A Practical Checklist for the First Ten Minutes of a Slowdown

When performance drops, a calm sequence prevents random changes. First, confirm the symptom: which endpoint, region, and user action is slow, and when did it start? Check error rate and traffic volume together; latency with no traffic change suggests a different path than a sudden surge. Look at CPU, memory, disk wait, and network for the hosts serving the request. Then inspect dependency health: database locks, cache hit rate, queue depth, and external API timing. Compare current values with the same period yesterday to spot a pattern. Capture a trace or slow query sample before restarting anything. Finally, note every action and observation in a shared timeline so the team can reason about cause and effect. Which step do teams in your organization skip most often, and what was the cost?