Quiz — Kubernetes Fundamentals¶
Quiz Overview¶
Validate Kubernetes fundamentals used on the job: workload APIs, networking abstractions, probes, config/storage, RBAC, and production operations judgement.
| Attribute | Value |
|---|---|
| Topic | Kubernetes |
| Questions | 40 |
| Passing score | 70% (28 correct) |
| Estimated time | 45–60 minutes |
| Format | Multiple choice (single answer) |
How to use this quiz
Attempt each section without peeking. Use Reveal answer only after you commit to a choice. Score yourself honestly — gaps are the point.
Learning Objectives¶
This quiz assesses whether you can:
- Explain Pods, Deployments, Services, and Namespaces
- Distinguish liveness vs readiness failure behaviour
- Use kubectl for logs, exec, and describe-driven triage
- Diagnose ImagePullBackOff, CrashLoopBackOff, scheduling, and probe mistakes
- Choose sensible production patterns (HPA, PDB, security baselines)
Section 1 — Fundamentals¶
Question 1¶
What is a Pod in Kubernetes?
Difficulty: Beginner
Options:
- A. The smallest deployable unit: one or more containers sharing network namespace and volumes
- B. A physical rack
- C. A cloud region
- D. A Helm repository only
Reveal answer
Correct answer: A
Pods group containers with shared networking/storage abstractions.
Related concepts
- Pods
Question 2¶
Which control plane component schedules Pods onto nodes?
Difficulty: Beginner
Options:
- A. CoreDNS only
- B. etcd alone
- C. kube-scheduler
- D. Ingress controller
Reveal answer
Correct answer: C
kube-scheduler assigns Pods to nodes.
Related concepts
- Control plane
- kube-scheduler
Question 3¶
What does a Deployment manage?
Difficulty: Beginner
Options:
- A. PersistentVolumes directly
- B. Only Nodes
- C. etcd snapshots
- D. ReplicaSets/Pods for declarative, rolling updates of stateless workloads
Reveal answer
Correct answer: D
Deployments own ReplicaSets and provide rollout strategies.
Related concepts
- Deployment
- ReplicaSet
Question 4¶
A ClusterIP Service provides:
Difficulty: Beginner
Options:
- A. A public load balancer on every cloud by default
- B. A stable virtual IP/DNS name for Pods inside the cluster
- C. Physical disk attachment
- D. TLS certificates automatically
Reveal answer
Correct answer: B
ClusterIP is internal service discovery and load balancing.
Related concepts
- Service
- ClusterIP
Question 5¶
What happens when a liveness probe fails repeatedly?
Difficulty: Beginner
Options:
- A. kubelet restarts the container
- B. Pod is removed from Service endpoints only
- C. The node is cordoned automatically
- D. etcd is compacted
Reveal answer
Correct answer: A
Failed liveness → container restart. Failed readiness → remove from endpoints.
Related concepts
- livenessProbe
Question 6¶
What happens when a readiness probe fails?
Difficulty: Beginner
Options:
- A. Container is always killed immediately
- B. Namespace is deleted
- C. Pod is taken out of Service endpoints until ready again
- D. API server restarts
Reveal answer
Correct answer: C
Readiness controls whether the Pod receives traffic via Services.
Related concepts
- readinessProbe
- Endpoints
Question 7¶
ConfigMaps are primarily for:
Difficulty: Beginner
Options:
- A. Storing private keys as best practice
- B. Replacing etcd
- C. Billing data only
- D. Storing non-sensitive configuration data
Reveal answer
Correct answer: D
Non-sensitive config belongs in ConfigMaps; use Secrets for sensitive material.
Related concepts
- ConfigMap
Question 8¶
kubectl apply is associated with which management style?
Difficulty: Beginner
Options:
- A. Imperative only
- B. Declarative configuration applied to the cluster
- C. SSH into nodes to edit binaries
- D. Manual etcd edits
Reveal answer
Correct answer: B
apply reconciles desired manifests with live objects.
Related concepts
- kubectl apply
Question 9¶
Namespaces are used to:
Difficulty: Beginner
Options:
- A. Partition cluster resources and scope names for multi-tenancy/environments
- B. Replace VPCs in all clouds
- C. Store container images
- D. Disable RBAC
Reveal answer
Correct answer: A
Namespaces scope objects and policies within a cluster.
Related concepts
- Namespaces
Question 10¶
Which object requests durable storage for a Pod?
