Veem B&R 8 seems to take much more time processing replicas

[unsichtbarre]

We recently upgraded to Veeam B&R v8 and have noticed that it is slowing down our replication jobs significantly. We have gone from surpassing a 30min RPO to requiring greater than 2 hours/job

The slowdown happens during the step(s) "Using target proxy Blah Blah Blah for disk Hard disk X [nbd]" and holds the job for 4-12 minutes per VM. We have parallel processing enabled and enough proxies/CPUs to run multiple disks at a time - and even when only one VM is processing, there is a tremendous delay we did not see before.

If you add 4-12 minutes per VM on a job with 10 VMs/replicas, we are no longer to meet a RPO less than 2 hours, whereas in Veeam 7, we were easily able to meet sub-30 min. RPO

After the data starts to replicate, replication it seems to move as fast as it ever did.

Any ideas?

This case has been sent to Veeam Support and log files have been uploaded.
Case #00733966

Thanks,

[Vitaliy S.]

Have you tried using hotadd backup mode at the target location to see if it makes any difference or not?

[unsichtbarre]

We have been burned by hotadd in the past. If possible, we would like to avoid it.

Also, v7 was not configured for hotadd, prior to the upgrade.

THX for the response.

[Vitaliy S.]

Ok, let's see what our support team figures out after reviewing the debug logs. BTW, does this happen on all target proxy servers?

[unsichtbarre]

At this point, there are two proxies at source side w/4 vCPU each and only the Veeam server at target side w/8vCPU. I can add an additional proxy at the target and test, although the problem occurrs even when only one VM with one disk is running.

The support team has responded and said that "it looks like encryption in enabled." We have double-checked and encryption is not enabled between host/proxy. Since it is a replication, we dont know of any other encryption feature.

[veremin]

Most likely, they're talking about traffic encryption that can be found in the global throttling settings. Thanks.

[unsichtbarre]

We are still struggling with a halt of about 6 minutes per VM, per job.

Encryption was never enabled anywhere in the GUI, but on the advice of Veeam support we have applied the following Registry key:

DisablePublicIPTrafficEncryption

Type: REG_DWORD
Default value: 0
Description: By default, traffic between different subnets and on public addresses is encrypted. To globally disable network encryption between agents, set to 1.

In the following image, you can clearly see where Veeam just stops for over 6 minutes. In Veeam 7, this job was finalizing both VMs in 3-5 minutes. In our other job with 10 servers, we are adding 60 or more minutes, just waiting around

New logs have been gathered and uploaded to the same case referenced before.

Thanks,

[unsichtbarre]

Thanks to Ken Sauer at Veeam Support, this is now resolved.

Network Connections > [ALT] > Advanced > Advanced Settings > Adapters and Bindings

Place the Replication network first in the list and the local vSphere Management Network second

Thanks Ken!