Exchange 2019 Backup: Hyper-V Guest or Failover Agent

[ruddj]

We have just upgraded to Veeam 10 and I am trying to determine the best way to back up our two Exchange 2019 servers running on Windows Server 2019. They operate in an Exchange cluster with 2 databases as part of a DAG with each server having 1 active and 1 passive. They run on 2 separate Server 2016 Hyper-V hosts.

For backup we seem to have two options:

• Use Hyper-V Backup on the hosts to backup the guest VMs with Application-aware processing enabled. This can be run as a single job for better deduplication or as separate sequential jobs to avoid snapshotting at the same time but using more storage.
• Install the Windows Agent on the VMs and use the Agent backup with Failover Server chosen on the DAG group. (Failover Agent install: https://www.veeam.com/kb2463 https://www.veeam.com/blog/windows-2019 ... agent.html )

Testing appears to show Hyper-V is faster as it processes both nodes at once while Agent method backups up the nodes sequentially by making the Exchange DAG resource unavailable on second node while first node is getting backed up. However, some community threads recommend backing servers up sequentially instead of in parallel so this may be better.

I have read the Best Practise but it is extremely basic and doesn’t really recommend one over the other https://www.veeambp.com/applications/exchange

The instructions on backing up DAGs are from 2014 and don’t mention the Agent option at all. https://www.veeam.com/blog/how-to-backu ... ation.html

I have been reading through as many community threads as I can, but most are for much older B&R versions, or just say to read the Best Practise link above.

We use Socket licensing and with version 10 as the Hyper-V host is already backed up, the agent license is covered, so no difference in price.

Does anyone have any suggestions of which option is better or any other pros/cons I have not considered?
Thanks.

[HannesK]

Hello,
and welcome to the forums.

There is no "definite" best practice. If you don't have any failover issues, then I would go with VM backup (easier, keep it simple). https://www.veeam.com/kb1744 recommends multiple jobs. So the deduplication argument is not relevant.

Agents are usually used, when snapshots result in unplanned DAG failovers.

Best regards,
Hannes

[ruddj]

Thanks Hannes, as we are using Hyper-V with production checkpoints rather than vSphere we don't seem to have the 'stun' period triggering failover, so seems OK with with going with the simpler VM backup.
Thank you.