Skip to content
Ameen Abozaid
Englishالعربية
MID-048

Monitoring

  • 5 Lessons
  • 2.5 h practice (estimate)
Start module
On this page 8 topics

Overview

Topics

  • Prometheus
  • Grafana
  • dashboards
  • alerting

Learn first

2.5 h of practice (estimate)

Learning outcomes

  1. Scrape metrics with Prometheus and write PromQL for rates, ratios, and percentiles.
  2. Build Grafana dashboards using the RED and USE methods.
  3. Write alerts on symptoms rather than causes, with clear runbook links.

Lessons

  1. 01Prometheus model
  2. 02PromQL essentials
  3. 03Dashboards: RED and USE
  4. 04Alerting that people trust
  5. 05Demo: a service dashboard from scratch

Lab and exercises

Lab

MID-048-LAB1 · Shop dashboards

Outcomes: 1–3 · Time: ~2.5 h Build one dashboard per service plus a business overview. Add alerts for error rate and latency, each with a runbook. Acceptance: a fault injected by the instructor triggers exactly one actionable alert.

Exercises (assessment: dashboard and alert review)

IDOutcomeTask
MID-048-EX011Configure Prometheus to scrape a Shop service and confirm the target is up.
MID-048-EX021Write PromQL for request rate, error ratio, and p95 latency of orders.
MID-048-EX031Explain why rate() is needed for counters and what a range of [5m] means.
MID-048-EX042Build a RED dashboard and a USE row (heap, GC, DB pool) for one service.
MID-048-EX053Write an error-rate alert with for:, severity, and a runbook link.
MID-048-EX063Rewrite a CPU-above-80% alert as a symptom-based alert.
MID-048-EX072–3Review a given dashboard and alert set: list what to delete, fix, and add.
Show solutions
  • EX01 Add a scrape config or a service/pod monitor; the targets page shows UP.
  • EX02 Rate: sum(rate(..._count[5m])); ratio as in the lecture; p95 with histogram_quantile.
  • EX03 Counters only increase and reset on restart; rate gives per-second increase over the window and handles resets.
  • EX04 Panels: rate, errors, p50/p95/p99, heap used, GC pause, pool active versus max.
  • EX05 expr on the error ratio, for: 5m, label severity: page, annotation runbook_url.
  • EX06 Alert on checkout error ratio and p95 latency; keep CPU as a dashboard panel (cause), not a page.
  • EX07 Typical findings: averages, no runbook links, alerts on causes, duplicate alerts for one fault.

Outcome 3 is also checked by the lab criterion "exactly one actionable alert".

Quiz

Outcomes: O1 Prometheus and PromQL · O2 dashboards (RED and USE) · O3 alerting.

A self-check for you: nothing is graded, saved or sent anywhere. Choose an answer, then check it. You can try again as often as you like.

  1. 1Prometheus collects metrics by:O1
  2. 2A counter is read with:O1
  3. 3Which gives an error percentage?O1
  4. 4histogram_quantile(0.95, ...) returns:O1
  5. 5RED stands for:O2
  6. 6USE dashboards cover:O2
  7. 7A dashboard nobody reads should be:O2
  8. 8Alerts should be based mainly on:O3
  9. 9Every actionable alert has:O3
  10. 10The for: clause in an alert rule:O3

Practical part

Seven exercises MID-048-EX01 to EX07 (EX07 is the dashboard and alert review) plus the dashboards lab. Pass: at least five exercises and a lab where an injected fault triggers exactly one actionable alert.

Common mistakes

Alerting on CPU; dashboards nobody reads; alerts without runbooks.

Assessment

Dashboard and alert review.

Career relevance

SLO engineering (SRC-012).

Type to search levels, phases, modules and lessons.