Learn Kubernetes survival kit
Intermediate Free
Kubernetes survival kit
Enough Kubernetes to be dangerous in a good way: the core objects, safe rollouts and rollbacks, debugging the three failures you'll actually hit, config and secrets, and autoscaling including event-driven scaling with KEDA.
- 5 lessons
- 37 min total
- Quiz in every lesson
What you'll learn
- A Pod is the smallest deployable unit - one or more containers that share network and storage, scheduled together on the same node.
- A Deployment's default rolling update replaces old Pods with new ones gradually, controlled by maxSurge and maxUnavailable, instead of an all-at-once replacement.
- CrashLoopBackOff means the container keeps starting and exiting - the fix is in the application/config, not the cluster; kubectl logs (with --previous) and kubectl describe pod are the first two commands, always.
- A ConfigMap holds non-sensitive configuration; a Secret holds sensitive values - they're used almost identically (env vars or mounted files) but Kubernetes treats them differently for access control and display.
- The Horizontal Pod Autoscaler (HPA) adjusts a Deployment's replica count based on observed metrics (CPU/memory by default) against a target you set - it changes replica count, not per-Pod resource limits.
Curriculum
- 1. Pods, Deployments, and Services The three objects almost everything else in Kubernetes builds on, and how they relate to each other. 8 min (completed)
- 2. Rollouts and rollbacks How a Deployment replaces Pods safely, readiness probes as the real safety net, and rolling back a bad release fast. 7 min (completed)
- 3. Debugging CrashLoopBackOff, OOMKilled, and Pending pods The three failure states you'll actually hit, what each one means, and the exact commands to diagnose them. 8 min (completed)
- 4. Config and secrets ConfigMaps and Secrets, how to actually inject them into a Pod, and why a Secret alone isn't encryption at rest. 7 min (completed)
- 5. Autoscaling, including KEDA HPA scaling on CPU/memory, why that's not enough for many real workloads, and event-driven scaling with KEDA. 7 min (completed)