Backing up a Metro Storage Cluster

[gpomanti]

I have a couple of questions for anyone who has implemented a vSphere metro storage cluster with peer persistence.

First some background on the layout. On our site, we have two data centers separated by about 500 yards. In data center 1, we have 16 ESXi hosts running on a 3PAR 7400 array and an HP EVA8400 for a primary backup target. In data center 2 we have an additional 16 ESXi hosts running on a 3PAR 8400 array and an HP EVAP6300 for a primary backup target. There is a physical Windows 2012 R2 backup proxy in each data center. Every thing is connected via Fiber Channel.

We are planning to configure a stretched cluster where every datastore on the 7400 is fully synchronous to the 8400 and vice versa. The vSphere clusters will be split between the two data centers with DRS rules set so VM's should live on hosts in the same data center as their primary storage. In the event of any sort of failure, VM's will be able to run in the other data center (single array failure, host failure, entire data center failure, etc.). For backups, everything will be kept local to the data center. For example, VM's in data center 1 will be backed up through Direct SAN access to the backup target in data center 1. So backup layout is: Data Center 1 - 3PAR 7400 -> proxy 1 -> EVA8400. Data Center 2 - 3PAR 8400 -> proxy 2 -> EVAP6300

My question is, in the event of some kind of failover scenario, how can I ensure my VM's are still backed up? For example, the 3PAR 7400 in data center 1 fails. The VM's should remain running but their storage will now be in data center 2 on the 3PAR 8400. Will Veeam have any issues with this or should everything continue to be backed up?

[dellock6]

Hi,
I never worked on such a scenario, but I got some details from a partner doing many of these deployments, here it is what he wrote me a year ago:

Since 3Par PP is using native T10 ALUA specification, everything works out of the box.
In the Veeam Backup Server, you just need to connect the two 3Par arrays in the SAN management pane.

In a normal situation, each LUN has 8 path, 4 active I/O and 4 standby.
On both arrays, each replicated LUN has the same WWN and is exported with the same lun number.
Path Policy is set to RoundRobin on each host (VMware or Windows).
When backup starts, Veeam Backup finds the active array by looking which target is holding the active path. At least, it is what we understood when analyzing the behavior.

After a failover, active paths are switched to the secondary array. Veeam connects to the array looking at the target holding the active path, here again this is our understanding of the process and in our case it all went well.

We didn't try to do a switchover during backups. If using storage snapshot it would certainly fail since the storage snapshot only exists in one array. If using standard direct SAN without storage snapshot integration, it would be transparent and follow the path switchover.

Hopefully, there is now the option to fallback to normal mode if storage snapshot isn't working.

Last comment was referred to v8 with the option to allow failover from storage snapshots to "simple" directSAN.

hope this helps (and thanks to our vanguard Eric M. for the email I copied here)

[gpomanti]

Great, thanks for the info! That all makes sense and should be pretty easy to test, I just wanted to make sure I wasn't missing anything. Once we have the array configured, I will test backups and fail over scenarios to see if we can get storage snapshots to work.