# 30 DevOps and SRE interview questions, and what a good answer shows

> Thirty DevOps and site reliability engineer interview questions on Linux, networking, containers, CI/CD, observability and incidents, each with what a strong answer covers.

Source: https://hintvo.com/blog/devops-sre-interview-questions | Last updated: November 20, 2026 | Reviewed: 2026-11-20

DevOps and SRE interviews test how you think when production misbehaves: what you check first, how you limit the damage and how you stop it from happening again. Below are 30 common questions, grouped by topic, each with what a good answer shows.

## Linux and networking

- **A server is slow. What do you check first, and why?** A good answer looks at CPU and load, memory and swap, disk space and I/O, and the network, and says what each number would tell them instead of reciting commands.
- **What happens when you request an HTTPS address with curl?** A good answer walks through DNS, the TCP handshake, TLS with certificate checks and the HTTP request, and how to tell which step failed.
- **A disk is full on a production host. What do you do?** A good answer finds what grew, such as logs, temporary files or deleted files still held open, frees space safely, then fixes the cause with rotation, retention and an alert before it fills again.
- **What is a zombie process, and is it a problem?** A good answer explains it is a child process that has exited but has not been collected by its parent, harmless alone, and a sign of a bug in the parent when they pile up.
- **Why do some users still reach the old server after a DNS change?** A good answer explains time-to-live values and caching resolvers, and lowers the TTL well before a migration.
- **How does a load balancer decide where to send traffic?** A good answer covers balancing methods, health checks, sticky sessions and what they cost, and the difference between balancing connections and balancing HTTP requests.

## Containers and orchestration

- **What is the difference between a container and a virtual machine?** A good answer explains that containers share the host's kernel and are isolated processes, while a virtual machine runs a whole guest system, and what that means for isolation and start time.
- **A pod keeps restarting. How do you find out why?** A good answer reads the pod's events, the previous container's logs and the exit reason, such as being killed for memory, and checks probes that fail too early and missing configuration.
- **What are liveness and readiness probes, and how can they cause an outage?** A good answer says readiness takes a pod out of traffic and liveness restarts it, and that an aggressive liveness probe under load can restart healthy pods in a loop.
- **How do you set CPU and memory requests and limits?** A good answer sets them from measured usage, knows that exceeding a memory limit kills the container while a CPU limit throttles it, and that requests drive scheduling.
- **How would you make a container image smaller and safer?** A good answer uses multi-stage builds, a minimal base image, no build tools in the final image, a non-root user, pinned versions and an image scan.
- **How do you roll out a change safely?** A good answer uses rolling or canary deploys with automatic rollback on errors or latency, and [feature flags](https://hintvo.com/blog/feature-flags-in-plain-words) to separate deploying code from releasing a feature.

## Pipelines and infrastructure as code

- **What should a CI pipeline check before a merge?** A good answer runs the build, tests, linting, type checks and security scans, keeps it fast enough that nobody skips it, and fixes or quarantines flaky tests.
- **Why manage infrastructure as code, and what goes wrong without it?** A good answer names repeatable environments and reviewed, recorded changes, and the drift and one-of-a-kind servers that appear without it.
- **How does a team share Terraform state safely?** A good answer uses remote state with locking, separate state per environment, no manual edits to resources, and a reviewed plan before every apply.
- **How do you handle secrets in pipelines?** A good answer uses the CI system's encrypted secrets or a secret manager, prefers short-lived credentials, masks logs and rotates on a schedule.
- **A deploy failed halfway through. What now?** A good answer stops the rollout, restores service first by rolling back or forward, then finds the cause, and makes deploys safe to run again.

## Observability and incidents

- **What is the difference between metrics, logs and traces?** A good answer uses metrics for trends and alerts, logs for detail and traces for one request across services, and says which answers which question.
- **What makes an alert good?** A good answer wants alerts that need action, follow what users feel rather than every internal cause, reach someone who can act and link to a runbook, and removes noisy ones.
- **Explain SLIs, SLOs and error budgets.** A good answer says an SLI measures what users experience, such as the share of successful requests, an SLO is the target for it, and the error budget is the failure allowed before reliability work takes priority over features.
- **How would you run an incident call?** A good answer gives the call a lead, someone who communicates and someone who keeps the timeline, restores service before finding the cause, and posts updates on a schedule. See [incident calls](https://hintvo.com/blog/incident-call-find-the-code-path-fast).
- **What belongs in a postmortem?** A good answer includes the timeline, the impact, the contributing causes, what went well and changes with owners and dates, written without blame. See [blameless postmortems](https://hintvo.com/blog/blameless-postmortem-meeting).
- **Latency spiked for ten minutes and recovered on its own. How do you find the cause?** A good answer reads metrics around the window, looks for deploys and configuration changes, traces slow requests, and checks dependencies and saturation of CPU, connections and queues.
- **How would you make on-call less painful for a team?** A good answer counts pages, removes noisy alerts, fixes the most frequent causes, writes runbooks and automates the common fixes.

## Cloud, cost and security

- **How would you design a service to survive losing one availability zone?** A good answer spreads instances across zones, removes single-zone dependencies, lets health checks move traffic and tests the failure on purpose.
- **The cloud bill doubled this month. How do you find out why?** A good answer breaks cost down by service and tag, lines it up with recent changes, checks data transfer and storage growth and idle resources, then adds budgets and alerts.
- **What does least privilege look like for a service's permissions?** A good answer gives each service its own role with only the actions and resources it needs, avoids long-lived keys and reviews access regularly.

## Behavioral, for DevOps and SRE roles

- **Tell me about the worst incident you handled.** A good answer covers their role, how they stayed calm and kept people informed, the fix and the lasting change.
- **Tell me about manual work you automated.** A good answer measures the work it removed and explains how they made the automation safe to trust.
- **How do you push back when a team wants to ship something risky?** A good answer brings data about the risk, offers a safer path such as a flag, a canary or more tests, and reaches a shared decision instead of a veto.

## How to practice these

1. Pick the five questions you would find hardest, not the ones you already know.
2. Answer each one out loud in about two minutes, and record yourself if you can.
3. Compare your answer with the note: did it cover what a good answer shows?
4. For every question, write down one real example from your own work that you could use.

For interviewers: the notes work as a scoring guide. Decide before the interview what a strong answer covers, and score the evidence you heard, not the confidence it was said with. See [how to run a fair technical interview](https://hintvo.com/blog/run-a-fair-technical-interview).

More roles and the questions every engineer gets: [interview questions by role](https://hintvo.com/blog/software-engineer-interview-questions-by-role).

> **Where Hintvo fits:** a practice interview asks questions out loud, gives you a score out of 5 and a stronger version of each answer, and on coding questions you can ask for a hint, the approach or the full solution. Practice interviews need a pack of hours. Use Hintvo only where AI help is allowed. Check the rules of your company, school or interviewer first, and respect a no.

## Related

- [Incident calls: find the code path while people are talking](https://hintvo.com/blog/incident-call-find-the-code-path-fast.md): How do I find the failing code path quickly during an incident call?
- [Blameless postmortem meetings: an agenda and a template](https://hintvo.com/blog/blameless-postmortem-meeting.md): How do I run a blameless postmortem meeting after an incident?
- [200 software engineer interview questions by role, and what good answers show](https://hintvo.com/blog/software-engineer-interview-questions-by-role.md): What interview questions do software engineers get for each role, and what makes a good answer?
