Gateway Server Required

[ratkinsonuk]

I'm migrating my Veeam B&R environment to use S3 for long-term storage. I've configured a SOBR repository so that the Capacity tier has the 'Copy backups to object storage as soon as they are created' box checked.

My question is, should I be building a separate gateway server to handle to CPU/IO load caused by the SOBR Offload process, or does this get handled by the proxy server?

What I'm trying to do is reduce the workload on the Veeam application server, as it's already pretty stretched.

Cheers, Rob.

[chris.childerhose]

The actual repository server does SOBR offload, and there are some things you can fine-tune as per this blog post - https://www.veeam.com/blog/sobr-archite ... CPU%20core.

[ratkinsonuk]

Thanks for the reply Chris. I've had a look at that article, particularly "Veeam Backup & Replication uses a “SOBR Offload” background task to copy and/or move the backup data to the Capacity Tier."

The article doesn't definitively say where this task runs, but reading between the lines, I would say it runs on the main application server. Looking at the offload that's currently running on my server, I can clearly see the 'VeeamAgent' processes and the network load the task is generating.

I'd like to shift that load elsewhere. If not, then I'm guessing I need to seriously boost the performance of my B&R server?

Cheers, Rob.

[ratkinsonuk]

Also, in reference to your comment "The actual repository server does SOBR offload" - our source repositories are stored on a Dell Data Domain, so there's no 'server' involved.

Not sure what this looks like if you use something like a Linux repository instead?

[chris.childerhose]

Then it would make sense that your VBR server is doing the offloading since the DD is not an actual server which I assumed (my bad). So yeah, you need to beef up the VBR server if that is the Gateway server specified for the DD or switch to a different repo type like Linux XFS or Windows ReFS. The article I posted references a repo server versus an appliance like DD.

[ratkinsonuk]

The settings for a Data Domain (DDBoost) device do allow for a gateway server, which I'm guessing is where the Data Mover would sit. I'm fairly confident that building a VM and assigning it as the gateway should see the bulk of the SOBR Offload workload move as well.

I'll let you know!

As it happens, I'm moving the a Hardened Linux Repository setup with block based storage, so your original comments regarding the repository server should still be valid.

Cheers, Rob.

[RubinCompServ]

Then it would make sense that your VBR server is doing the offloading since the DD is not an actual server which I assumed (my bad). So yeah, you need to beef up the VBR server if that is the Gateway server specified for the DD or switch to a different repo type like Linux XFS or Windows ReFS. The article I posted references a repo server versus an appliance like DD.

Is there a sizing guide to appropriately size the VBR for the S3 offloading tasks?

[chris.childerhose]

You can check this blog article which explains and gives some tips - https://www.veeam.com/blog/sobr-architecture-guide.html

This page also helps as it is the Best Practice site for Veeam - https://bp.veeam.com/vbr/2_Design_Struc ... teway.html

[ratkinsonuk]

Page 11 of this document covers the Repository Server David, but it's a bit ambiguous regarding CPU. I've gone for 8-core 32GB as a starter, and I'll see which way I need to move once the data is travelling through it.

https://www.veeam.com/veeam_backup_11_a ... tes_rn.pdf

Cheers, Rob.