DevOps Engineer Interview Experience at IBM – 2025

Candidates are asked to explain Kubernetes architecture and the role of each control plane component - the API server as the single entry point for all cluster operations, etcd as the cluster's source of truth, the scheduler's job of assigning pods to nodes based on resource availability and constraints, and the controller manager's role in reconciling actual cluster state toward desired state. From there, the discussion walks through creating and managing pods via kubectl, and explaining how a Service differs from a Pod - a Service provides a stable virtual IP and DNS name that load-balances across a set of pods, which is necessary because individual pod IPs are ephemeral and change whenever a pod is recreated.

This round covers building and managing Docker images efficiently - specifically why layer ordering in a Dockerfile matters for build cache reuse, and how a multi-stage build reduces final image size by discarding build-time dependencies from the final image. The Linux scripting portion asks for a script that tails a log file and alerts if a specific error pattern appears more than a set number of times within a rolling time window. The Python portion asks for a script automating a repetitive deployment task, specifically checking a service's health endpoint and retrying with exponential backoff before declaring a deployment failed.

This round is a discussion of previous DevOps projects, typically focused on a migration from a VM-based deployment to containers, the specific challenges encountered (primarily around persistent storage and configuration management that had previously been handled by hand on the VM), and the solutions applied (moving configuration into ConfigMaps and Secrets, and persistent data into a properly provisioned PersistentVolume rather than relying on local disk).