Proxmox cluster formed cleanly but its heartbeat ran over the management network
A three-node Proxmox cluster appeared healthy and reported quorum after creation and joining. On closer inspection, one node's corosync cluster heartbeat was using its management interface instead of the intended server network, because the join command had no flag to pin the link address. The setup worked only because the management network happened to route to the other nodes, leaving a latent risk. Such a problem could have caused a node to fence and reboot if that network were firewalled or disrupted. The author fixed it by editing the corosync configuration, correcting the ring address, incrementing the config version, and restarting corosync node by node while quorum was maintained.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in