new deployment, recovery at DR site question

[prolix21]

We're in the process of deploying Veeam Backup & Replication in our datacenters, the goal being to copy all backup jobs from our primary data center to our secondary.

The current thoughts on architecture are the VBR server is at our primary site, repositories, along with proxy servers handling our local jobs and a WAN accelerator to help shuttle stuff to the secondary site. At the secondary site all we have is a WAN accelerator, repositories, and a couple proxy boxes for local jobs there. Pretty much straight out of the Veeam docs.

Our storage is all 10GB iSCSI and in testing is currently mounted to our VBR at the primary site and to the WAN accelerator at the secondary site. The WAN link is 200MB and I can bump that if we run into issues.

Initially we were thinking we wanted a VBR on both sites, but in all the v8 docs that seems to be mentioned nowhere. If you look at the v70 docs it does mention this being a recommendation, but that recommendation seems to disappear in v8.

So question 1 is if our primary site disappears, and we don't have a VBR at the secondary, what do you do to recover? I could replicate our VBR there and that might work? I've searched around here and everyone seems to think two VBR boxes is unnecessary, so I'm wondering what I am missing if you want a truly HA config and both sites to be able to run independently should one go down.

Question 2 is more me asking for critiques and if there are suggestions out there for this setup I'd love to hear them. End goal is we'll have 300+ VMs being processed locally and a good chunk of those jobs will get shipped to the secondary site. I know it is going to take forever to get our initial copies sent to the DR site.

[alanbolte]

The topic has been addressed in detail elsewhere so I'm going answer this sort of obliquely: when it comes to running everything from backup side VS from DR side VS having two separate backup servers, very little changed about the pros and cons of each option between v7 and v8, so most of what you've read from v7 is still valid. The biggest change in v8 was the introduction of replication from backup, which doesn't support using two separate backup servers. While that option has many benefits, it pits the cons of running local backup from the DR site against the cons of running replication from the production site, which led to a complex web of less than ideal choices. In v9, the Guest Interaction Proxy and Mount server eliminate the downsides of running backups from the DR site, so I expect that to become best practice (deciding that isn't my job though).

All that said, if you're only doing backup copy and not replication, the matter becomes rather simpler. If your DR site just contains Veeam infrastructure, what are you planning to restore to?

[prolix21]

All that said, if you're only doing backup copy and not replication, the matter becomes rather simpler. If your DR site just contains Veeam infrastructure, what are you planning to restore to?

We have a full vSphere environment at the DR site, that will be the restore target should something occur. We aren't using Veeam for replication, we're using something else for that currently. Out of Veeam all we want is a copy of the backups at a secondary site for worst-case scenario, and for compliance reasons.

[veremin]

If VB&R server is replicated via different solution, the idea would be to failover it during DR, point it to existing repositories (backups) and start restoration procedure.

Otherwise, you can just backup its configuration, and in case of DR deploy a new backup server, restore its configuration and proceed with other VMs restoration.

Thanks.

[foggy]

Also, I don't think you really need WAN acceleration for 200MB link. It was designed for slower connections.

[rreed]

Subscribe; needing to do the exact same thing. Our two data centers are to mirror each other, but our B&R server is in only one of them. What happens when that data center goes down? Our backup data is replicated to the other data center but no B&R server to manage recovery. We don't feel restoring a B&R server THEN restoring our VM's is the best solution for us - additional time to recovery - so to keep that time as short as possible is replicating the B&R server preferable? Other options? I would be open to building a second B&R server at the other site and manage each site as truly independent but not where we would stand on additional licensing costs if that applies since our backup footprint doesn't change - only two consoles vs. one.

[foggy]

Any mentioned approach is viable. Some people prefer to go to vSphere Client to fire up Veeam B&R replica manually and then start failover, others - to make several clicks through Veeam B&R setup, import configuration, and failover. Having second (or the only) Veeam B&R instance in DR location to be able to immediately start failover is also an option.

[rreed]

Awesome, thanks foggy. Are there additional licensing requirements if we were to split our B&R server into two, each managed seperately? The CPU count wouldn't change, it would just be split b/t the two rather than lumped onto one.

[foggy]

No issues with that as far as the number of sockets across both backup servers is inline with the licensed amount. You can set up Enterprise Manager for easier license management.