Feature request: Replica fail back to production timing

[nismoau]

Hi,

We've recently implemented Veeam replica, and one thing I noticed during testing is that when you go to fail back to production, it occurs in 2 passes. Pass 1; the running replica is online and able to service requests whilst the delta is being compared to the original source VM, and Pass 2; the running replica VM is powered off, and the final delta sync to the original source VM is made, after which the original production VM is powered back on.

What would be nice, in terms of new features is before Pass 2 is started - which is the start of the outage to the server - is some sort of dialog that pops up so that you can schedule the outage with the business etc.

I understand that it's likely not ideal to be in a 'partial failback' state (i.e. between Pass 1 and Pass 2) for very long, but in terms of co-coordinating outages this would be great.

End of the day I also understand that it's difficult to predict how long the failback to production will take, as factors such as storage speed, amount of changes to the system during in failover state, time spent in failover state etc come into play but from an outage to the business perspective it would be great to be able to communicate when, and approximately how long, the outage will be.

Cheers!
Justin

[veremin]

We understand your feedback. Nevertheless, it will be really hard to predict time required for failback operation (even approximately) due to the number of variables present.

You can replicate failovered VMs back to production and, then, execute planned failover to have more control over the process as well as more accurate prediction regarding the time needed.

Thanks.

[nismoau]

True, you could do that, but it's very manual.

In Hyper-V replica, you power off the VM first at least, therefore you know when the outage will start (not when it will finish, as same applies whereby it depends on speed of storage, amount of data to be replicated etc).
https://youtu.be/GP49y92TYVA?t=225

[mcz]

Hello,

I would like to add my feature request to this topic. As mentioned before, planned failover (or failback) is performed in two stepts.
When I used the planned failover the last time, I tried to save as much time as possible and therefore I shut down the os before I performed failover.

But veeam didn't notice that the vm has been shut down and started the second replication pass which had a delta of zero (what else). So the whole failover/failback process could be accelerated If veeam noticed that the vm is not running that only one replication pass would be neccesary.

I know that it doesn't make any difference in terms of availability of the vm's (because there will always be a point where the vm isn't running and replication pass is being done) but the whole failover/failback process would be sooner finished.

Thank you,

Greetings
Mike

[fcardarelli]

Hi,

What I was wondering if there is any way to synchronize disks from cloud vm back to production before doing a failback? Because we can’t calculate how much time the failback will take and how much data will be written.
Once we have the vms on a failover mode (Customer migration Cloud) there is no way replicate in the other way (Cloud to Customer) until except a failback to production is done.

During the process is going to calculate signature for the disk and we do not know how much time it will take on a 1TB disk, or bigger disks.

After calculating signatures and having replicated changes the VM is failedback to production. But the time is unknown, I can’t decide at what time the vm must be powered off, it just does it at the end of calculating the signature and replication changes.

The problem that I see on this process is that Cloud Connect might power off a VM during a peak hour, because it does it automatically at the end of signature and replication.

The feature request i want to submit is the possibility to replicate changes during a period of time until the vm is fully synchronized so the downtime of the vm is during a schedule window and not random.

[mcz]

I totally agree with you!