High IOPS VM Server Backup

[Lambpix]

Hi,

I have had a question from one of our departmenst about backing up their public facing VM's using NetBackup, the servers have very high IOPS and they do not want to have any interuptions during backup.

We currently use the VSphere API on our infrastruture VM's using VIP's but they are concerened that the server would beinterupded during the snapshot process, not sure what is the best way to bak these servers up without any performance hit on the severs.

I could use non-quiescent backups but these would not be crash consistent 

Any advice would be helpful.

Thanks

Kev

[Marianne]

Please have a look at Resource Limits:

https://www.veritas.com/content/support/en_US/doc/21902280-127283730-0/v47183490-127283730

You can limit the amount of concurrent backups per datastore, per ESX Server, max concurrent snapshots per vCenter, etc. etc...

Have a look at the list of resources that you can limit, then discuss with your VMware admin which resources to limit and what the value should be.

[Mouse · Accepted solution]

I concur with @Marianne - your first port of call has to be resourcing the backup appropriately. Unfortunately VIPs don't deal with IOPS provisioning as such but leaving as much resouces as possible to the backup process.

However, for a really high-transactional VMs with heaps of IOPS on its own exclusive datastore where no other VMs running, a VM stun - the effect when snapshot consolidation or removal takes place, can be a real thing. This is a feature of the current state of VADP and known to VMware.

There is a solution in the works inhouse between Dell and VMware that will enable another API, from what I heard, initially available exclusively to Dell, that would deal with this issue. Not sure if it will be portable across all datastore types or just vSAN.

Now what other people do as you might ask. The best solution to this issue so far is to do in-guest backup using agent, because the OS inside the VM deals with resource allocations so there will be no stun. Yeah, you won't be able to quickly restore the entire VM but it's the trade off.

[Lambpix]

Thanks Marianne, we are currently using the resource limits on our current set up and that is working fine for our current needs, I think they are being a bit more cautious with the public facing, high IOPS, servers and do not wnat any diruption at all so snapshot may not be the best metod of backups for these.

Kev

[Lambpix]

Cheers Mouse, 

I think you maybe correct and going down the agent backup rather than the VM API route if they want to avoid any blip in the server.

Kev