AI HR
Prova

Interview questions

DevOps engineer interview questions

DevOps and platform interviews test breadth and calm. You will be asked how you build pipelines and infrastructure, then asked what you did when production went down and how you made sure it did not happen again.

Rehearse this interview by voice2 minutes without an account, 3 more after sign-up. No card.

Questions by stage and what a strong answer includes

  1. 01

    What does your current infrastructure look like?

    Strong answer: Cloud provider, compute (Kubernetes, ECS, VMs), IaC tool, CI/CD system, observability stack, and roughly how many services and environments. Your part in building it.

  2. 02

    Are you on call now, and what does a typical week look like?

    Strong answer: Rotation length, page volume, and what you did to reduce noise. It tells the recruiter how close you are to production.

  1. 03

    Tell me about an outage you led the response for.

    Strong answer: Detection, communication, mitigation, root cause and the actions from the postmortem, with times. Blameless tone.

  2. 04

    Describe a time you convinced developers to change how they deploy.

    Strong answer: The problem with the old way, how you made the new way easier, not just mandatory, and adoption numbers.

  1. 05

    Walk me through a CI/CD pipeline you built.

    Strong answer: Stages from commit to production: build, test, security scans, artifact storage, deploy strategy (blue-green, canary), approvals and rollback. Why each step exists.

  2. 06

    A pod is in CrashLoopBackOff. How do you debug it?

    Strong answer: kubectl describe and logs, previous container logs, exit code, resource limits and OOM kills, config and secrets, readiness and liveness probes. Systematic, one hypothesis at a time.

  3. 07

    How do you manage Terraform state and changes across a team?

    Strong answer: Remote state with locking, modules, separate state per environment, plan in CI with review, and how you handle drift and imports.

  4. 08

    What would you monitor for a web service, and when would you page someone?

    Strong answer: Golden signals or RED metrics, SLOs with error budgets, alert on symptoms users feel rather than every CPU spike, and runbooks linked to alerts.

  1. 09

    Design infrastructure for a service that must survive the loss of a region.

    Strong answer: Multi-region strategy, data replication and its consistency trade-offs, DNS or global load balancing, failover testing and the cost of each option.

  2. 10

    Our cloud bill grew 40% in six months. How would you approach it?

    Strong answer: Break down spend by service and team with tags, find idle and oversized resources, reserved or savings plans, storage lifecycle, and make costs visible to the teams who create them.

  1. 11

    What are your expectations around on-call and compensation?

    Strong answer: Clear questions about rotation size and on-call pay or time off, plus your range for the level. Experienced candidates always ask about on-call.

One answer, before and after

Tell me about an outage you led the response for.

Weak answer

We had a big outage when our Kubernetes cluster went down. I worked all night with the team to fix it. After that we improved our monitoring.

Strong answer in STAR

Situation
On a Tuesday morning our main Kubernetes cluster for a B2B SaaS product stopped scheduling new pods. Autoscaling could not add capacity and API latency climbed past 5 seconds.
Task
I was the on-call incident commander. I had to restore service and keep customer success updated so they could talk to clients.
Action
I opened an incident channel, assigned one engineer to investigate and one to communications, and posted updates every 15 minutes. We found that a node group upgrade had used an AMI that failed to join the cluster. I pinned the previous AMI, scaled a healthy node group manually and drained the broken nodes.
Result
Latency was back to normal 48 minutes after the first alert. The postmortem led to a canary node pool for upgrades and an alert on pending pods older than five minutes, which caught a similar issue a month later before customers noticed.

Why it works

The weak answer shows effort, not leadership. The strong one shows incident roles, a communication rhythm, a precise root cause and a fix that proved itself later. Numbers are illustrative: replace them with your own.

How a practice session works

  1. 01

    Pick a stage, paste the job posting

    Recruiter screen, HR interview, hiring manager, final round or offer talk. Add the posting and your resume so the questions are about your role, not generic.

  2. 02

    Talk it through by voice

    The interviewer speaks the language you chose, interrupts, follows up and digs into details like on a real call. If the connection drops, you pick up where you stopped.

  3. 03

    Get the review and drills

    A score, 7 skills, what you said next to a stronger way to say it, plus pace, filler words and talk share. Weak spots turn into short drills.

Rehearse this interview by voice

Related roles