Node completely evacuates from Proxmox VE and rejoins the cluster

Node completely evacuates from Proxmox VE and joins the cluster again

Scene description

When a node in the Proxmox VE cluster is damaged and cannot be repaired quickly, the faulty node needs to be kicked out of the cluster cleanly and the remaining information must be cleaned up. Otherwise, new nodes using the IP address used by the faulty node will not be able to join the cluster normally; similarly, after the faulty node that has separated from the cluster is repaired, although it has nothing to do with the cluster, it will not be able to access the web management of this single node. In the background, information about other nodes in the original Proxmox VE cluster will appear, which is very annoying.

Evict nodes from the cluster

If Proxmox VE is a Ceph hyper-converged cluster, you need to log in to the host system Debian of any node in the cluster (except the node you want to delete), and completely delete all OSDs of the node you want to evacuate from the command line [see my book "Proxmox VE Hyper-converged Cluster Practice" Chapter 12 Related Content], continue to use the command line "pvecm delnode nodename" to evict the node. Switch to the Proxmox VE cluster web management background, refresh the browser, and check whether there are still evicted nodes in the "Server View" on the left side of the page. Even if the Web page no longer has information about the evicted nodes remaining, the information about the evicted nodes is not completely cleaned up and further processing is required.

Switch to the Debian command line of the host system of any node in the Proxmox VE cluster, enter the directory "/etc/pve/priv", and edit the files "authorized_keys" and "known_hosts".

Use the Vi editor or Sed command to delete the node information that was expelled from the cluster, save and exit. Other nodes in the cluster will also be automatically synchronized by corosync, so there is no need to repeat the process.

The file "Known_hosts" is also processed accordingly. After completing this step, use the command "grep pve169 /etc -r" to search to determine whether it is completely cleared.

Processing of residual information of the evacuated node cluster

The nodes that were evacuated from the Proxmox VE cluster have restored their functions without reinitializing (reinstalling the system) and want to join the original cluster again. When we use a browser to access the single-node Proxmox VE Web management backend, we may still see other nodes in the cluster clearly visible, and refreshing the page and restarting the system will not work. So, how to get rid of this annoying display?

In addition to stopping the cluster service, stopping the synchronization service Corosync, deleting corosync.conf and other operations, you also need to enter the Proxmox VE single-node hosting system Debian, enter the directory "/etc/pve/nodes", and delete except this node other node directories.

After deletion, refresh the browser, Proxmox VE single node Web management background, only the single node itself is left.

After completing the above two steps, the outlier node will join the cluster with its original IP and will no longer be excluded by the cluster.

The above is the detailed content of Node completely evacuates from Proxmox VE and rejoins the cluster. For more information, please follow other related articles on the PHP Chinese website!