PVE - Removing a node from a cluster

From Da Nerd Mage Wiki
Revision as of 11:12, 11 August 2025 by Tinker (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

To safely remove a node from a Proxmox VE (PVE) cluster, follow these steps:

  1. Migrate Workloads:
    • Before removing the node, ensure all virtual machines (VMs) and containers are migrated to other active nodes in the cluster. This can be done via live or offline migration using the Proxmox web GUI or command line tools like qm migrate.
  2. Verify Cluster Status:
    • Log in to an active node in the cluster and run pvecm nodes to confirm the list of nodes and identify the name or ID of the node you wish to remove.
  3. Remove the Node from the Cluster:
    • Execute the pvecm delnode command from an active node, replacing [NODE_NAME] with the actual name or ID of the node to be removed. If the command fails due to a loss of quorum, particularly in a two-node cluster, you can temporarily set the expected votes to 1 using pvecm expected 1 before running pvecm delnode again.
  4. Power Down the Removed Node:
    • Once the node is successfully removed from the cluster configuration, power it off and disconnect it from the network to prevent it from attempting to rejoin the cluster with its old configuration.
  5. Clean Up Configuration Files:
    • To completely remove the node's presence from the cluster's configuration, log in to one of the remaining active nodes and delete the node's directory from /etc/pve/nodes/. This step is necessary to remove the node from the Proxmox web interface. Additionally, on the removed node itself, you should delete the remaining cluster files to ensure it can be reused in another cluster without issues. This includes removing files from /etc/pve/corosync.conf, /etc/corosync/*, and /var/lib/corosync/*.


For clusters using Ceph storage, additional steps are required after removing the node from the cluster: stop and destroy the Ceph Metadata Servers (if running CephFS), remove the node's OSDs from the Ceph crush map, and finally remove the node's bucket from the CRUSH hierarchy.

The Official Instructions

  • Remove a Cluster Node (There IS some dumb in these instructions...)
    • If you still see the removed node in the GUI afterwards...
    • You need to remove the whole directory /etc/pve/nodes/<nodename>
Retrieved from ‘https://wiki.nerdmage.ca/index.php?title=PVE_-_Removing_a_node_from_a_cluster&oldid=3797’