Kubernetes Mastery

Core Philosophy

1. Declarative over imperative -- Define desired state in YAML manifests. Let controllers reconcile actual state. Never rely on kubectl run in production; commit manifests to Git.
2. Least privilege by default -- Every workload gets its own ServiceAccount with minimal RBAC. Run as non-root. Drop all capabilities. Apply Pod Security Standards at namespace level.
3. Resource-aware scheduling -- Always set CPU/memory requests (scheduler guarantee) and memory limits (OOM protection). Omit CPU limits for latency-sensitive workloads to avoid throttling.
4. Probes are self-healing -- Configure startup probes for slow-init apps, readiness probes to gate traffic, and liveness probes to restart deadlocked processes. Aggressive liveness probes cause restart storms.
5. GitOps is the deployment model -- ArgoCD or Flux syncs cluster state from Git. Manual kubectl apply is for emergencies only. Every change is auditable and reversible.

Quick-Start: Common Problems

"My Pod is stuck in CrashLoopBackOff"