This plan ok?

[backupquestions]

Setup: 2 dell servers with lots of DAS. 60 TB volume formatted ReFS on each one.

First server is in local datacenter where production storage/vms reside. It will take 7 day retention of all systems via main backup job. It will then do a backup copy job to itself via another folder on same volume which will be 3 particular VMs with GFS enabled for monthlies and yearlies. All these will be spaceless fulls via block clone etc so no extra space will be used.

Second server is in remote datacenter but connected via 10GB fiber to main datacenter. This server will receive a backup copy job with 7 day retention all VMs. Then it will also receive backup copy job with 3 particular VMs with GFS enabled for monthlies and yearlies.

This way, if either server physically dies or the ReFS volume dies then I have not only my short term but also my long term retention on both systems. Does this seem like a good plan?

I ask because Veeam's recommended best practice seems to be only a short term retention so say 7 days on the first server, and then your GFS would go to the other server. To me, if the second server dies, I've just lost all my GFS archival retention, so why wouldn't I do a complete mirror of all restore points on both systems? Especially with ReFS, it is not even costing space! Or am I wrong about something here?

Lastly, with planning cloud backups, would it take essentially two backup copy jobs from first repo to the cloud to achieve getting everything up there? One backup copy job for all VMs 7 restore points, and then one copy job for GFS on 3 systems?

[backupquestions]

Another question I would have is when going to the cloud, does it make more sense that I should break up my initial main backup job into 2 jobs, one containing all systems minus the file servers and one with just the file servers(gfs required)? Or would I be able to leave the main backup job with all systems included and then separate out all cloud offloading via multiple backup copy jobs? Because it seems to me like each backup copy job requires 2 regular restore points, a full and an incremental, so you could end up with more than one full for the GFS VMs when your initial backup copy job already had that uploaded...

Like for example I have first backup copy job to cloud, containing all VMs, so that means a full of those 3 large file servers is already up there...

Now I make a second backup copy job containing only the 3 large file servers, that is going to make a new full for those 3 that I don't really want, and then also GFS restore points will create more fulls... I just want one full for all systems in the cloud and then only additional fulls would be the GFS points of those 3 large file servers.

[backupquestions]

Anybody? Bueller....? I know I rambled a lot and anything at all on any of my topics that you can address would be appreciated.

[bdufour]

you may want to throw some replicas in there at your DR site - if you have the storage for it. u can use a backup as a seed as well. i replicate all vms off site to a dr datacenter and also replicate critical vms at the primary site as well, to protect from os level corruption. for the critical vms, its nice to have a replica ready to go at the primary site vs waiting several hours to restore from a backup.

[backupquestions]

If you used instant recovery, or if you restored entire vm with the box "quick rollback" selected, a restore shouldn't be several hours I wouldn't think.

That is a good idea though on replication as well as the backups.

[bdufour]

if youre backing up to a dedup appliance it will, depending on the size of the vm and other factors. im actually happy with my restore results from dedup, but like having a replica of critical vms ready to roll at the primary site and dr site. also, instant recovery doesnt work that great if youre backing up to a dedup appliance (in my experience). doesnt sound like you are, but a lot of us are. regardless, replicas are important.

[backupquestions]

Ahh ok, I got ya. I'm looking at Exagrid because they have their appliance split in two inside it where you have a non dedupe "landing zone" etc. But they are expensive.

[bdufour]

exagrid is cool, i like their landing zone concept (def helps with restores) - have other admin friends that use them and like it. we use data domains with ddboost and are happy with the results.

[foggy]

Another question I would have is when going to the cloud, does it make more sense that I should break up my initial main backup job into 2 jobs, one containing all systems minus the file servers and one with just the file servers(gfs required)? Or would I be able to leave the main backup job with all systems included and then separate out all cloud offloading via multiple backup copy jobs?

You do not need to split your regular jobs. In backup copy jobs you can specify the VMs you want to copy individually as a source, not necessarily the entire backup jobs.