41_patch_and_reboot_nodes.yml - Usage Guide

Modifies State Maintenance

⚠️ Can update packages and reboot live nodes

A default run only reports pending packages and whether a reboot is needed. Changes require node_patch_confirm=true.

Confirmed maintenance flow:

  1. If it is a Kubernetes node, drain via the first control-plane
  2. apt dist-upgrade on Debian, dnf update on RedHat
  3. Reboot if reboot was also confirmed and is required
  4. Uncordon the node
  5. On error, attempt uncordon from the rescue block

Hosts are processed one at a time with serial: 1.

Variables

Variable Default Description
node_patch_confirm false Enables package updates
node_reboot_confirm false Allows a reboot when required
node_reboot_always false Reboots even if there is no reboot flag
node_allow_single_node_maintenance false Extra confirmation for singlenode maintenance
kubernetes_node_name ansible_hostname Node name in the Kubernetes API

How to run 

# Report only:
ansible-playbook -i inventories/musteri_a/hosts.ini playbooks/41_patch_and_reboot_nodes.yml

# Patch a worker and reboot if required:
ansible-playbook -i inventories/musteri_a/hosts.ini playbooks/41_patch_and_reboot_nodes.yml \
  --limit worker1 \
  --extra-vars 'node_patch_confirm=true node_reboot_confirm=true'

A singlenode cluster also requires node_allow_single_node_maintenance=true. Before maintenance, verify a current etcd backup with 32_verify_etcd_backup.yml.