How to protect large cluster

[samh22a]

Hi,

Currently looking at Veeam for an internal project to protect a big Hyper-V cluster (5000 Vms).

I can't figure out how to automatically protect newly created VM, and to split the backup load onto multiple jobs..let's say 50 VMs per job.

Is there a way to specify 1 job per CSV ? And then automatically detect newly added CSV ?

Thanks,
Sam

[Gostev]

Hi, you can populate the job with hosts and clusters (in addition to individual VMs). Thanks!

[samh22a]

Hi,

Isn't this not very optimal ? as VM changes very frequently on a specific host..the same job may have 100 vm one day, and 100 totally different vm the next day ?

[Gostev]

Yes, it's not very optimal for highly dynamic environments. For those, creating the job on a cluster lever is better.

[samh22a]

Make sense.
I'm still new to Veeam..so I'm trying to understand how everything works together.

As for the job at the cluster level: can we have multiple job to split the load at the cluster level ? If there's 5k vm within my cluster, I really doubt a single job can handle it all.

[tsightler]

For more sophisticated management of a large cluster, the best option is probably to use Powershell, potentially in concert with some orchestration, to automate the creation and maintenance of jobs. It should be fairly easy to create a Powershell that list VMs for each CSV and adds them to a job, although I'll have to admit I've only done this for VMware. I would guess the first step would be to grab a list VM objects with Get-VM, then the list of CSV objects with Get-ClusterSharedVolume. At that point simply looping through the CSV objects to determine which VMs live on them should be a simple matter of comparing field (I think the VMCPath property of the VM object would be matched against the FriendlyVolumeName property of the CSV object).

So once you have the list of VM objects it should be pretty easy to create the job and/or add VMs to an existing job. If you use some orachstration you can even trigger a script to add a VM to the appropriate job automatically as soon as the VM is created, otherwise a simple scheduled run of the script each day would pick up new objects and create new jobs.

That's the conceptual idea anyway. I've seen this done for plenty of environments with VMware so I don't think it would be too difficult with Hyper-V, but I'll admit that's a platform I'm still ramping up my knowledge on.

[paulhardy]

Is there anyway of grouping the VM's in some sort of folder structure like you can with VMware which Veeam can then use as the backup selection criteria?

[Vitaliy S.]

Hi Paul,

Not sure there are native ways to do that. The only workaround I can think of is hosts group, but again it is not the same as VMware VM folder.

Thanks!

[Gostev]

Unfortunately, unlike VMware, Hyper-V does not provide VM folders. But, we have some ideas on how we can resolve this in the future versions.

[samh22a]

Hi,

What would be the maximum recommended number of VM per job when using a hardware VSS provider ?

Until Veeam can work with folders or CSV, I think I'll have to do it manually.

Being able to create 1 job per CSV from the GUI would be ideal in my case.

Thanks

[Gostev]

Hello, there is no limit on number of VMs per job (no matter of VSS provider). Thanks!

[foggy]

Just take your backup window as well as VBK size and dedupe rate considerations into account when combining VMs into jobs.

[jake1110]

Definitely break the jobs down by CSV, makes it easier to run the jobs concurrently. Although there's no hard limit to the number of VM's a job can have, I always experienced performance issues if added say 50+ VM's in a job. Also, if you have too many jobs running concurrently on a Veeam server, performance tends to suffer from what I observed. For my 250+ jobs(remote sites), I broke down those jobs into three different Veeam management servers to balance the load. I tried kicking off the jobs at once on one Veeam server and it choked, even with excessive hardware specs. After splitting the jobs into three servers, I now run 25 jobs at a time each starting within 30 minutes of each other. Works well for me.

Good luck with a cluster that has 5000+ VMs though. It's going to take a lot of fine tuning on your part to make it work.

[samh22a]

How did you split the job per CSV ? Did you used a custom script ?

[veremin]

If you want to create backup jobs, using CSV as their source, then, the best option will be PS scripting, indeed. Especially, in case of large clusters.

Should any assistance with PS scripting be needed, feel free to create a topic at the corresponding subforum.

Thanks.

[bungle]

I am using VBR ver 7.0.0.871 and have added a Hyper-V Cluster ver 2012 R2.

From the Virtual Machines tab > View I can change to datastores and see the CSVs presented to the cluster.

I want to create a backup job against a datastore rather than cluster, host or selected VMs but I cannot seem to find the option in job settings can someone please let me know if this is possible.

[Vitaliy S.]

Hi Mark,

Currently, hosts, clusters or individual VMs are available as source for the backup job. How many CSVs do you have per Hyper-v Cluster?

Thanks!

[bungle]

At the moment 5 CSVs, wanted to split the backup jobs across 2 repositories as not enough room to fit all VMs on one

[Vitaliy S.]

Got it, it seems like it could be possible to create jobs based on CSV containers via PS script (according to Vladimir's comments), though I'm not a great expert in PS.

[bungle]

Okay thanks I have found a workaround, from the Virtual Machines > View Datastores select the corresponding CSV and expand to list the VMs it contains, Ctrl-A to select all VMs then right click and add to new backup job, name the job with the corresponding CSV name.

Its not dynamic so spinning up new VMs on a particular CSV wont be protected by the corresponding job so will have to be a process in place so new VMs are added manually to backup job.

But at least I have grouped my VMs by CSV so they can be sent to different repository.

[Vitaliy S.]

Nice trick