Replication questions

[theviking84]

I've got backups working across various datacenters and use various proxies per datacenter to achieve this.

Now I am looking at replicating vms from remote sites back to an HQ datacenter so that if we have long vmware esxi host outage at a remote site we can power up and run the replicas at the main site.

My thoughts are

Do I need to spin up dedicated additional proxy at each remote site to use for "only replication" purposes? If I don't, then won't Veeam server be dealing with competition for backups/replication across same proxies? The backups only run usually once per night, but the replications would be something that we probably do every hour or less. What will happen during the nightly backup time when replications need to happen at the same time?

Are there any big gotchas or things I need to think of with this plan? We will have to re-ip at the HQ datacenter and I see that Veeam is able to do this automatically, and then maybe if we need to fail back to the original we can do that still?

How are backups usually handled in the event you end up failing over to replica and then eventually back to production? Like for example, when the backup job tries to run, now that we are failed to replica the backup will fail I'm guessing, or will it detect the new replica as the same vm (same venter, moref id changed or no? replica cause change?) So will incrementals just resume as normal? Then when I fail back and the backups go back to how they were off the original vms, does that resume incremental fine?

We use sobr and scale out to s3, copy mode and move mode, so this could get messy if we have to make veeam take full backups of all these same servers again in the cloud just because we temporarily failed over to replicas. We also use immutability so its like our storage use would skyrocket and not be able to delete if we went into one of these events...

Maybe disable backup job until we fail back to production vms?

[PetrM]

Hi David,

I would recommend to deploy a proxy on each site in any case to provide better read performance for both backup and replication scenarios, for example you can spin up a virtual machine and read data in HotAdd mode. One more idea is to run remote replica from backup as you reduce load on production environment, already existing backups will be used as a source of data.

You still have the option to failback to an original VM, Re-IP is applied to replica on the target and should be used if you have different addressing schemes on source and target sites.

Backup and replication are completely independent from each other, there is no special logic which "knows" that a VM you're backing up is failed over to its replica and vice versa. Thus, backup jobs will try to continue its runs according to schedule.

Thanks!

[foggy]

Hi David,

Do I need to spin up dedicated additional proxy at each remote site to use for "only replication" purposes? If I don't, then won't Veeam server be dealing with competition for backups/replication across same proxies?

You need to have a proxy on the target side for optimal data transfer and target datastores population but that doesn't need to be a proxy dedicated to replica - it can be shared with backup and other tasks just fine, provided it is capable of taking all the load without resource contention.

What will happen during the nightly backup time when replications need to happen at the same time?

Backup jobs do have a higher priority and replication jobs targeted to the same VMs will wait (if started when the backup is underway) or be terminated (if already running when the backup starts).

How are backups usually handled in the event you end up failing over to replica and then eventually back to production?

Replica VMs are different VMs and you need to protect them separately, by creating a dedicated backup job, for example.

[theviking84]

Thank you both!

One more clarification. So assume I decide to share the proxies I already have (I have some at every site already for backup, hot-add)

In this case, are backup/replication still considered just "tasks" and so if the proxy in question can handle 4 tasks then it can be doing a backup task and also a replication task etc?

[PetrM]

Hi David,

One proxy task literally means one VM disk. If you have 4 task slots on proxy: 4 disks in total can be processed by all jobs. You may refer to this page to get more information regarding resource scheduling and limitation of concurrent tasks.

Thanks!