Difficulty: Beginner
Options:
- A. Ingress
- B. ServiceAccount token only
- C. PersistentVolumeClaim (bound to a PersistentVolume)
- D. HorizontalPodAutoscaler
Reveal answer
Correct answer: C
PVCs request storage; PVs are the provisioned volumes.
Related concepts
- PVC
- PV
Section 2 — Practical Knowledge¶
Question 11¶
Which command tails logs from a Pod named api-x in namespace prod?
Difficulty: Beginner
Options:
- A. docker compose logs
- B. journalctl -u api-x
- C. helm uninstall api-x
- D. kubectl logs -n prod api-x -f
Reveal answer
Correct answer: D
kubectl logs with -n and -f follows Pod logs.
Related concepts
- kubectl logs
Question 12¶
You need to open a shell in a running container:
Difficulty: Beginner
Options:
- A. kubectl delete pod
- B. kubectl exec -it -n
-- sh - C. kubectl cordon
- D. kubectl proxy only
Reveal answer
Correct answer: B
exec attaches a command/shell into the container.
Related concepts
- kubectl exec
Question 13¶
ImagePullBackOff most often means:
Difficulty: Intermediate
Options:
- A. kubelet cannot pull the image (name/tag/auth/registry/network)
- B. The scheduler is missing
- C. The Service CIDR is wrong
- D. HPA is disabled
Reveal answer
Correct answer: A
Check image reference and pull secrets/registry connectivity.
Related concepts
- ImagePullBackOff
Question 14¶
CrashLoopBackOff indicates:
Difficulty: Intermediate
Options:
- A. Node memory is always healthy
- B. DNS is perfect
- C. Pod starts then the container exits repeatedly
- D. RBAC is unused
Reveal answer
Correct answer: C
Investigate logs, probes, and command failures.
Related concepts
- CrashLoopBackOff
Question 15¶
Which field sets CPU/memory requests and limits on a container?
Difficulty: Intermediate
Options:
- A. replicas only
- B. hostNetwork: true
- C. strategy.type
- D. resources.requests / resources.limits
Reveal answer
Correct answer: D
requests affect scheduling; limits cap usage.
Related concepts
- Resources
- QoS
Question 16¶
A Deployment rolling update should keep capacity. Which helps?
Difficulty: Intermediate
Options:
- A. delete the namespace mid-rollout
- B. maxUnavailable / maxSurge settings appropriate to SLO
- C. set replicas to 0 first always
- D. disable readiness probes permanently
Reveal answer
Correct answer: B
RollingUpdate parameters control surge and unavailability.
Related concepts
- RollingUpdate
Question 17¶
RBAC: a Role is scoped to:
Difficulty: Intermediate
Options:
- A. A namespace (Role) vs cluster-wide (ClusterRole)
- B. Only etcd
- C. Only Ingress
- D. Only container images
Reveal answer
Correct answer: A
Role is namespaced; ClusterRole is cluster-scoped.
Related concepts
- RBAC
- Role
- ClusterRole
Question 18¶
Helm primarily helps you:
Difficulty: Intermediate
Options:
- A. Replace container runtimes
- B. Compile the Linux kernel
- C. Package, templatise, and release Kubernetes manifests as charts
- D. Provision bare metal only
Reveal answer
Correct answer: C
Helm is the common package manager for Kubernetes apps.
Related concepts
- Helm
Question 19¶
HPA scales Pods based on:
Difficulty: Intermediate
Options:
- A. Git commit messages
- B. Docker Hub stars
- C. Manual /etc/hosts edits
- D. Metrics such as CPU/memory/custom metrics (when metrics pipeline exists)
Reveal answer
Correct answer: D
Horizontal Pod Autoscaler uses resource/custom metrics.
Related concepts
- HPA
Question 20¶
PodDisruptionBudget (PDB) is meant to:
Difficulty: Advanced
Options:
- A. Encrypt etcd automatically
- B. Limit voluntary disruptions so a minimum number/percentage of Pods stay available
- C. Replace NetworkPolicies
- D. Disable node upgrades forever
Reveal answer
Correct answer: B
PDBs protect availability during drains/upgrades.
Related concepts
- PDB
Section 3 — Scenario-Based Questions¶
Question 21¶
Intermittent 503s during deploy. Readiness path /healthz 404 while / works. Effect?
Difficulty: Intermediate
Options:
- A. Pods may never become Ready, so Services send no traffic (or only to old pods) — fix probe to a real path
- B. Nodes reboot
- C. etcd loses quorum always
- D. Images become unsigned
Reveal answer
Correct answer: A
Bad readiness blocks endpoints. Align probes with the app contract.
