Proxmox Cluster NFS: Frontend Connects, Backend Fails

The PVE cluster can connect to NFS via the dashboard from the frontend but fail from the backend? It’s the routing. Your NFS traffic is going out the wrong interface. Here’s the fix.

Why the frontend works but the backend doesn’t

The cluster dashboard runs on the web UI, which uses the node’s management IP. That IP is on your frontend network, and it can reach the NFS server just fine. But when Proxmox actually mounts the NFS share, it uses the storage backend, which tries to connect from a different network—usually the cluster or storage network. If that network doesn’t have a route to the NFS server, the mount fails. Simple as that.

Check your routing table

First, see what routes you have. On the node that’s failing, run:

ip route show

Look for a route to your NFS server’s subnet. If it’s missing or going through the wrong interface, that’s your problem. You’ll probably see the frontend subnet routed correctly, but the backend subnet missing or pointing to a dead gateway.

Add the missing route

To fix it temporarily, add a route for the NFS server’s network through the correct backend interface. For example, if your NFS server is at 192.168.10.50 and your backend interface is vmbr1 with gateway 192.168.10.1:

ip route add 192.168.10.0/24 via 192.168.10.1 dev vmbr1

To make it permanent, add the route to your network config. In Proxmox, that’s /etc/network/interfaces. Add a post-up line to the backend interface:

auto vmbr1
iface vmbr1 inet static
    address 192.168.10.10/24
    gateway 192.168.10.1
    post-up ip route add 192.168.10.0/24 via 192.168.10.1 dev vmbr1

Then restart networking or reboot. Now try adding the NFS storage again from the cluster dashboard. It should connect from both sides.

Don’t overthink it

This is a classic split-network mistake. You set up a separate backend network for storage traffic, but forgot to tell the node how to reach the NFS server on that network. The frontend works because it’s a different subnet with a default route. The backend fails because there’s no route. Add the route, and you’re done.

If you’re still stuck, check your NFS server’s export settings—make sure it allows connections from the backend subnet, not just the frontend. That’s another common gotcha. And if you’re dealing with NFS on Proxmox, you might also run into IO delay issues later, but that’s a different beast.