[SOLVED] Proxmox SDN fabric, access shared storage in fabric

[Knogle]

Hey folks—hope you’re doing well.


I’m running a 3-node Proxmox cluster in a hyper-converged setup.
On node 2 there’s a TrueNAS VM that exports two NVMe pools over NFSv4.


Previous setup (worked fine):

• All 3 nodes connected to a 10 Gbps switch via LACP, classic VLANs.
• VLAN 3100 was the default VM network.
• The TrueNAS VM NIC was bridged into VLAN 3100.
• Each Proxmox host also had an IP interface in VLAN 3100 to mount the NFS share.

What changed:

• I removed the ~300 W Nexus 3000 switch and cabled the nodes in a ring.
• I now run a VXLAN fabric, and VLAN 3100 exists inside that overlay (VXLAN + VLAN tag, actually VM bridged to the fabric interface + VLAN tag).
• VM-to-VM networking and live migration work.

My problem

I can’t figure out the clean way for the Proxmox hosts themselves (not the VMs) to reach the NFS server that now resides inside the VXLAN/VLAN-3100 segment.
In other words: what’s the appropriate method to give the hosts IP reachability into VLAN 3100 inside the VXLAN overlay so they can mount NFS from the TrueNAS VM?


My question

• Is it supported/reasonable to put an IP on the overlay bridge (e.g., assign an address to the VXLAN bridge) and/or create a VLAN sub-interface on that bridge (e.g., vmbrX.3100) on each host?
• Alternatively, should I attach a veth from the host into the VXLAN/VLAN-aware bridge to “place” the host stack inside that segment?
• Is there a recommended Proxmox SDN way to give the host an interface inside a VNet/VNI for this purpose?

Thanks in advance for any help!!!

 

[Knogle]

I was able to solve the problem!
The solution is quite simple, but would be nice to have something in the GUI.
The solutiion is to simply add a VLAN interface to the "VXLAN" bridge interface, and it works