Skip to main content

Module 3: Container Orchestration: Worked Examples

Example 1: Diagnose Pending Pods​

Pending is a symptom. Inspect events, requests versus node allocatable capacity, selectors/affinity, taints/tolerations, PVC binding, and quotas. Adding nodes is wrong if a selector makes every node ineligible or a volume cannot bind.

Example 2: Separate Probes​

Startup probes protect slow initialization; readiness decides whether traffic should arrive; liveness detects a process that cannot recover without restart. Pointing all three at a deep dependency check can restart healthy pods during a database outage and amplify failure.

Example 3: Make a Rollout Reversible​

Pin immutable image digests, define requests/limits, use multiple replicas and disruption constraints, set rollout progress deadlines, run smoke checks, and retain a compatible previous version. Database changes must follow expand/migrate/contract rather than assuming application rollback reverses schema.

Completion Standard​

  • Diagnose scheduling from evidence.
  • Demonstrate correct probe behavior during dependency failure.
  • Execute and roll back a deployment without losing compatibility.