Best practice for file server cluster off-site backups

[smd32]

Hello,

I wanted to get some opinions on the following method of backing up Windows file server clusters at two different sites.

At site A, there is a two server file server cluster that replicates via DFSR to site B, where there is another two server file server cluster. The cluster at site B also has replicaton via DFSR to site A. There is some data unique to users at A, some that is unique to users at B, and some data that everyone uses. DFSR is continuously running.

Veeam runs locally at each site once a day and creates a backup of the local cluster. If DFSR has done its job and stays caught up, Veeam A's backup job will have a Veeam backup of A+B, and B's backup job will have a Veeam backup of B+A. No Veeam backup copy job needed to get things off-site.

Aside from DFSR lagging or totally failing and resulting in backups of newer data only being at one site and not the other, does this seem reasonable compared to backing up each site with Veeam and using a backup copy job to get backup data off-site to the other location? I like using backup copy jobs better because it doesn't rely on DFSR even though it could wind up with more disk space being used if I have no dedupe, and more network traffic. There is a 50Mbps WAN link between the two sites so bandwidth isn't that bad. It's about 3TB of data in total, not sure of the rate of change but it's probably not more than 2-3% a day.

My colleagues insist that the above method of relying on DFSR for in effect shipping backup data is fine, since normal production use by users needs DFSR to be working anyway. But this just feels somehow like it's not safe or enough.

Thanks for any suggestions, real world experience with this being ok or not, etc.

[skrause]

Just for clarification, does all of your data get replicated via DFSR?

We use DFSR to replicate between two datacenters and only backup the system drive on the secondary site while backing up the entire VM in the primary site. We then use backup copy jobs to copy the primary backups to our third site where we keep 5 weeks and send fulls to tape once a month. It works pretty well for us.

[smd32]

Everything gets replicated via DFSR. Veeam is backing up the cluster at A and the cluster at B, after A<->DFSR<->B happens.