WAN link failure scenario

[gparker]

Hi,

I'm about to deploy Veeam v6 to replicate some VMware VM's from one site to the other. What happens if the WAN link fails during replication? When the link is restored, will the replication job(s) resume from where they left off? Is there any manual intervention required?

Regards,
George.

[Gostev]

Hi George, no manual intervention is required. The job will automatically kick in the retry cycle (according to its schedule), and will start processing VM on which it failed over again. Thanks!

[gparker]

Will it start processing the VM on which it failed from the beginning, or will it compare the source VM with what's been transferred already to the target and continue from that point?

[foggy]

It will start processing the VM from the beginning.

[gparker]

Wow, what sort of a DR product is this? Are you saying that if I'm replicating critical VM's to my DR site and the WAN link fails in the middle of the night or example and the replication has to start from scratch the following morning and 1/2 an hour later the production site dies, I won't be able to start the replicated VM in DR because it was still replicating!

[J1mbo]

Veeam replication includes definable restore points, for example you might choose to have 5 or 10 or whatever available at the remote site. Only the latest replica would be unavailable, of course because it hadn't finished. Have a read up on VMware Change Block Tracking for info on how the replication works.

[Gostev]

The response from Foggy is simply incorrect (or may he is talking about different thing) - only difference (delta) between source and replica disk states is always transferred, VM will NEVER be processed "from beginning".

Our replication is truly forever incremental, so at any point only delta between the latest "good" (consistent) state of replica and source VM is what gets transferred. If the link dropped during transferring of the specific delta, that part of the delta that was already transferred is "lost", and the next job cycle will transfer that delta again (there is no "resume" on that delta transfer). However, you are still always able to failover to the "last know good" state produced by most recent successful replication run.

To give a visual explanation, let's say the first letter is source VM disks state, second letter is replica VM disk state:

A = A (seeding done, replica VM is the same as source VM snapshot)
Source VM disks state changes from A to B
B > A (AB delta is transferred to replica)
B = B (replica VM is now the same as source VM snapshot)
Source VM disks state changes from B to C
C > B (BC delta is transferred to replica)
Link drops during CB tranfer!
Source VM disks state changes from C to D
D > B (BD delta is transferred to replica)

[foggy]

I didn't mean that the previous restore points were dropped, of course. Just thought that George asked about resuming transfer during last restore point processing (as he mentioned this in his first post here), and there's no resume functionality in VBR currently. Sorry for confusion though.

[gparker]

Thanks for all your comments and clariffications.

I should have elaborated a bit more:- I was reffering to "what happens if the link fails during the first replica/seed job". I understand that once the first job has succeeded and I have a full replica copy of my source VM, I can use the resore points feature etc, but my concern is around if the WAN link fails during the very fist replication job and the answer seems to be that the job won't resume from where it left off but it will need to start from scratch.

George.

[J1mbo]

You can seed the job via removable media if you prefer.

[gparker]

Yep, thanks for that

George.

[Gostev]

my concern is around if the WAN link fails during the very fist replication job and the answer seems to be that the job won't resume from where it left off but it will need to start from scratch

I'm replicating critical VM's to my DR site and the WAN link fails in the middle of the night or example and the replication has to start from scratch the following morning and 1/2 an hour later the production site dies, I won't be able to start the replicated VM in DR because it was still replicating!

Well, in that case your concern is totally not valid. What if the production site died 1 or 2 week ago, when you did not even have Veeam deployed? You are essentially asking "what if my production site fails and I am not ready for that yet". You should have been ready for that months ago!

As far as doing the first sync over WAN, internally we do this for individual VMs as needed (we are replicating across Atlantic over 6 Mbps link). Works fine most of the times, I remember link dropping may be twice in the whole history during this operation. However, when we need to do this for many VMs, we of course just FedEx the USB hard drive of course (otherwise, it will take about 1 month to complete over WAN even without any disconnects, and this is unacceptable of course).

No doubt resume would be "nice" to have (just like hundreds of other pending features), but I do not see how resume will affect our current strategy or give us significant benefits. Links are pretty reliable these days, even when replicating across the globe - resume is no longer as critical as in dial-up world. Bandwidth, on the other hand, is and will remain an issue for most, so seeding via removable media is here to stay and is the proper way to do the first sync.