AWS · Domain
Operations and delivery
Observability, automation, stacks, delivery pipelines and recovery. 1 world, 6 stages, about 2.5 hours to complete.
Ops World 1 · Associate
Watchtower
A metric is a number with a period, an alarm is arithmetic over periods, a log is a token with a retention, and a trace is a request remembering where it waited.
- 1Measure ItFreeKestrel's sorting fleet runs out of memory at the cutoff and nobody sees it. Give every requirement a source at the lowest monthly bill.
- 2Alarm ActionsThe scan API fleet must page ops within three minutes of sustained high CPU, scale out by itself, and never page for a single spike.
- 3Keep the LogsLog events as tokens through retention, class, filters, destinations and masks. Each one lands where the requirement says, or the rule says why not.
- 4Trace the RequestLabel requests through API Gateway, Lambda and DynamoDB. Each one is sampled or not, carries segments or not, and can be found by depot or not, by a rule you can read.
- 5One PaneFive audiences, five panes. Each one sees every account and Region it needs and nothing it should not.
- 6Watch DeskTickets from the watchtower. The metric, the history, the agent's log or the group's settings names the cause before you touch anything.