no matter from where you've picked the VM - from the infrastructure or from replicas. It will by default failover to the most recent snapshot created while replicating the source VM:
https://helpcenter.veeam.com/backup/vsp ... n_vms.html

This applies unless you specify to the exact point in time where you want to failover to.

We appreciate your help on this matter since our documentation was not clear enough on this feature."

I replied that the absolute logic kind of made sense, but the usability side of things doesn't, e.g. if you're selecting a specific replica instance, then that's the one that you'd expect to start.

Anyway, at least until documentation is updated I thought I'd spread the word here in case it caught someone else out...

[foggy]

Paul, thanks for paying our attention to this inconsistency and for sharing with the community. Much appreciated!