W2K12 Proxies/Repositorys using CSVFS (Clustered Shared Vol)

[dkvello]

I'm curious about re-configuring my Veeam B&R Proxies (They are FC SAN attached and use their own LUNS hosted on a Nexsan e48) to use w2012 and Clustered Shared Volumes.
Each Proxy is also a Repository server.

It's end-to-end Fiber Channel (Reads from Storwize FC, writes to Nexsan FC)

Problems I'm hoping to solve:

1) Each proxy has its own set of NTFS Volumes that reside on the same Nexsan FC storage. This does not allow for sharing free-space between clusters. i.e.
If on LUN on one proxy is close to empty, I cant use free space dedicated to another proxy

2) Proxy's won't be able to load-balance since

I could create a NAS gateway or a separate Repository server that owns all FC Storage and have all Proxies push their data through it, but that would defeat the load-balancing aspect.

or

Can I upgrade my Proxies to W2012 and create a CSVFS from my Nexsan FC Storage, haveing all proxies/repository nodes seeing the same clustered filesystem ?

http://blogs.msdn.com/b/clustering/arch ... 91490.aspx

Anyone tested this ?

[Vitaliy S.]

Hi Dag,

If you want to have a redundancy for all backup components (proxy and repository servers), then I believe you need to create an SMB 3 share out of the CSV and then target your backup jobs to this share. In this case your backup proxy servers will be retrieving data directly from the production storage, but will be sending data over the network.

If you want to send data in the "lan-free" mode then, creating a CSV on both proxies will you give you an ability to share free-space between two repositories. The drawback of this approach (as you've correctly noted) would be the loss of redundancy for the backup repositories, as if you lose one proxy server (backup repository), then this repository will not be accessible by the job as this repository will be explicitly named (by the host name) in the backup job wizard.

Me personally, I haven't tried these scenarios, but if you can have a dedicated network for your backup traffic and need a redundancy for the repositories, then the first approach looks a bit more reasonable to me, though it also has its own drawbacks (for ex. CIFS repository was never considered as best target for the backup jobs).

Hope this helps!