What it is
CloudWatch collects metrics, logs, and events from AWS resources. You can graph them, search them, and set alarms when something crosses a line.
It is how you know the box is bored, the error rate spiked, or the disk is full. It is not magic. Someone has to choose what to watch and who gets the page.
Azure Monitor is the twin. Same job. Same empty dashboard if nobody wired an action.
Why it matters in the meeting
A green architecture slide without alarms is optimism. When Bart asks "how would we know," he means CloudWatch (or Monitor), not vibes.
JJ discovers monitoring when the outage is already a customer thread. Context before the meeting is cheaper than theatre after.
Real world
Alarms that email a distribution list nobody reads are decoration. Alarms that page a human with a runbook are operations.
Logs you never retain cannot help an audit. Logs you retain forever without a lifecycle are another storage bill. Pick a story and stick to it.
CloudWatch is the instrument panel. An alarm with no subscriber is a check-engine light in a locked garage. Wire it to a person, or admit you are flying by feel.
What to ask
- Which production alarms page a human, versus emailing a black hole?
- What is our log retention, and who pays for it?
- Can we answer "is it broken" in one dashboard before the war room?
- When did we last test that an alarm actually fires?
You just knew a little more Jack than you did five minutes ago.
All concepts