Level 5 exam — Docker · GUIDINEE.COM

Exam — fresh questions

12 new questions to validate Orchestration & production: scale, Swarm, Kubernetes and prod checklist.
Answer everything, then click Check: your score and explanations will appear.

5.1 · Recall

1. What does Compose NOT do, justifying a move to an orchestrator?

Re-reading a services file Spreading across multiple hosts with self-healing Replaying a stack locally Launching containers on one machine

Compose stays single-host: no multi-machine scheduling, no automatic replacement of a dead node.

5.1 · Application

2. Idir runs on 1 server with simple updates. What do you advise him?

Doing everything by hand with no versioning Setting up a full fleet to learn Staying on Compose + CI + backups, simple and sufficient Multiplying tools before needs

For a simple need, a simple tool: well-equipped Compose beats an oversized orchestrator.

5.1 · Arbitration

3. The team wants Kubernetes “for the hype”, with no multi-host need. Idir's call?

Moving up a step only on proven need: staying on Compose Tinkering with a cluster in prod to learn Adding five tools around Accepting to look modern

Move Compose → Swarm → Kubernetes by real pain, never by hype.

5.2 · Recall

4. What is needed for a zero-downtime Swarm update?

Replicas + one-by-one update + healthcheck proving readiness Stopping everything then relaunching Deleting volumes Luck

Replicas absorb, the one-by-one update preserves, the healthcheck validates each new one.

5.2 · Application

5. After moving multi-node, data goes missing intermittently. Why?

Too many logs displayed Wrong time zone Too much CPU consumed Local volumes not shared: each node has its disk

A local volume does not travel: you need shared storage or an external database. Pitfall #1.

5.2 · Application

6. Idir wants to see the state of each replica of his service. Which command?

No command, everything is hidden The command that lists services The command that shows each replica (state per task) The command that displays passwords

Listing replicas shows node, state and history: the deployment dashboard.

5.3 · Recall

7. What does the Deployment do when a Pod dies?

Nothing, everything must be relaunched It recreates a Pod to keep the desired replicas It deletes the Service It stops the cluster

The controller maintains the declared state: missing Pod = automatically recreated Pod.

5.3 · Application

8. Idir sometimes reaches an address that changes, sometimes one that stays. What do you remind him?

The Service is ephemeral Pods must be reached directly Both are the same The Pod is an ephemeral envelope, the Service a stable address in front

Pods are born and die, the Service stays: reach the stable, not the moving.

5.3 · Arbitration

9. Idir hesitates between a moving tag and a fixed tag in his Deployment. Your call?

Putting no tag Changing registry on each deployment Taking the moving one, it is simpler Taking an immutable tag: you know exactly what is running

Fixed tag = traceable updates and reliable rollback; the moving one drifts without warning.

5.4 · Application

10. One point on Idir's checklist is a “no”. Verdict?

Switching checklist Deploying anyway, we will see Fixing and re-testing before any prod Hiding the checklist

The checklist is a gate: a single “no” blocks the production release.

5.4 · Application

11. The Docker host disk fills up. Trio of suspects?

Keyboards and screens Accumulated images + rotation-less logs + orphan volumes RAM alone Network alone

The filling trio: old images, unbounded logs, abandoned volumes. Rotation + cleanup + alerts.

5.4 · Arbitration

12. Idir, a junior, wants to open prod alone after his audit. Your call?

Having checklist, secrets and monitoring validated by the senior/IT before opening Skipping the tested restore to save time Versioning secrets to go fast Opening alone, the audit is enough

A junior does not put into prod alone: supervised validation, tested backup, secrets out of the repo, proven monitoring.

What's next?

Score ≥ 80%: move to Level 6 “In practice” — real cases, troubleshooting and Docker toolbox (NO exam).
Score 50-79%: re-read the explanations above — each re-explains the concept — then redo the 48h challenges and the templates of the failed sections (Swarm stack, manifests, prod audit).
Score < 50%: resume cards 5.1 (scale by need) and 5.4 (eight yes before prod), your two vital moves, before continuing.