Focused, hands-on problems for sharpening your DevOps and server-side skills. Some teach specific concepts while others mirror real-world scenarios. Every challenge includes hints, feedback, and automated solution checks.
A helper script that lists every container image running in the cluster — along with its on-disk size and the workload that uses it — has been downloaded to cplane-01. Install it as a kubectl plugin named kubectl-images so that kubectl images works from any directory.
Create a Pod that uses a generic ephemeral volume backed by a dynamic StorageClass. Unlike emptyDir, ephemeral volumes are provisioned as PersistentVolumeClaims and support storage requests and access modes.
Add the official Argo CD Helm repository, then use helm template to render the Argo CD chart for a dedicated namespace while skipping the chart's bundled CRDs, since they are already installed in the cluster.
Configure a Vertical Pod Autoscaler (VPA) with a per-container resource policy that provides recommendations for the main container while excluding the sidecar container from any VPA-managed scaling.
A webapp deployment in the prod namespace is stuck in CrashLoopBackOff. The application's startup probe checks that its existing config directory is intact before the container is marked healthy. Fix the deployment so configuration is injected correctly and the pod reaches Running state with its health check endpoint responding on port 8080.
A MariaDB deployment and its PVC were accidentally deleted. The PersistentVolume survived due to its Retain policy. Restore the database by rebinding the existing PV without losing any data.
A deployment is running without health probes or graceful shutdown configuration. Add startup, liveness, and readiness probes plus a preStop hook and terminationGracePeriodSeconds to make it ready for real traffic.
Create a PriorityClass for production workloads and patch an existing deployment to use it, ensuring critical services get scheduled first during resource contention.
A service in the mesh namespace returns only a single IP during DNS lookup, preventing the peer-server pods from finding each other. Investigate why DNS does not return one address per pod and change the service configuration so that peer discovery works correctly.
Add a new, separate Service CIDR range for the dev team using the ServiceCIDR API object — without touching or restarting the cluster's default Service CIDR — then create a Service with an explicit ClusterIP inside that new range and verify DNS resolution.