Recommendation for Presenting FC SAN Disks to Backup Repository

[morit]

I have some questions about a new Veeam Backup setup (V10.0)

There are 3 x ESXi hosts in this setup and all datastore volumes are created from shared FC SAN Disks (HPE 3PAR).
1 x Windows VM will be deployed for Veeam Backup Server & Proxy.
There is another MS Windows Server 2019 VM as the backup repository.
The backups are to be stored on another FC SAN Storage (HPE MSA 2060 storage - 150 TB).

Q1. What would be the recommendation to achieve this design and present the MSA storage to the Windows VM as a repository?
Should we use RDM disks for MSA LUNs? Understand from below article that we should avoid placing backups on VMFS volumes.

https://www.veeam.com/blog/vmware-backu ... tices.html
vmware-vsphere-f24/vmware-rdm-instead-o ... 47024.html

Q2. If we were to carve out multiple LUNS from the MSA storage and present it to the Windows VM as RDM, would there be any concerns or additional configuration required by Veeam?

Q3. Should we present all datastore LUNs to both Veeam Proxy VM and backup repository VM so that we can use Virtual Appliance transport mode?
https://helpcenter.veeam.com/docs/backu ... ml?ver=100

[morit]

Veeam Support Case # is 04620254

[HannesK]

Hello,
and welcome to the forums.

1 x Windows VM will be deployed for Veeam Backup Server & Proxy.

There is another MS Windows Server 2019 VM as the backup repository.

I recommend to make it one physical server with FC connection. Then you can use the 3par properly with backup from storage snapshots. And you don't have any chicken-egg issues. And you don't need to think about ugly workarounds. And you just have a happy life

Q1: both is bad from my point of view. RDM is better than VMFS
Q2: yes, there is a recommendation around 200 TB for REFS volumes (while other customers go beyond 400 TB). But that's probably irrelevant for that environment.
Q3: No. A physical machine with backup from storage snapshot would be the recommendation. But yes, you can do it.

I recommend https://bp.veeam.com/vbr

Best regards,
Hannes

PS: I told support that they can close the case. It's a design question (not a technical problem)

[morit]

Thank you for the advice.

I agree with the idea which one physical server should be deployed. We’ve highlighted this option to the customer.
But, they’ve deicide to proceed with their original design (which is one VM for VBR & Proxy, and another VM for repository).
They have following questions, and may need your advice.

Q1) They are planning to use Virtual Appliance Mode.
Since Proxy and Repository are sitting on different VMs, I assume that the backup data flow still go via their backup VLAN/network on step5 below (from Proxy VM to Repository VM).
Is it correct understanding?

The process of data retrieval in the Virtual appliance transport mode includes the following steps:
1. The backup server sends a request to the ESXi host to locate the necessary VM on the datastore.
2. The ESXi host locates the VM.
3. Veeam Backup & Replication triggers VMware vSphere to create a VM snapshot.
4. VM disks are attached (hot-added) to the backup proxy.
5. Veeam Backup & Replication reads data directly from disks attached to the backup proxy.
6. When the VM processing is complete, VM disks are detached from the backup proxy and the VM snapshot is deleted.
https://helpcenter.veeam.com/docs/backu ... ml?ver=100

Q2) There are some Windows VMs with MS SQL cluster and MS Active Directory (which requires Veeam Agent for Windows).
Should they add a backup network adapter for these MSSQL and AD VMs? Do they need to set the backup vlan as the preferred network?

Q3) For the rest of standard VM backups, each VM does not require any backup vlan connection. Is it correct?
Currently the Backup server and Vcenter have 1 x IP each and are sitting in the management vlan.

[HannesK]

Q1) probably yes, that depends on the setup. I would go with one network card. That means simple routing and nothing to worry.

Q2) there is no requirement for agents. Per default, Veeam is agentless. The only exception are "half-physical" servers where VM-snapshots are impossible (e.g. SQL servers with shared disks on FC / iSCSI / RDM LUNs). I recommend to avoid complex setups with multiple network interfaces. Even with multiple network interfaces, the Windows routing table should do the job.

Q3) correct. VMs have no access to the backup LAN. I'm surprised that the ESXi host management ports are not in the management VLAN. Anyway, everything is possible, as long as you allow the required connections: https://helpcenter.veeam.com/docs/backu ... ml?ver=100

[morit]

Thank you for the inputs. Noted.