Mixing Backup & Replication - or Split into multiple VBR Servers

[[OIT]Francis]

Hi fellow Veeam enthousiasts and engineers,

For a Customer we are rebuilding their DR strategy by adding a passive datacenter with spare compute/storage capacity after a consolidation. Suffice to say, we have enough capacity to run all workloads in the 2nd datacenter, so we want to leverage the existing Veeam licenses & infrastructure and replicate the VM's that are needed in a DR/BCP scenario (~50-100).

Currently we are on the fence on how to proceed, as all backups are being managed from a physical VBR server (12.3) in the primary datacenter, multiple local proxies and 3x HPE Appollo servers configured as LHR's. We also have a physical server + 3x Appollo's available in the 2nd datacenter to setup a VBR infra.

Option 1 - Migrate the current backup server (export/import) to the 2nd datacenter and use it for both Backups, Copy-Backups & Replication
- PRO: Leverage existing proxies for all requirements
- CONTRA: Performance ? Job conflicts due to proxy limitations & # of jobs/vm's.

Option 2 - Run everything from the primary datacenter, including the replica/failover jobs, and create an export/import procedure to use in case of DR
- PRO: Least amount of work involved, leverage existing proxies and avoid having to reconfigure agent-based jobs etc ...
- CONTRA: Performance (see option1) and reliance on DR import procedure to revive the VBR infrastructure

Option 3 - Keep the existing infra in the primary DC for Backups & Copy-Backups (to 2nd DC), Setup a separate VBR server in the 2nd DC for replication/failover jobs.
- PRO: No potential proxy conflicts, role-based management, replica infrastructure readily available in case of DR
- CONTRA: Additional proxies required for replication, management overhead, job/snapshot conflicts (?)

Additional question ... Let's say a VM is being replicated with a Veeam Failover Policy. Do we backup the primary/running VM, or backup the replicated entity in a powered-off state (for VM's not requiring application integrations) ?

If there are any architectural guidelines that explain this and/or provide best-practices, it would be much appreciated to link them.

Thanks !

[Brad.Barker]

Option 1 and option 2 should be where you are focusing your considerations.

Option 3 would go against best practices and design from the perspective of processing the same workloads with two different Veeam servers. The main concern is these systems won't know what the other is doing, and there are possible scenarios in which, even if not scheduled for it, jobs might overlap with one another from the different Veeam servers which can lead to unexpected outcomes and issues.

I would lean towards option 1. This specifically revolves around the concept of recovery in the event of a large scale outages. If this customer in question were to lose all of the primary site, you will be in a scenario where you either are manually powering on replicas (which isn't the best idea because it complicates failing back to the production after repairing it) or rebuilding the Veeam server in DR and importing the configuration. As part of your design I would recommend sending the configuration backup of your Veeam server that you're creating at the DR site to the primary site. This way if you lose your DR site, you can rebuild your Veeam server in primary to get backups running again while repairing or rebuilding DR for the copy jobs and replication jobs. Finally, Veeam has options available to minimize the performance concerns from option 1 depending on the frequency of replication being targeted. If you intend to only replicate the servers once a day, you can actually source the replica data from a backup or backup copy (See Data Sources in the v12 UG here https://helpcenter.veeam.com/archive/ba ... ce_vm.html). So if the backup copy for these systems is already placed at the DR location, you could reduce the amount of data sent between the links. If the intention is to run the replication more often, or to leverage Continuous Data Protection, the amount of data sent over should be lower and at that point link speed would more than likely be the main concern. This is because Veeam can always leverage virtual proxies that you can spin up to meet demands, and in some cases replication may benefit from virtual proxy writes at the DR site because of the ability to utilize HotAdd operations for the writes.