From: Oluwaseun Adeyemi, on-call, Harrow Museum
"Ticket sales for the new exhibition opened ten minutes ago and the site is crawling. Nothing was deployed today. Find out what is wrong, fix it, and give me three lines for the incident log."
Done when: the node list and the kubelet journal are saved, the node is Ready, tickets runs 4 Ready replicas, and the summary names node, cause and fix.
A node incident in four steps: record the cluster's view, ask the node itself through its journal, restore the kubelet and confirm the workload, then write a summary that separates the trigger from the cause.
The cluster's view, the node's view, the fix.