Failback or Perm Failover--> Rep ?

[ysm]

Hi folks,
I was wondering if Prod site and DR site has low bandwidth link, what strategy should I use for DR test?
1 use planned failover w/o perm failover, run on DR site for a few days, then failback to Prod site,
or
2 use planned failover with perm failover, run on DR site, initiate replication job back to Prod by mapping to backup copy found in Prod site until we are ready to failover again to Prod site?

Note the DR test period varies between 1-2 days to a week. And our bandwidth between the two sites is not great. Based on what I have tested, if using Method 1, failback will have to sync the delta. So if a few days have passed since the planned failover, the sync could potentially take very long to complete.
Whereas Method 2 seems to allow me to sync immediately after DR failover.
(background, we have 1 Veeam server on each site, but at any 1 time, only 1 Veeam server can do backup job i.e. when the VM in that site is in production use, we initiate the backup on that site)
So, do you guys agree that Method 2 sounds more feasible to shorten the sync time? Feel free to comment or suggest alternative ways.
Thanks.

[foggy]

The second method will also require sync upon mapping the job to original VM. However, the switch itself in the second approach will probably be faster, due to less amount of data to sync at the final stage (planned failover), since most of the data will already be sent during replication back to the original site (provided it was performed just prior to failover and not "in advance"). You can test both of them to compare the switch times.

Also, there's not much need for two backup servers, everything can be controlled by a single DR one.

[ysm]

I tested the first method and it is working. But I am having some issue with method two on the part where I need to map the backup as the source. I am pretty sure that there are restore points in the DR site but it could not detect it. Need more reading and checking on that.
You have a point on the redundancy of our backup servers. Indeed, in a stable operation day, only the DR backup server is probably required.

[skrause]

When you do a permanent failover, all of the snapshots on the replica you are failing to (target on original replication jobs) are consolidated. When you map back to the production site, just do replica seeding and search for the original VM. On first job run it will scan the entire disk for differences which takes a bit of time and then it will replicate normally.