Related concepts
- Readiness
- Probes
Question 22¶
Deployment stuck. describe shows FailedScheduling / Insufficient cpu. Fix direction?
Difficulty: Intermediate
Options:
- A. Delete kube-apiserver
- B. Change the app language
- C. Reduce requests, add nodes, or free capacity — scheduling cannot place the Pod
- D. Disable all Services
Reveal answer
Correct answer: C
Scheduler needs fitting node resources.
Related concepts
- Scheduling
- Requests
Question 23¶
Secret mounted as env still shows old value after you apply a new Secret.
Difficulty: Advanced
Options:
- A. Kubernetes never mounts Secrets
- B. Only Helm can update Secrets
- C. Need to reboot every node
- D. Pods do not automatically reload env from updated Secrets — rollout restart / redesign for dynamic reload
Reveal answer
Correct answer: D
Env-injected secrets are fixed at process start for typical apps.
Related concepts
- Secrets rotation
Question 24¶
GitOps reports drift: live Deployment differs from Git. Correct response?
Difficulty: Advanced
Options:
- A. Always kubectl edit and ignore Git
- B. Change Git (or allowlisted sync) — avoid long-lived kubectl edit that fights the reconciler
- C. Turn off the cluster
- D. Delete Git history
Reveal answer
Correct answer: B
Git is the source of truth; reconcile toward desired state.
Related concepts
- GitOps
- Drift
Question 25¶
Ingress returns 404 but Service curl from inside cluster works. Likely layer?
Difficulty: Intermediate
Options:
- A. Ingress rules/host/path or Ingress controller config
- B. Container runtime missing
- C. PVC pending only
- D. HPA minReplicas
Reveal answer
Correct answer: A
Ingress is the external HTTP routing layer.
Related concepts
- Ingress
Question 26¶
You need zero-downtime drain of a node. Useful sequence?
Difficulty: Intermediate
Options:
- A. power off the node immediately
- B. delete the Deployment
- C. cordon + drain (respecting PDBs), ensure replicas elsewhere, then maintain
- D. remove CoreDNS first
Reveal answer
Correct answer: C
cordon/drain is the supported maintenance workflow.
Related concepts
- kubectl drain
- PDB
Question 27¶
NetworkPolicy default-deny in a namespace means new Pods:
Difficulty: Advanced
Options:
- A. Automatically get Internet without rules
- B. Bypass Services
- C. Ignore DNS
- D. May be isolated until explicit allow policies select them
Reveal answer
Correct answer: D
Default-deny requires careful allowlists for expected traffic.
Related concepts
- NetworkPolicy
Question 28¶
Init container fails; app container never starts. Why?
Difficulty: Intermediate
Options:
- A. Init containers run after the app
- B. Init containers must succeed before app containers start
- C. Init containers replace the scheduler
- D. Init containers only run on Windows
Reveal answer
Correct answer: B
Init sequence gates startup.
Related concepts
- Init containers
Question 29¶
type LoadBalancer on bare metal without cloud integration. What happens?
Difficulty: Advanced
Options:
- A. EXTERNAL-IP may stay pending unless MetalLB/equivalent provides IPs
- B. Kubernetes invents a public IP always
- C. The Pod becomes a Node
- D. RBAC is disabled
Reveal answer
Correct answer: A
LoadBalancer needs an implementation to allocate addresses.
Related concepts
- LoadBalancer
- MetalLB
Question 30¶
Liveness probe too aggressive on a slow-starting JVM. Symptom?
Difficulty: Intermediate
Options:
- A. Faster cold starts magically
- B. More endpoints join early
- C. Container killed/restarted before it becomes healthy — use startupProbe or retune thresholds
- D. etcd grows slower
Reveal answer
Correct answer: C
startupProbe or longer delays prevent premature liveness kills.
Related concepts
- startupProbe
Section 4 — Troubleshooting¶
Question 31¶
Pod Pending; events: failed to pull image myapp:latst (typo). Root cause?
Difficulty: Beginner
Options:
- A. CoreDNS crash only
- B. Missing Deployment name
- C. PDB conflict
- D. Wrong image reference/tag
Reveal answer
Correct answer: D
Typos in image names cause pull failures.
Related concepts
- Image references
Question 32¶
Service selects app=api but Pods are labelled app=backend. Effect?
Difficulty: Beginner
Options:
- A. Automatic label rewrite
- B. Endpoints empty — no traffic to Pods
- C. Nodes cordon
- D. HPA scales to zero always
Reveal answer
Correct answer: B
Selector/label mismatch yields empty Endpoints.
