Twin-datacenter backup with StoreOnce

[mkaaden]

I have a twin-datacenter setup with a VMware Metro Storage Cluster on 2 3PAR 8200's with Peer-persistence
I also have 2 StoreOnce 5100 Appliances (136TB) and 2 physical back-up proxies in each datacenter.
ISL between datacenters is 4x 10GbE (4 switches, full mesh)

Back-up retention policy:
- Short term: 14 restore points, daily incremental, weekly synthetic full
- Long term: 2 restore points, 2 weeks, 12 months, 4 years

I am looking for the optimum back-up setup to use and was thinking of the following:

Create 2 Catalyst Stores per Appliance: 1 for short term and 1 for long term.
Create 6 Repositories in Veeam, 3 on each gateway: to the short and long term store on the local appliance and to the long term store on the remote appliance.
Both proxies wil process back-ups of datastores local to the 3PAR on their datacenter (so active-active)
I will be using a few large datastores (10TB+) and create jobs per datastore using storage snapshots.
To have both StoreOnce Appliances store a complete copy of all the back-up data, I want to setup 2 copy jobs per site:
1 job to copy short term to long term on the local appliance.
1 job to copy long term to the remote appliance

Is this a feasible setup ?

[emachabert]

Hi,

Your setup is totaly feasible but not optimal as a backup copy job from a catalyst store to another catalyst store on the same array won't be very efficient in term of throughput. You will have to reread (rehydrate everything) to send to the target store which is on the same array.

The best way to do what you want is to have a repository on each Proxy (DAS) to hold the short term on each DC , and then use the StoreOnce for mid and long term retention (that's the pruprose of such arrays). You also should use backup from storage snapshots for any jobs of any kind.

I do have 3par Peer persistence over 2 DC (production) and a third DC (DR Site) with a StoreOnce 5100. On each DC I have a physical proxy with lot of cores and a DAS (DL380 25Tib RAID 50 4GB cache, NTFS with windows dedupe) to hold short term retention. Then I have a backup copy jobs to build mid/long term retention on the third DC (I also have replication job on the third DC using the storeonce stores to rebuild all the production services every morning on a third 3Par, but this is a different story). With this setup I do have really good performance and stability over time (~900MB/s incremental and over 2,7GB/s for weekly full backups).

Another thing to note, pointing two repositories to the same storeonce endpoint is not supported as per Veeam documentation. What I have observed is that while this limitation appears to be true for CIFS (periodic repo rescan will screw up the configuration database) it appears that Catalyst store aren't impacted (I repeat: per my observation).

[mkaaden]

Yeah, I know copying from store to store on the same appliance is far from ideal.
But I think this is the only way to have all backup data with full retention in both DC's
I will use catalyst stores, so that would ease the 'pain' a little.

It would be easier if we could build the retention on the same store using the catalyst features and then have the appliance copy the store to a remote appliance. (hint @veeam)

I also would have liked a third appliance for Long Term copy.
Then it would be 2 smaller StoreOnces and 1 fairly big one.

[emachabert]

I don't know how much data you will move but depending of the volume you may face serious performance problem that will result in very long times because read speed from the source store wiil be limited. You should really look into a first level repository based on DAS storage as per Veeam reference design guide.

I agree that cross site copy would be better if handled by a catalyst copy job but Veeam does not yet manage such jobs on its own.
Have you looked at HPE StoreOnce Catalyst Copy Utility ? It is a standlaone tool to manage catalyst copy jobs. You could use it to trigger the copy of the long term store it would be very efficient for the long term replication.

[mkaaden]

So, an ideal solution with StoreOnce would be:
- Have a piece of DAS storage in the proxy/gateway as landing zone for the initial daily full-incr chain
- Use the StoreOnce as secondairy copy and long term storage

I will look in to the HPE StoreOnce Catalyst Copy Utility.

Thanks !