Kubelings βΈοΈ
An interactive, client-side WebAssembly learning platform and comprehensive reference manual for Kubernetes.
β‘ The Modern Way to Master Kubernetes
Kubelings combines a zero-install, 100% client-side WebAssembly interactive playground with 26 comprehensive architectural reference guides spanning the entire cloud-native ecosystem.
- :material-play-circle-outline: **Zero-Install Web IDE**
---
Run Monaco Editor, Pyodide WebAssembly, and real-time schema validation 100% in your browser. No Docker, no minikube, and no cluster setup required.
[**Launch Playground β**](playground/index.html){ .md-button .md-button--primary }
- :material-book-open-page-variant-outline: **26-Chapter Reference Manual**
---
Deep architectural documentation, annotated YAML specs, production best practices, and diagnostic workflows for modern Kubernetes.
[**Explore Reference Guides β**](#-comprehensive-26-chapter-reference-guides){ .md-button }
π Comprehensive 26-Chapter Reference Guides
Explore in-depth architectural guides and launch linked practice exercises directly into the playground:
- ### Core Workloads & Storage
---
- [**01. Pods & Core Workloads**](guides/01-pods.md) • Pod specs, multi-container pods, initContainers, ports
- [**02. Controllers & Replication**](guides/02-controllers.md) • Deployments, ReplicaSets, StatefulSets, DaemonSets, Jobs, CronJobs
- [**03. Configuration & Secrets**](guides/03-config-secrets.md) • ConfigMaps, Secrets, projected volumes, envFrom
- [**04. Storage & Persistent Volumes**](guides/04-storage.md) • PVs, PVCs, StorageClasses, access modes, volumeMounts
- ### Networking & Traffic Routing
---
- [**05. Services & Networking**](guides/05-services-networking.md) • ClusterIP, NodePort, LoadBalancer, Headless services, Endpoints
- [**06. Ingress & Gateway API**](guides/06-ingress-gateway.md) • Ingress controllers, path rules, TLS termination, annotations
- [**21. Gateway API Deep Dive**](guides/21-gateway-api.md) • GatewayClass, Gateways, HTTPRoute, canary splits, ReferenceGrant
- [**09. Network Policies**](guides/09-network-policies.md) • Default deny, ingress/egress CIDR blocks, port rules
- ### Scheduling, Security & Scaling
---
- [**07. Scheduling & Advanced Placement**](guides/07-scheduling.md) • nodeSelector, node/pod affinity, taints, tolerations, topology spread
- [**08. Security, RBAC & ServiceAccounts**](guides/08-security-rbac.md) • Roles, ClusterRoles, RoleBindings, SecurityContext, PSS/PSA
- [**10. Health Probes & Lifecycle**](guides/10-lifecycle-probes.md) • Liveness, readiness, startup probes, termination grace periods
- [**11. Autoscaling (HPA, VPA, KEDA)**](guides/11-autoscaling.md) • Horizontal Pod Autoscaler v2, VPA, event-driven KEDA
- [**18. Admission Webhooks**](guides/18-admission-webhooks.md) • Mutating, validating webhooks, sidecar injection, CRD conversion
- ### CRDs, Troubleshooting & Packaging
---
- [**12. Custom Resources & Operators**](guides/12-crds-and-operators.md) • OpenAPI v3 schemas, subresources, reconciliation loops
- [**13. Production Troubleshooting**](guides/13-troubleshooting.md) • CrashLoopBackOff, ImagePullBackOff, Pending pods, ephemeral debug
- [**19. Package Management with Helm**](guides/19-helm-packaging.md) • Chart.yaml, Go templates, _helpers.tpl, values schemas, subcharts
- [**20. Declarative Customization with Kustomize**](guides/20-kustomize-overlays.md) • Bases, overlays, generators, strategic merge patches
- ### Modern Cloud Native Ecosystem
---
- [**14. GitOps with ArgoCD**](guides/14-gitops-argocd.md) • Applications, ApplicationSets, sync policies, Argo Rollouts
- [**15. Service Mesh & Cilium eBPF**](guides/15-service-mesh-cilium.md) • L7 routing, mTLS, FQDN egress, Hubble observability
- [**16. Policy as Code (Kyverno & OPA)**](guides/16-policy-as-code.md) • ClusterPolicies, mutating/generating rules, Gatekeeper constraints
- [**17. Multi-Tenancy & Virtual Clusters**](guides/17-multitenancy-vcluster.md) • HNC anchors, tenant quotas, vcluster control planes
- [**22. Infrastructure as Data with Crossplane**](guides/22-crossplane-iac.md) • XRDs, Compositions, Managed Resources, claims
- [**23. Kernel Security with eBPF Tetragon**](guides/23-ebpf-tetragon.md) • sys_execve tracing, credential monitoring, kernel Sigkill
- [**24. Distributed AI Orchestration with KubeRay**](guides/24-kuberay-ml.md) • RayCluster, RayJob batch tuning, RayService serving
- [**25. Batch AI Scheduling (Kueue & Volcano)**](guides/25-batch-kueue-volcano.md) • ClusterQueue cohorts, gang scheduling, fair-share
- [**26. Hardware Acceleration & DRA**](guides/26-hardware-acceleration-dra.md) • NVIDIA MIG, Apple Silicon GPU, DRA, vLLM serving
π‘ How the Playground Works
The Kubelings web playground runs entirely on client-side WebAssembly technology:
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β Web Browser Tab β
β βββββββββββββββββββββββββ βββββββββββββββββββββββββββββββ β
β β Monaco Editor β YAML Edit β Web Worker (Pyodide Wasm) β β
β β (VS Code in Browser) ββββββββββββββΊβ β’ PyYAML Manifest Parser β β
β βββββββββββββββββββββββββ β β’ 26 Schema Validators β β
β β β’ Progressive Hint Engine β β
β βββββββββββββββββββββββββ ββββββββββββββββ¬βββββββββββββββ β
β β Interactive xterm ββββββββββββββββββββββββββββββ β
β β Terminal Output β Instant Test & Validation Result β
β βββββββββββββββββββββββββ (< 1ms in WebAssembly) β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
- Monaco Code Editor: Full syntax highlighting, error diagnostics, and side-by-side solution diffing.
- Pyodide WebAssembly Engine: Executes the exact Python schema validation engine in an isolated Web Worker thread.
- Local Progress Persistence: Tracks completed exercises and user edits in
localStoragewith JSON export/import support.
π The *lings Ecosystem
If you enjoy the hands-on learning model of Kubelings, check out the other interactive projects in the suite:
- ποΈ Terralings β Master Terraform and OpenTofu through interactive infrastructure-as-code exercises.
- πͺπΈ Spanglings β Developer-grade interactive TUI for learning Spanish (B1βC1).
- β‘ Raylings β Learn distributed AI, Ray Core actors, and scalable clusters through hands-on Python exercises.
π License
Kubelings is open-source software licensed under the Apache License 2.0.