Skip to main content
KBY Technologies Logo
All Engineering Labs
SimulatorAdvanced

Kubernetes Troubleshooter

A scenario-based diagnostic Lab that practises moving from Kubernetes symptoms to evidence, containment and a safe first remediation. It rewards disciplined investigation instead of broad restarts or destructive guesses.

Around 10 min Saved private report Advanced level
Purpose and audience

Who this Lab is for

Designed for

  • Platform and DevOps engineers
  • On-call engineers supporting Kubernetes
  • Graduates building production diagnostic habits

Use it when

  • Practising incident response before joining an on-call rota
  • Teaching evidence-led Kubernetes troubleshooting
  • Refreshing diagnostic choices for common failure modes
How to use it

A complete run, step by step

1

Choose a scenario

Select scheduler pressure, a failed rollout or intermittent service errors.

2

Identify the primary symptom

Separate the observed symptom from an assumed root cause.

3

Select discriminating evidence

Choose evidence that safely narrows the hypothesis before changing the cluster.

4

Contain and remediate

Prefer a reversible first action that preserves evidence and limits customer impact.

Input guide

What you will need

Prepare the following information before starting. Use measured evidence where possible; defaults are examples and should not be treated as recommendations.

Available scenarios

Scheduler pressure

Pending workloads after a node-pool change.

Failed rollout

Crash loops and readiness failures after deployment.

Intermittent 503s

Endpoint and dependency failures across replicas.

Primary symptom

select

Choose the strongest observed signal.

Choices: Pods remain Pending · CrashLoopBackOff after rollout · Intermittent 503 responses

First evidence to inspect

select

Prefer evidence that narrows the hypothesis safely.

Choices: Pod events and scheduling conditions · Restart every pod · Increase all resource limits · Node CPU dashboard only

Initial response

select

Select the most reversible first move.

Choices: Contain rollout and preserve evidence · Delete the affected namespace · Disable probes globally · Scale every workload to zero

Results and methodology

What the result tells you

Your report includes

  • A scored diagnostic sequence
  • Feedback on evidence quality and remediation safety
  • A repeatable practice record across scenarios

How it is determined

Choices are scored for diagnostic value, reversibility and blast-radius control. Evidence-first actions receive more credit than global restarts, probe removal or destructive namespace operations.

Indicative rubric · low confidence · v2026.07.1

The simulator grades diagnostic discipline; it has no cluster telemetry and cannot identify a production root cause.

Model assumptions

  • The selected symptom is the strongest observed signal.
  • Preserving evidence is safer than broad mutation.
  • The scenario omits provider and workload-specific behaviour.
Worked example

Pods remain Pending

Situation

A workload stops scheduling after a node-pool change.

Result

Inspecting pod events and scheduling conditions before changing resources leads to a strong score and a focused capacity or constraint hypothesis.

Important limitations

Use the result with engineering judgement

  • Scenarios simplify real clusters and do not execute commands.
  • Always follow your organisation's incident and access procedures.
  • Production remediation should be validated against current cluster state.
Frequently asked questions

Questions before you begin

Do I need a Kubernetes cluster?

No. The Lab is a decision simulator and can be used without infrastructure access.

Are commands included?

The Lab focuses on diagnostic reasoning. Use your approved runbooks and current Kubernetes documentation for exact commands.

Can I try every scenario?

Yes. Each variant can be run and saved separately to build a broader practice history.

Kubernetes Clinic is under review

This legacy judgement-based Lab has been retired. Existing saved reports remain available, but new execution is disabled.

Open deterministic utilities