aichallengesafrica
Take a challenge
HomeChallengesRepsProblemsLeaderboardTake a challenge

The dashboard that answers one question

Instrument one operation with rate, errors and duration, keep the tag cardinality survivable, and write the single alert that would have paged somebody before the customers noticed.

Advanced20 min on the clock6 graded checks · 100 pointsA Request You Can TracePro
Graded by reading your code against the checklist below.

What you build

  • A Meter, and counters for accepted and rejected
  • A histogram for how long it took
  • Tags of low cardinality only
  • The duration recorded even when it throws
  • An alert on a ratio, not on a raw count
  • A runbook link on the alert

Done means

One operation, three signals, a bounded number of series, and an alert somebody could act on at 3am.

How it is graded

Published in full, before you start — every point is one of these and there is nothing else.

  1. Accepted and rejected are counted separately+25Running it in production
  2. Duration is a histogram, so percentiles exist+20Running it in production
  3. Tags stay low cardinality+20Running it in production
  4. Duration is recorded even when it throws+15Error handling
  5. The alert is a ratio, not a raw count+15Running it in production
  6. The alert says what to do next+5Running it in production

What it teaches

The rest of A Request You Can Trace

What is worth loggingIntermediate · 20 minCorrelate across two servicesAdvanced · 25 min