proxies, concurrent vms from the same job

[tomas.olsen]

Lets say I have 3 esx hosts. I create 3 proxies and create affinity rules to keep each one of them locked down to one host each. Lets say I have one big backup job. How can I utilize all my proxies at the same time with just one backup job?
Now it seems that the veeam backup server only starts backup of one virtual machine and uses one proxy at a time.

[Vitaliy S.]

Hi Tomas,

VMs are processed sequentially within the backup job. If you want to utilize all proxy servers available, you should be using separate backup jobs.

Thanks!

[tomas.olsen]

Thank you for your answer.
I actually suspected you would say that

But if I create several jobs, lets say one job for each host. And start them simultaneously, that will actually do the trick. The problem is DRS. All of a sudden maybe 5 of the vm's have changed hosts, and the backup jobs is no longer aimed towards one physical host.

I would love the possibility to utilize several proxies with one backup job.

[Vitaliy S.]

Not sure I fully understand what the issue is with DRS. Are you concerned about proxy server migration or migration of the VMs being backed up?

[tomas.olsen]

In my three host cluster I was thinking about making three backup jobs, making sure only vm's from one host was added to each backup job. So when I schedule one backup job with either automatic proxy selection or manual, I will be sure that the job contains only vm's from one host, and will be run with only one proxy. If I do this for all three hosts, creating one backup job for all the vm's of each host, I can run the backup jobs at the same time, utilizing all three proxies on all three hosts at the same time. Yet I all the proxies will only handle backup of one vm at a time to make sure I will get the best performance on each host, each proxy, each virtual machine.

[chrisdearden]

if your cluster uses shared storage (which it will if you use DRS) - then the hot add proxy can add disks from VM's its not homed on. Simply divide your VM's into 3 jobs , either by datastore or folder or by specific VM.

[Vitaliy S.]

Got it, though be aware that you can always keep your VMs in the folders and configure your backup jobs to backup these folders. This approach should do the trick with DRS "issue".

[Vitaliy S.]

Ahh....Chris beat me

[tomas.olsen]

Could this functionality be considered as a feature request?

That veeam could temporary disable DRS during backup, take not of where each vm was resided and then divide the job in several subjobs. One subjob for each proxy. And then do multiple backup jobs of vms of each host. And then, we could adjust each proxy as we do today when it comes to how many concurrent jobs each proxy would run. And the same limit still applies to the repository. If repository is set to four only four concurrent jobs (subjobs) is run. All other must wait in line.

[chrisdearden]

Disabling DRS during backups ? doesn't sound like a good idea to me , especially when you can potentially run backups 24/7 ( in different jobs ) - I can see where you are coming from , but I don't think thats necessarily the angle I would use for it. I don't think I would advocate a design that requires 1Vm per host to be deployed - you could end up with a large number of proxies to deploy which would not always scale that well. Prom a performance point of view I like to think of the target storage as the ultimate rate limiting step, I then design the proxy infrastructure and job setup to maximize that.

That said, non serial processing of VM's within a job has also be put in as a feature request to my knowledge.