Metrics, health checks and alert design
Create alerts that signal actionable conditions rather than noisy symptoms.
By the end of this lesson
- Explain metrics, health checks and alert design and when it is useful
- Perform a safe inspection or structured practice task
- Verify the result and describe remaining uncertainty
Your environment: Use an unprivileged account and a disposable Linux lab. Commands and available utilities vary by distribution and version. Where a command is unavailable or would require production access, complete the scenario as an offline written exercise. Setup help
Understand the concept
Metrics quantify behaviour over time, such as request latency, error rate, queue depth, saturation and resource use. A health check may test process liveness, readiness to receive traffic or a deeper dependency. These checks answer different questions. Alert thresholds should reflect service objectives, normal variation and the cost of delayed detection.
A useful alert explains what happened, where, how long it has persisted, likely user impact and the next safe action. Avoid high-cardinality labels and checks that overload the service they monitor. Test alert delivery and define ownership, escalation and suppression rules for planned work.
Apply a careful workflow
Monitor user-visible outcomes as well as infrastructure components. Confirm scope before consequential changes, keep a short record of expected and observed results, and consult local documentation when behaviour differs. Protect credentials and personal data in notes and logs.
Check your understanding
The goal of this lesson is not to memorise a single command. Explain why the proposed check is relevant, what its output can establish and what remains unknown. A successful command status is only one piece of evidence.
See it in context
Worked example
A bounded example to inspect in your own lab; adapt paths and options to your installed tools.
printf "%s\n" "metric=api_error_rate value=8.2% window=10m" "metric=p95_latency value=1.8s window=10m" > "$HOME/training4linux/alert-sample.txt"
cat "$HOME/training4linux/alert-sample.txt"Read each command and identify its target, inputs, expected evidence and possible failure modes. Do not run administrative or modifying commands on a production system.
Your turn
Practical lab
Design one alert for a fictional API error-rate increase, including threshold rationale, duration, impact text, owner and first diagnostic check.
Run commands in your own lab. This website displays instructions; the practice terminal below helps you rehearse the syntax.
1. Prepare your workspace
mkdir -p "$HOME/training4linux/extended-labs" && cd "$HOME/training4linux/extended-labs"
printf "%s\n" "Practice workspace for Metrics, health checks and alert design" > lab-notes.txt2. Complete the task
- Complete the scenario using the supplied synthetic information or a disposable local lab. Record what you observed and what you inferred.
- Compare your work against the success criteria and explain any differences.
- Write one verification step, one stop condition and a safe cleanup or recovery step.
3. Check your result
- The alert is actionable, time-bounded and linked to a documented response.
- The work remains within the disposable workspace or an offline case study.
- The report identifies at least one limitation or remaining uncertainty.
Need a hint?
Start with a read-only check, define the scope and choose the smallest next step that distinguishes competing explanations.
Reveal a worked solution
Compare this with your approach after trying the task. More than one valid solution may exist.
Use the success criteria as a review rubric. Record the reasoning and evidence behind your result rather than treating one exact command sequence as the only valid answer.After the lab: Retain your notes if useful. Remove only files created in the disposable practice directory; no production system should be changed.
Common mistakes to avoid
- A ping response does not establish that a business-critical request works.
- Do not treat a successful command exit as proof that the overall service or business outcome is correct.
Check your understanding
Knowledge check
Answer all 3 questions correctly to pass. Retry as often as you need; your best score is kept.
Your lesson progress
Not started · Best score: 0% · Attempts: 0
Practical completion is self-reported. Progress is linked to a private cookie in this browser and stored on the site. Clearing that cookie loses access to it. Manage or erase progress.