Network configuration for NFS shared datastore

serg2002

New Member
Jan 8, 2026
2
0
1
Hi
What would be the best Proxmox nodes network configuration to connect NFS VM datastore as a shared storage?
As far as I understand Proxmox recommends to use bridges and do not assign IP directly to interfaces.
I configured a switch with LACP and it passes two VLANs (10 and 20) to Proxmox hosts. One is for NFS, the other one is for VM migration.
So, it will be just Proxmox node to storage communication for NFS traffic and VM migration traffic.
There will not be VMs that will pass any traffic through this bond interface.

I tried two configurations:

1. Bond
auto bond2
iface bond2 inet manual
bond-slaves ens1f0 ens2f1
bond-miimon 100
bond-mode 802.3ad
bond-xmit-hash-policy layer2+3
#Storage. Port channel 10 to storage sw

auto bond2.10
iface bond2.10 inet static
address 192.168.100.1/24

auto bond2.20
iface bond2.20 inet static
address 192.168.1.1/24
#Migration

2. Bridge
auto bond2
iface bond2 inet manual
bond-slaves ens1f0 ens2f1
bond-miimon 100
bond-mode 802.3ad
bond-xmit-hash-policy layer2+3
#Storage. Port channel 10 to storage sw

auto vmbr2_10
iface vmbr2_10 inet static
address 192.168.100.1/24
bridge-ports bond2.10
bridge-stp off
bridge-fd 0
#Storage switch

auto vmbr2_20
iface vmbr2_20 inet static
address 192.168.1.1/24
bridge-ports bond2.20
bridge-stp off
bridge-fd 0
#Migration switch

In both configurations I can ping NFS storage from Proxmox hosts (haven't tried to actually add it) and 'migration' IPs.
Should I use bonded interface or a bridge? Any benefits of using a bridge? Does creation additional layer (in this case - vmbr) adds extra latency?

Thank you.
 
Hi @serg2002 , welcome to the forum.

Unless you plan to use the underlying interface for VMs, there is no reason to create bridge on top of the bond.


Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
That is what I thought initially.
There is no plan to use VM via this link. It's even not physically possible, because storage network is isolated.
I tried to search online and tried to ask ChatGPT (not the best source of truth, but still)
It firstly advised to utilize the bond interface directly, but as long as I mentioned the NFS storage will be used for VMs, it immediately changed it's recommendations to create a bridge.
"...you should use a vmbr for the NFS VLAN if the datastore will host running VMs."
"...Proxmox treats storage networks that back running VMs differently from “pure host services”.
"The VM lifecycle now depends on that network."