DevOps Engineer Interview Questions: CI/CD, K8s, Incidents

DevOps, SRE and cloud engineer interviews focus on how you ship changes safely, run systems reliably, and respond when production breaks. Expect questions about CI/CD, containers and Kubernetes, infrastructure as code, observability, and incident response, plus scenario questions and behavioral stories. This page covers the questions that come up most, grouped by topic, with what a strong answer includes. It also covers the Docker, Kubernetes and cloud questions that used to have their own pages.

How the process usually runs

1. Recruiter screen: experience with cloud, tooling, on-call.
2. Technical screen: Linux and networking fundamentals, scripting, sometimes a short troubleshooting scenario.
3. Loop: system or pipeline design, a troubleshooting round, and behavioral questions about incidents and collaboration.

Linux, networking and fundamentals

• "What happens when you type a URL into a browser?" DNS lookup, TCP connection, TLS handshake, HTTP request, load balancing, server response, rendering. Know where each step can fail.
• "A server is slow. How do you investigate?" Check load, CPU, memory, disk I/O and network (for example with top, vmstat, iostat, and ss), then look at logs and recent changes, and narrow down before changing anything.
• "What is the difference between a process and a thread?" A process has its own memory space; threads share memory within a process.

CI/CD and safe delivery

• "Describe a good deployment pipeline." Build once, test automatically, scan, deploy to a staging environment, then roll out to production gradually with automated checks and a fast rollback.
• "How do you roll out a risky change?" Canary or blue-green deployment, feature flags, defined metrics to watch, and a rollback trigger agreed beforehand.
• "How do you handle database migrations in a continuous deployment flow?" Backward-compatible changes in stages (expand, migrate, contract) so old and new code can run together.

Containers and Kubernetes

• "What is the difference between a container and a virtual machine?" Containers share the host kernel and isolate processes; VMs run full guest operating systems on a hypervisor. Containers start faster and are lighter; VMs give stronger isolation.
• "Explain a Kubernetes pod, deployment and service." A pod is one or more containers scheduled together; a deployment manages replicas and rolling updates; a service gives a stable network endpoint for a set of pods.
• "A pod is stuck in CrashLoopBackOff. What do you do?" Check the pod's events and logs (kubectl describe, kubectl logs, including the previous container), look for failing probes, missing config or secrets, resource limits, and image problems.
• "How do liveness and readiness probes differ?" Liveness decides whether to restart a container; readiness decides whether it receives traffic.

Infrastructure as code and cloud

• "What is infrastructure drift and how do you handle it?" Differences between declared and actual infrastructure. Detect with plan or drift tools, reconcile through code review, and limit manual changes.
• "How do you manage Terraform state safely?" Remote state with locking, separate state per environment or component, restricted access, and careful review of plans.
• "How do you design for high availability on a cloud provider?" Multiple availability zones, stateless services behind a load balancer, managed databases with replication and backups, and tested failover.
• "How do you manage secrets?" A secrets manager, least privilege, rotation, and no secrets in code or images.

Observability and incidents

• "What do you alert on?" User-facing symptoms (error rate, latency, availability) rather than every internal cause, with alerts that are actionable and routed to the right people.
• "Tell me about a serious incident." A strong answer has five parts: detection, mitigation first, diagnosis, communication, and the lasting fix. Include time to detect and recover if you know them.
• "What is an SLO, and how do you use an error budget?" An SLO is a target for reliability; the error budget is the allowed unreliability, which guides how fast to ship versus stabilize.

Worked answer: "How would you roll out a risky change?"

Outline: Say what makes it risky. Propose a staged rollout: a small canary, watch error rate and latency against a baseline for a set period, then expand. Define the rollback trigger before starting (for example, error rate above a threshold for five minutes). Mention a feature flag to turn off the behavior without a deploy, and what you will tell stakeholders. Close with how you would verify success.

Common mistakes

• Naming tools without the failure they prevent.
• Going to root cause in an incident story before mitigation.
• No numbers: time to detect, time to recover, impact.
• Never admitting a change you made contributed to an outage.

Practice

Tell your best incident in five steps in under three minutes, and re-answer the rollout question until the rollback trigger is explicit. The DevOps practice page has a follow-up chain for a resume line, and the system design framework helps with design rounds.

Frequently asked questions

Which cloud provider should I focus on?

The one in the job description. Core concepts (networking, IAM, compute, storage, managed databases) transfer between providers.

How much coding is expected?

Usually scripting and automation in Python, Go or Bash, plus reading configuration. Some roles include a coding round.

What is the difference between DevOps, SRE and platform engineering?

The boundaries vary by company. SRE emphasizes reliability targets and error budgets, platform engineering builds internal tooling for other developers, and DevOps emphasizes delivery practices and collaboration.