Monitoring

How this question breaks down

1. 1MetricsCore20 minTurn every event into a few numbers per series, collect them on a fixed beat, and store them compressed so a fleet's health fits on one screen and costs the same at 10 or 10,000 requests a second.
2. 2AlertingCore21 minPage a person only when users are hurting or about to be, and turn a storm of firing rules into one notification that says what is wrong and who should fix it.
3. 3TracingCore22 minGive each request an ID that every service passes along, record a timed span for each piece of work, and keep a well-chosen sample of the resulting trees so you can see where one slow request spent its time.

Work through the parts in order and you’ve covered the full design. Each part is one topic.