The most detailed free FDE + DevOps library: 140+ lessons, 70+ labs and 80 long-form articles, in English and Turkish. Start learning →

TürkçeSearch

DevOps Practitioner · Module 7: Applied observability and SLOs · Lab

Fix a Wrong Alert Threshold with Fixture Data

45 min hands-on · Core

A local machine with a small metric fixture (a file series or a tiny metrics endpoint); no cluster needed. All numbers are fixtures and labeled as such.

Local guide: run the steps below on your own machine in order, then check the validation list.

Objectives

  • Show a threshold alert firing on noise and missing on real burn
  • Recompute the threshold from the fixture SLI and budget
  • Prove the corrected alert fires on burn and stays quiet on noise

  1. Step 1

    Catch the wrong threshold

    Take the provided alert with a fixed error-count threshold. Replay the fixture series: quote one noise spike that pages and one slow burn it misses. Both failures go in the record.

  2. Step 2

    Recompute from budget

    Compute the fixture SLI and error budget, then set burn-rate thresholds (fast burn pages, slow burn tickets) from the budget math. Write the calculation down.

  3. Step 3

    Replay and compare

    Replay the same series against the corrected alert. Quote the burn firing the page, the noise staying quiet, and the trend raising a ticket. Label every number a fixture.

How to confirm it worked

  • Both threshold failures quoted from the same series
  • Budget math written with SLI and burn rates
  • Corrected alert quoted firing on burn, quiet on noise
  • Fixture labeling present throughout