CDP question - mix with traditional B&R

[FrenchBlue]

Hello,

There's something I don't get with CDP.
We have 2 distant Vsphere clusters, let's call them A and B, with the need for bi-directional DRP, each with its own Veeam VBR, repositories, proxies etc...
We need 2 levels of DRP: low RPO and there we want to use CDP now (was done by another well known tool up to now), and "normal" RPO for which backups replications are fine.
Veeam VBR from site A does local backups and replicates them to a dedicated repository in site B. If we lose site A, we just import this replications repository in Veeam VBR "B" and can restore the replicated backups locally.
But for CDP, we need to define them for site A from site B, if we lose site A, the failover has obviously to be triggered from Veeam VBR "B". Unfortunately it doesn't work. When testing a CDP replication, it fails right at the beginning , complaining that the source cluster is already managed by Veaam VBR "A":

"5/4/2022 4:49:45 PM :: VM configuration for the initial sync completed with errors Error: The specified cluster domain-c8 is managed by a different backup server par01veeamvbr.xxxxxx: first occurrence at 5/4/2022 4:46:38 PM, last occurrence at 5/4/2022 4:50:23 PM, 7 retries"

What are we doing wrong here?

Thanks.

[FrenchBlue]

OK, I found the issue after a few hours of trial&error, might be useful if someone has the same problem. I/O filters are not just code pushed to the clusters/hosts. They are tied to the VBR instance which pushed them there. We had initially deployed them with Veeam "A" , so it simply doesn't work if you try to define cdp policies on these clusters with Veeam "B". I had to remove I/O filters from Veeam "A", then re-install them from Veeam "B", and then re-create the cdp policy. After this, everything was fine.

It also means that for bi-directional DRP with cdp, it is mandatory to have separate production and DRP clusters.

[veremin]

Correct, I/O filters on a cluster should be managed by one backup server only. This is mentioned along other requirements here; might be worth reviewing. As to disaster recovery scenario, you will need to add target cluster to the second backup server, after initial rescan the backup server should expose the existing CDP replicas and allow you to perform failover.

We re-test this case internally (just to be 100% sure), and I will post back the results of our findings.

Thanks!