DevOps & Cloud Interview Questions
DevOps interviews probe two things: can you build paths to production that are boring and repeatable, and do you stay calm when production is on fire. Tools change yearly; the underlying judgement is what gets scored.
Q1Design a CI/CD pipeline for a team shipping ten times a day. What gates does a change pass?
What they're checking: Whether you balance speed and safety or worship one of them.
A strong answer: Fast unit gate, parallel integration, artifact build once and promote, staged rollout with automatic rollback on error budgets. Strong answers include what deliberately is NOT gated to keep speed.
Q2Explain the difference between a Docker image and a container, and what layers actually are.
What they're checking: Fundamentals — many candidates use Docker daily without understanding it.
A strong answer: Image = immutable layered filesystem + metadata; container = running instance with a writable layer. Explain layer caching consequences for Dockerfile ordering and image size.
Q3A Kubernetes pod is CrashLoopBackOff. Walk your diagnosis.
What they're checking: Systematic debugging under the most common k8s failure.
A strong answer: describe pod for events, logs --previous for the dying container, check probes vs startup time, resources vs OOMKilled, config/secret mounts. Order matters — cheapest checks first.
Q4Why infrastructure as code? And what goes wrong when someone edits the console directly?
What they're checking: Drift understanding, not IaC evangelism.
A strong answer: Reproducibility, review, rollback — and drift: state no longer matches reality, next apply either reverts a hotfix or fails. Mention drift detection and break-glass procedures.
Q5How do you manage secrets across environments? What never goes in the repo?
What they're checking: Security hygiene and blast-radius thinking.
A strong answer: Secret managers with IAM-scoped access, short-lived credentials, rotation, audit. Nothing sensitive in code or CI logs — and per-environment isolation so staging leaks cannot touch prod.
Q6Define SLI, SLO and error budget — and what actually happens when the budget is spent.
What they're checking: Whether reliability is a number you manage or a slogan.
A strong answer: SLI measures, SLO targets, error budget is the allowed failure. Spent budget should gate feature releases in favour of reliability work — and a strong answer admits the org friction that causes.
Q7Blue-green vs canary deployments — when is each the right call?
What they're checking: Rollout strategy judgement including cost.
A strong answer: Blue-green for instant rollback and schema-compatible releases at 2x infra cost; canary for gradual risk exposure with real traffic but needs solid metrics to judge. Tie choice to failure detection ability.
Q8You get paged: p99 latency tripled but error rate is flat. Where do you look?
What they're checking: Performance triage instincts distinct from availability incidents.
A strong answer: Saturation first: CPU throttling, connection pools, GC pauses, a slow downstream dragging tail latency. Check recent deploys and traffic mix shift — p99 with flat errors usually means queueing.
Now try answering these under interview pressure
InterviewGPT runs a mock DevOps interview — asks questions like these tailored to your situation, listens to your spoken answers, and scores every one with written feedback. First session free, no credit card.
Start a free DevOps interview