Virtual Machine with multiple NICS serving as a proxy

[Stiddy1979]

I am trying to track down a performance issue with replication. I hope this is the correct forum "VMWare vSphere" for this question. Here is a general overview of my environment. We are running ESXi6.7u3m and Veeam Build 10.0.1.4854. My primary and DR datacenters have Cisco Hyperflex (3) nodes each all connected via 10gb to core switches. I have a dedicated backup proxy assigned to each node in my environment having 6 vCPUs and 8GB of memory. Each of these Backup Proxies have two vmnet3 NICs assigned, one for the primary data network and one with direct access to the storage vlan. While watching a job run for replication I RDP'd into the backup proxy at HQ and also at the DR and pull up Performance monitor to watch the NIC utilization. It seems that my NIC with direct access to storage is never used and I would suspect that with Cisco HyperFlex taking a storage snapshot the traffic would travel over the NIC with direct access to storage. Am i looking into this the wrong way or do i possibly have a misconfiguration?

My source proxy mounts the virtual machines disk that is being replicated as NBD, my target proxy mounts as a HOTADD.

[Andreas Neufert]

Hello,

please allow me to share the dataflow.

NBD
Traffic between ESXi host and Proxy.
We will get from the vcenter APIs (or from the host if no vcenter is there) the full qualified domain name of the VMKernel interface. We contact the host over the IP we receive from DNS. You can manipulate the used VMKernel interface over DNS or the local hosts entry of the Proxy and B&R server.


Hotadd
Hotadd mounts the disk within the ESXi host to interact with it. There is no traffic for this on the network. Of cause the ESXi host speaks with the storage then.

Veeam traffic
For traffic that flows between the Veeam Server, you can influence by preferred network setting.https://helpcenter.veeam.com/docs/backu ... ml?ver=110

[Stiddy1979]

Thank you for the reply and explanation. I guess what really has me questioning why the source proxy continues to utilize NBD over DSA (Direct Storage Access). If i look at the transport mode of each source proxy I have them setup for Transport Mode DSA with the option to "Failover to network mode...." enabled. It seems the proxies never use DSA. How can i trouble shoot why DSA is never utilized?