Several Veeam questions

[marieelieg]

Hi everyone, hope you are doing well.

I have several questions regarding Veeam. Hopefully you guys can help me.

First, I have a hyper-v cluster environment which contains several CSVs. Whenever a backup is run against a vm in a volume, the whole volume (all vms in the volume) gets affected performance wise. Is this behavior normal?

Second, I have also replication to DR. Whenever I do replication or failover, the vm in the DR must be off and the VM in production can be on. When I do failback to production both VMs are off and the syncing processes are taking a lot to finish (each sync phase 20 min for around 30GB vm so you can imagine if it's a 200-300 GB vm). Is this behavior normal? Can't we keep the DR vm running or the production vm running or both in failback situation? And if we use wan acceleration how much time will I be able to save approximately for both failover and failback?

Third, I want to know how much toll it will take on my production environment if I do replication directly from live VMs and not from backups. So can you please post a link here to the requirements for replication using live vms? (in the meantime i will keep searching).

Thanks everyone in advance. Your help is much appreciated

[Mike Resseler]

Hi,

What version of hyper-v are you running? Is it windows server 2012 R2? The backup architecture is different in 2012 and before that, in 2012 R2 and finally now in 2016 it is completely changed. Let me know the version and I'll try to explain as good as possible what happens when you take a backup so you can understand the consequences of it

Thanks
Mike

[marieelieg]

Hi Mike.

Hyper-V cluster is 2012 datacenter.

Thanks.

[Mike Resseler]

2012 uses the same framework as 2008 and 2008 R2. The first change to the framework was with 2012 R2 and now with 2016 it is completely changed. What happens (in short)

Veeam asks the Hyper-V VSS writer on the host to get everything ready for backups. Hyper-V VSS talks to the integration components inside the guest and those also do the entire VSS work. When everything is ready, there is a guest snapshot. After that, Hyper-V VSS (on the host) gets informed and prepares to do a host snapshot.

Now here is the problem. The host snapshot is a volume snapshot which means each VM on that volume is affected in performance. Secondly, the actual VM that you wanted to backup has a guest snapshot but it was earlier than the host snapshot so there is inconsistency. This means the VHD(x) needs to be mounted again and through VSS rolled back to the exact moment of the guest snapshot.

All these operations are not good for the performance of your VM's

Hope it makes sense

Mike

[marieelieg]

Thank you Mike for your feedback.

Any thoughts on my other questions?