Architecture Advice: Replicated SAN, snapshots & offsite bck

[cjack03]

Dear All,

Firstly thanks for such an amazing forum and all of the information that is available here, it's a fantastic resource.

I have a question to put to all of you about a future Veaam B&R install we are planning to implement in the next few weeks. We are about to replace 5 legacy SAN bays with 2 HP 3PAR 7200 bays. There will be one bay in our primary DC (DC1) and the other in our secondary DC (DC2). DC1 and DC2 are on the same campus connected by fiber.

We plan to replicate all of our vcenter (vsphere 5.1) datastore LUNs from DC1 to DC2 for DR, using the SAN hardware, as well as in order to be able to use HP Peer Persistance technology. We will reuse one of the newer legacy bays as a repository for Veaam B&R at a 3rd site on the same campus. My idea was to use physical direct SAN proxies for Veeam to take the real backups of our VMs from the replicated LUNs in DC2 and take regular snaps of these replicated LUNs for use with explorer for SAN snapshots. The idea being that the primary production storage would never be touched/impacted. However I think that this wont be possible as Veeam needs to use vcenter for the initial Vmware snap of the VMs. Am i correct in thinking this, maybe i'm completely lost

How would you experts go about implementing the perfect Veeam install in such an architecture

Thanks so much for any help you can give me!!

[veremin]

Am i correct in thinking this, maybe i'm completely lost.

Yep, you’re completely right in your assumption. Even with backup from storage snapshot, Veeam still requires normal VM snapshot to be taken, though, it’s just kept for seconds, until a storage one is created.

The more detailed description of how this functionality works can be found in the Veeam blog:

When this box is checked, we take a VM snapshot just like we do today, but then we take a storage snapshot. The storage snapshot includes the VM snapshot, so we can release the VM snapshot on the production datastore immediately after, and create the backup from the VM snapshot inside the storage snapshot. We create the backup in exactly the same way as we do today: We perform advanced application-aware processing (including log truncation). We use VMware Changed Block Tracking (CBT) to speed incremental backups. And you can perform Instant VM Recovery, SureBackup recovery verification, Universal Application-Item Recovery, etc. from backups created from storage snapshots. The only difference is where the VM snapshot is read (from the production datastore or from a storage snapshot) and how long it’s open.

Hope this helps.
Thanks.

[cjack03]

Thanks for the reply! So in my case this means I must take all my backups with Veeam from the storage bay in DC1 which is the production storage presented to my vsphere cluster?

[veremin]

Providing that Virtual Machines at DC2 weren’t visible in vSphere, it wouldn’t be possible to backup these VMs up. Thanks.

[dellock6]

Regardless if the storage is an active-active system or also called a stretched cluster (which means you can do vmotion from DC1 to DC2 without downtime) or DC2 is a standby replica of the DC1 storage (and VMs are always executed in DC1 unless a failover activity happens) vCenter and ESXi servers only see 1 version of the storage, and data needs to come out from this version.
So I do also agree there would be no real possibility to split the load on the two storage system and delegate backup activities to the DC2 array... even if it would be indeed a cool design.

Luca.

[cjack03]

@dellock6 and @v.Eremin thanks for you input it has helped me clarify our backup architecture, maybe one day i'll be able to do my back ups from the replicated LUNs but until then we'll stick with what works best!

[veremin]

You’re welcome. Should any additional help be needed, don’t hesitate to let us know.

[cjack03]

Dear All,

Just a quick update in case it can help anyone. We have implemented our new 3par arrays. Primary array in DC1 holds all vmware datastores, this is replicated to DC2 and the replicated stores are also presented to the vmware hosts using HP's peer persistance technology. The replicated LUNs have the same WWN as the primary array LUNs but are presented as read only. I have presented this secondary array to Veeam B7R v7 in the SAN infrastrucutre. Now when I do a backup job the snapshot is actually taken on the replicated LUN and this snapshot is presented to my proxy.

I have had a few problems and have a ticket open with support due to the snapshot disappearing occasionally on the proxy after a rescan of the FC bus. This means my backup jobs are not 100% reliable YET but i'm confident we'll get there.

I am very very pleased with this setup, please feel free to contact me if anyone out there would like some more info.