Proxmox: Removing a Ghost Node from the Web UI

This is a follow-on to Proxmox Cluster Going Sluggish? Your Offline Node Has a Stale Config. That post covers nodes that are misbehaving but still real. This one covers nodes that don't exist at all.

---

After sorting out a stale corosync config on a returning node, I noticed the web UI was still showing an extra node — PVE9 — on every host in the cluster. It wasn't causing any problems, just sitting there looking wrong. Here's how to get rid of it.

What a Ghost Node Is

A ghost node is a stale directory in /etc/pve/nodes/ with no corresponding corosync membership. The Proxmox web UI reads from the shared cluster filesystem, not from corosync directly — so a leftover directory shows up as a node even if the machine is long gone. It can appear after a node was removed uncleanly, rebuilt under a different name, or never properly decommissioned.

Check It's Actually a Ghost

First confirm the node isn't just offline — check corosync:

pvecm status

Membership information
----------------------
    Nodeid      Votes Name
0x00000001          3 10.140.3.10
0x00000002          1 10.140.3.80
0x00000003          1 10.140.3.70
0x00000004          1 10.140.3.82
0x00000006          1 10.140.3.20

If it's not in this list, it's a ghost. Confirm the directory exists:

cat /etc/pve/corosync.conf | grep pve9   # nothing
ls /etc/pve/nodes/
# pve1  pve2  pve7  pve8  pve9  xenon

Check for VMs Before Deleting

The node directory may contain VM or container configs:

ls /etc/pve/nodes/pve9/qemu-server/
ls /etc/pve/nodes/pve9/lxc/

If there are configs, check them before doing anything:

cat /etc/pve/nodes/pve9/qemu-server/102.conf

If the node is truly gone, any disks listed as local-lvm:vm-XXX-disk-Y were on that node's local storage and are already inaccessible. You won't be able to recover them. Make sure you're happy with that before proceeding.

Remove It

Try the proper route first:

pvecm delnode pve9

If the node was never in corosync you'll get:

Node/IP: pve9 is not a known host of the cluster.

In that case, remove the directory directly:

rm -rf /etc/pve/nodes/pve9

The deletion replicates across the cluster filesystem immediately. Verify:

ls /etc/pve/nodes/
# pve1  pve2  pve7  pve8  xenon

Reload the web UI — the ghost node is gone.

Quick Reference

Command	Purpose
pvecm status	Confirm node isn't in corosync
ls /etc/pve/nodes/	List all node directories
ls /etc/pve/nodes/<name>/qemu-server/	Check for VM configs
ls /etc/pve/nodes/<name>/lxc/	Check for container configs
pvecm delnode <name>	Proper removal (works if node was in corosync)
rm -rf /etc/pve/nodes/<name>	Manual removal for true ghost nodes