Related concepts
- Selectors
- Endpoints
Question 33¶
OOMKilled in container status. Meaning?
Difficulty: Intermediate
Options:
- A. Exceeded memory limit (or node pressure killing)
- B. CPU throttle only
- C. DNS NXDOMAIN
- D. Successful probe
Reveal answer
Correct answer: A
OOMKilled is a memory kill signal outcome.
Related concepts
- OOMKilled
Question 34¶
kubectl auth can-i delete deployments -n prod returns no. Next?
Difficulty: Intermediate
Options:
- A. Disable RBAC cluster-wide
- B. Use --force on API server flags casually
- C. Check RoleBindings/ClusterRoleBindings for your identity; request least privilege access
- D. Store kubeconfig in a public gist
Reveal answer
Correct answer: C
Fix RBAC bindings appropriately; do not disable RBAC.
Related concepts
- RBAC troubleshooting
Question 35¶
PVC Pending forever in a cluster without default StorageClass. Likely?
Difficulty: Intermediate
Options:
- A. Ingress misconfigured
- B. HPA missing
- C. Too many ConfigMaps
- D. No provisioner/StorageClass to satisfy the claim
Reveal answer
Correct answer: D
Dynamic provisioning needs a StorageClass; otherwise bind a matching PV.
Related concepts
- StorageClass
- PVC Pending
Section 5 — Architecture¶
Question 36¶
For a stateless HTTP API, which pairing is most typical?
Difficulty: Intermediate
Options:
- A. one naked Pod on a specific node name forever without probes
- B. Deployment + ClusterIP/Ingress + HPA + PDB
- C. hostNetwork for every Pod
- D. Privileged DaemonSet as the web tier
Reveal answer
Correct answer: B
Stateless HTTP commonly uses Deployment, Service/Ingress, and availability controls.
Related concepts
- Workload design
Question 37¶
Why prefer requests+limits over unlimited containers?
Difficulty: Intermediate
Options:
- A. Predictable scheduling and noisy-neighbour protection
- B. They slow the API server only
- C. Required to use ConfigMaps
- D. Disable metrics
Reveal answer
Correct answer: A
Resources enable fair scheduling and containment.
Related concepts
- Capacity planning
Question 38¶
Multi-tenant cluster security baseline includes:
Difficulty: Advanced
Options:
- A. A single cluster-admin kubeconfig for everyone
- B. Privileged pods by default
- C. Namespaces, RBAC least privilege, NetworkPolicies, admission controls, non-root images
- D. No audit logs
Reveal answer
Correct answer: C
Defence in depth across identity, network, and workload controls.
Related concepts
- Multi-tenancy
- Hardening
Question 39¶
Blue/green vs rolling update — trade-off summary?
Difficulty: Advanced
Options:
- A. They are identical
- B. Rolling always needs two clusters
- C. Blue/green forbids Services
- D. Blue/green shifts traffic between versions with more capacity cost; rolling gradually replaces replicas in one Deployment
Reveal answer
Correct answer: D
Choose based on risk, capacity, and traffic switching needs.
Related concepts
- Deployment strategies
Question 40¶
Where should cluster state backups focus first?
Difficulty: Advanced
Options:
- A. Only worker node /tmp
- B. etcd (or control plane backup strategy) plus application data volumes
- C. Only container stdout
- D. Only CNI IPAM files on one node
Reveal answer
Correct answer: B
Control plane state and persistent app data are critical.
Related concepts
- Disaster recovery
- etcd
Score Summary¶
| Score | Band | Meaning |
|---|---|---|
| 36–40 | Excellent | Ready for interviews and labs at this track level |
| 28–35 | Good | Pass — revise weak sections, then retry |
| 20–27 | Needs improvement | Revisit tutorials for missed topics before labs |
| ≤19 | Restart foundations | Work the track in order, then retake |
Total questions: 40 · Passing score: 28 (70%)
Recommended Study Areas¶
Weak on API objects → early Kubernetes tutorials. Weak on probes/rollouts → Health Checks tutorial + Deployment triage lab. Weak on RBAC/security → RBAC and hardening tutorials.
- Track: Kubernetes
- Lab: Kubernetes Deployment Triage
- Cheat sheet: Kubernetes
- Interview prep: Kubernetes
- Path: DevOps Engineer
Interview Connection¶
Be ready to explain probes, CrashLoopBackOff vs ImagePullBackOff, Service selectors, requests vs limits, and a drain/PDB story.