Kubespray Node Rejoin

Overview

Rejoining a node with kubespray is a scoped ansible run plus verification. The failure modes are mostly around where you run it from, how you watch it, and what you check after — not the playbook itself.

Run it right

1. Scope with --limit — never run cluster-wide for a one-node repair:

   ansible-playbook -i inventory/<cluster>/inventory.ini scale.yml -b --limit=<node>
   

   (Control-plane rejoin may need cluster.yml with the node in the right groups; scale.yml is for adding workers.) Use the repo's pinned environment (venv) — kubespray is version-sensitive to ansible.
2. Run from a host that can actually reach the nodes. If your workstation can't SSH to the cluster's network, run the playbook on a host that can (over SSH with a generous timeout), not through hacks on your side.
3. Watch, don't block. A rejoin runs 10–40+ minutes. Run it as a background task and poll for the PLAY RECAP line rather than sitting on a blocking terminal that a dropped SSH connection kills. tmux/nohup + tailing the log both work.

Read the recap correctly

PLAY RECAP has two distinct failure classes: