DevOps and SRE interview questions with answers for engineering rounds
DevOps and SRE interviews rarely ask for definitions alone. They hand you a symptom, such as a slow server or a stuck rollout, and watch how you narrow it down. The strong answer names what you check first, what the likely causes are, how you limit the damage and how you confirm the fix. This guide gives twenty questions with short answers and follows the Kubernetes documentation and Google's SRE book.
How DevOps and SRE questions appear in a live interview
Expect a troubleshooting scenario, a design question about a pipeline or deployment, and one on-call or incident story. Say your assumptions aloud and work from the outside in. The same habit helps in a technical phone screen and a system design interview.
Junior DevOps and SRE interview questions
1. What is the difference between DevOps and SRE?
DevOps is a culture and set of practices that bring development and operations together around fast, safe delivery. SRE is a specific way to run it: apply software engineering to operations, with reliability targets and error budgets. Many companies blend the two titles, so ask how this team defines it.
2. How do you troubleshoot a Linux server that is slow?
Check load and the usual resources in order: CPU, memory and swap, disk space and I/O, then network. Use tools such as top, free, df, iostat and ss, read the logs, and find the one process or resource that stands out. State a hypothesis before each command.
3. What happens when you type a URL into a browser?
DNS resolves the name to an IP address, the browser opens a TCP connection, a TLS handshake secures it for HTTPS, then an HTTP request is sent and the response is rendered. At each step name what could fail: a stale DNS record, a blocked port, an expired certificate, a slow backend.
4. What is the difference between TCP and UDP?
TCP is connection-oriented and provides ordered, reliable delivery with retransmission. UDP is connectionless and sends datagrams without those guarantees, so it has less overhead. Web traffic mostly uses TCP; DNS queries and real-time media often use UDP.
5. What do file permissions chmod 750 mean?
Permissions are read 4, write 2 and execute 1, set for owner, group and others. So 750 gives the owner read, write and execute, the group read and execute, and others nothing. Mention least privilege and why scripts need the execute bit.
6. What is CI/CD?
Continuous integration merges and tests changes frequently so problems surface early. Continuous delivery keeps every passing change releasable; continuous deployment releases it automatically. Describe a pipeline you know: build, test, scan, package, deploy, verify.
7. What is infrastructure as code and why use it?
Describing infrastructure in versioned files so environments are reproducible, reviewable and recoverable instead of hand-configured. Changes go through code review, and drift between declared and actual state can be detected. Tools differ, so name the one you used and what went wrong once.
Mid-level DevOps and SRE interview questions
8. What is a Pod and why do you usually not create one directly?
A Pod is the smallest deployable unit in Kubernetes: one or more containers that share storage and network. As the Kubernetes documentation says, you normally create Pods through a workload resource such as a Deployment or Job, which replaces them when they fail.
9. Explain liveness, readiness and startup probes.
Per the Kubernetes probe guide: a failing liveness probe makes the kubelet restart the container; a failing readiness probe removes the Pod from Service endpoints without restarting it; a startup probe holds off the others while a slow application starts. Misusing liveness for a dependency check can cause restart loops.
10. A Deployment rollout is stuck. How do you investigate?
Check the rollout status, then the Pod events and states: image pull errors, failing readiness probes, insufficient resources or a crash loop. Read the logs of the new Pods and compare configuration with the old version. If users are affected, roll back first and investigate afterwards.
11. How do you design a deployment strategy with minimal risk?
Release gradually: rolling, blue-green or canary, with a health signal that decides whether to continue. Keep changes small, make rollback fast and tested, and separate database changes so the old and new code both work during the transition.
12. How do you manage secrets in a pipeline?
Keep them out of the repository, images and logs. Inject them at runtime from a secret store, give each pipeline and service the least access it needs, and rotate anything that leaked. Explain how you would find out a secret had been exposed.
13. What is the difference between a load balancer and a reverse proxy?
A reverse proxy sits in front of servers and forwards requests, often adding TLS termination, caching or routing. A load balancer spreads traffic across several backends, and many proxies do both. Say which layer you mean: layer 4 works on connections, layer 7 on HTTP content.
14. How do you debug a DNS problem?
Check what the client resolves with a tool such as dig, compare against the authoritative answer, and look at caching and TTL. Then confirm the path, resolver configuration and whether only some clients are affected. Many outages that look like network faults are stale or wrong records.
Senior DevOps and SRE interview questions
15. What are SLIs, SLOs and error budgets?
An SLI is a measurement of service behaviour, an SLO is the target for it, and the error budget is the unreliability you may spend. Google's SRE book chapter on risk describes using the budget to balance releases against reliability: when it runs out, releases slow until reliability recovers.
16. What are the four golden signals?
Latency, traffic, errors and saturation, from the SRE book chapter on monitoring. Together they give a first view of a service's health. Add that alerts should be about symptoms users feel, not every internal metric.
17. How do you reduce alert fatigue?
Alert only on conditions that need a human now and have a clear action; send the rest to tickets or dashboards. Tie alerts to SLO burn, add runbook links, and review noisy alerts after each on-call week. Delete alerts nobody acts on.
18. Walk me through how you would handle a major incident.
Declare it, assign an incident lead and a communicator, and stabilise first: mitigate, roll back or fail over before hunting the root cause. Keep a timeline, update stakeholders on a schedule, then close with a postmortem. The goal during the incident is restoring service, not proving a theory.
19. What is a blameless postmortem?
A review that looks for contributing causes in systems and processes rather than blaming a person. The SRE book on postmortem culture says it assumes everyone acted on the best information they had. It ends in tracked action items, not a document nobody reads.
20. How would you scale a service that is hitting its limits?
Measure the bottleneck first. Then choose: scale out stateless tiers, cache, add queues, optimise the hot query, or shard data. Load test the change, set autoscaling with limits and watch cost. Scaling the wrong component only moves the bottleneck.
A concise answer shape
For an operations scenario, say what you would check first, the likely causes in order, how you would limit the damage, and how you would confirm the fix. Example: for elevated errors after a release, I would compare the error rate against the deploy time, roll back if it matches, then read logs and traces to find the cause before releasing again.
Before any live assessment, ask the recruiter which tools and documentation are permitted. Knowing the material is useful; using unapproved tools during an assessment is a separate policy question. If you want to rehearse out loud, the free plan is one option, subject to the same rules. Related guides: Docker interview questions, system design follow-up questions, technical interview preparation.
Sources
- Kubernetes: Pods
- Kubernetes: liveness, readiness and startup probes
- Google SRE book: embracing risk, SLOs and error budgets
- Google SRE book: monitoring distributed systems, golden signals
- Google SRE book: postmortem culture
Checked 12 November 2026.
FAQ
What are common DevOps interview questions?
Linux troubleshooting, networking basics, CI/CD pipelines, infrastructure as code, containers and Kubernetes, monitoring and alerting, secrets handling and incident response. Senior rounds add SLOs, postmortems and capacity planning.
What is the difference between DevOps and SRE interviews?
DevOps interviews lean on pipelines, automation and tooling. SRE interviews add reliability targets, error budgets, on-call and incident handling, and often a troubleshooting scenario. Both test Linux and networking.
What Kubernetes topics should I know?
Pods, Deployments, Services, probes, resource requests and limits, ConfigMaps and Secrets, rolling updates and how to debug a failing Pod. Know the reasoning behind each, not just the commands.
What are the four golden signals in monitoring?
Latency, traffic, errors and saturation. The Google SRE book names them as the minimum for monitoring a user-facing service, and they work as a starting checklist in an interview.
Will I get a live troubleshooting question?
Often. You may be told a symptom and asked what you check. Think aloud, form hypotheses, and say how you would limit impact. Ask the recruiter what format and which tools are allowed.