HPE Storage Users Group

Hello all,

I have recently setup a P10400 (or V400, not sure what the difference is?) anyway, I am looking at the vcenter integration options.

I noticed that I cannot create a virtual copy of a datastore that has snapshots? How do others work around this or implement it in best practice? I am just trying to understand why there is that limitation and what others do in their environment to leverage this license and technology?

We don't use the vsphere integration because I haven't found a good use for our environment. For backups, we use NetBackup and use SAN based backups in that manner.
In short, NetBackup has the same datastores mapped to the media servers that the ESXi servers do. Then netbackup interfaces with vsphere to create a snapshot per VM, then grabs those files direct from the datastore via FC.

not sure if that help you.

I would say use the Recovery Manager for VMware product with the vSphere environment. If you use application consistent snapshots, RMV will take a VMFS snapshot, then take a virtual copy of the volume, and then delete the VMFS snapshot. Honestly I think this would give you better performance and scalability. VMware can only handle 32 VMFs snapshots, and you lose IOPS/performance with each VMFS snapshot.

Could you please clarify a few things for me, pardon my ignorance on the subject...

I am using 3PARs Recovery Manager, or at least trying to. It tells me it cannot make a virtual copy of the datastore due to a snapshot existing.

Are you saying I have to use the option, "Quiesce guest file system" when making the VM snapshots? Then Recovery Manager will work?

Also you say vmware can only handle 32 VMFS snapshots? Do you mean 32 individual vm snapshots per datastore, or?

Right - in order to use RMV you'll have to delete the VMware snapshots from the vSphere client. Once you do that, RMV will take a 3PAR virtual copy of either a virtual machine or a datastore. If you are trying to snapshot an entire datastore, I believe all VMs in that datastore cannot have a VMware snapshot.


After the VMware snapshots are removed, and you use RMV to create a 3PAR virtual copy, it will queisce the VM and take an application consistent snapshot. Note that VMware tools has to be installed on the guest OS. The process RMV will use is to take a VMware snapshot on it's own, then perform a 3PAR virtual copy, and then delete the VMware snapshot automatically. This is what gives you the consistent state of the virtual machine on the 3PAR side.


I'm talking about the VMware snapshot through the vSphere client, normally per virtual machine. The 3PAR can handle hundreds (if not thousands) of virtual copies, so there is extra scalability there. For more on the VMware snapshots see here:
http://kb.vmware.com/selfservice/micros ... Id=1025279

Great, thank you for the clarification.

In our environment we test various client OS's and when testing is complete we revert back to a clean snapshot. So it sounds like I will be unable to leverage RMV in this environment... unless I change how the vms are returned to a clean state, and somehow incorporate the 3par into the equation.

You should be able to use it in a similar fashion, it's just to take the first 3PAR VC in your environment the client OS would need to go back to the first clean version. So if you revert the VMware snap (back to default), delete the VMware snaps, and then use RMV to take a 3PAR VC, I'm confident you'll have the same effect of what you want. Then after testing or whenever, you can use RMV to restore the VM back to defaults.

Once you have clean RMV created/managed snapshots, can Netbackup come into to play to back these up? If so, where would be a good starting point to read up on that? I looked at the off array snapshot backups with Netbackup, but that led me down a path of using Windows RDMs with the 3PAR VSS provider on each guest... not what I want/need in a VMFS datastore environment.

Thanks!

_________________
Richard Siemers
The views and opinions expressed are my own and do not necessarily reflect those of my employer.

We used to use Commvault to backup our VMs and it would snap the vm on the datastore and read it off the media agent using vmware VADP. There is not real reason to use array snaps with a third party backup that support VADP. We went away form that model because backing up 800 VMs every day put a huge IO load on the array. We now backup on the dat that is needed form the handful of VMs that need long term retention. All the other systems I have a script that enumerates all the VMs on a datastore, snaps the VM, then snaps the 3par lun with a 14 day expiration. We have proven that while the recovery in this model is more manual it is far more speedy for recovering a VM as we found Commvault could take 1-3 hours to recover a VM from a VADP deduped backup, the manual recover VM from array snap takes less than 30 minutes.

We also replicate our production array to an array in DR so every evening before the replication jobs run I snap all the luns on the DR array with 14 day retention, essentially provided instant DR recovery of up to 14 days of VMs.

Gentlemen,

I've used a lot of backup software over the years. Veeam Backup and Replication has been not only the best virtualization backup software I've ever used... its the BEST BACKUP SOFTWARE I'VE EVER USED. Especially with native support for Fiber Channel StoreServs and iSCSI support for StoreVirtuals (iSCSI support for StoreServs on the way) that uses storage snapshots rather than VM snapshots for backup... its just too fast. On top of that, you only have to buy licensing for hosts your backing up from. I was able to standup my Disaster Recovery site with replicated VMs to standalone hypervisors FOR FREE. You can't beat the licensing model. On top of that, their Active Directory support coming.

Edit: Oh did I mention it's UNBELIEVABLY FAST with direct SAN access to my 3PAR???? Plus a virtualized VeeamProxy that accelerates and compresses and dedupes? It reads 9.1TB off my virtual infrastructure, Dedupes it by 2.8X, Compresses it by 1.4x, and transfers only 2.4 TB to my repository, and then copies it onto tape. You can even crank it up, I just found that given my available processing power (the more cores you have the more synchronous tasks it can do at once... direct relation... its so great) this gave me good speed since space isn't that premium in my backup repository. We only have a 2 month retention for this data.