Country or region and language

    Choose your country or region and your language.

    Suggested for you

      All countries and regions

        Choose your character

        It shows in the top bar and on your profile. Change it any time.

        Projects/Velomint

        Recover from a failed release

        Release 2.14.0 broke one bike unlock in three until on-call rolled it back. Revert the bad commit in Git, ship 2.14.1 on Kubernetes behind a readiness probe, add the CloudWatch alarm that would have caught it, and write the incident report.

        ProjectVelomint · fictional company2 h4 stagesCloud and DevOps

        One brief, 3 toolsOne release carries through every tool: the commit you revert becomes tag v2.14.1, the image the cluster runs, and the service the new alarm watches on Friday's own numbers.

        Velomint

        Urban mobility

        About the company

        An app for renting e-bikes by the minute in Lisbon and Porto.

        Riders find a Velomint e-bike in the app, unlock it with their phone and pay by the minute, with a small card hold at each unlock. Commuters ride most on weekday evenings. The app talks to rides-api, a Node.js service that handles unlocks, rides and payments; it runs as a Kubernetes Deployment behind an AWS load balancer, and alerts go to an SNS topic the on-call engineer subscribes to. Releases go out from the main branch of one shared Git repository. You work in the platform team, which owns the cluster, the releases and the alarms.

        Founded
        2021
        Head office
        Lisbon, Portugal
        Cities
        Lisbon and Porto
        Employees
        About 60

        Who writes the briefs

        • Tomás PereiraHead of Engineering
        • Sofia RamosPlatform Engineer, on call on Friday

        A fictional company, invented for practice. Meet the other companies

        The brief

        From
        Tomás PereiraHead of Engineering
        To
        Platform team
        Subject
        Friday's unlock failures

        Release 2.14.0 of rides-api went out at 16:02 on Friday. For the next 45 minutes about a third of unlocks failed with a 502, until Sofia rolled back to 2.13.2. We heard about it from riders, not from an alarm.

        Before we ship weekend pricing again: find the commit that broke payments and undo it without rewriting main, deploy the fixed release so a broken build can never take traffic, and set up the alarm that would have paged us. Then write the incident report for the whole team.

        You have the load balancer's requests and 5xx counts per five minutes for Friday afternoon, and the timeline Sofia put together.

        Done looks like: v2.14.1 tagged on main, running on 3 ready pods behind a readiness probe, an alarm on the 5xx count that pages on-call, and a short blameless incident report.

        • Data: Friday 6 November 2026, 15:30 to 18:30, in five-minute buckets, plus the incident timeline.
        • You deliver a revert and a tag, a safe Kubernetes rollout, a CloudWatch alarm and an incident report.
        • Checks cover shared history, the ConfigMap key the pods read, rollout safety and the alarm rule.

        The project path

        1. 1
          Git25 min

          Revert the commit that broke payments

          Undo the bad commit with a new one, keep weekend pricing, and tag v2.14.1.

          Reads
          Produces

          The cluster deploys the image built from v2.14.1.

          Open stageOne Course in Git or All Access
        2. 2
          Docker and Kubernetes35 min

          Ship 2.14.1 behind a readiness probe

          Fix the manifest's ConfigMap key, add a readiness probe and a rollout that never drops a serving pod.

          Reads
          Produces
          release

          AWS watches the load balancer in front of these pods.

          Open stageOne Course in Docker and Kubernetes or All Access
        3. 3
          AWS25 min

          Add the alarm Friday did not have

          A CloudWatch alarm on the targets' 5xx count that pages on-call, checked on Friday's numbers.

          Reads
          error_rate
          Produces

          Your incident report cites when it would have fired.

          Open stageOne Course in AWS or All Access
        4. 4
          Readout20 min

          Write the incident report

          A short blameless report: what happened, why, what you fixed and how the next one gets caught.

          Reads
          incident_log · release · error_rate
          Produces
          incident_2026-11-06.md

          Optional: present it in Interview Hub.

          Open stageAll Access

        Each stage includes reference inputs. Starting in the middle or using them doesn't complete the earlier stages.

        Project access

        Free

        $0

        Read all briefs and paths. Tool stages and written work are locked.

        Keep browsing

        One Course

        $17.99/month

        Practice the stages in your chosen tool. Reference tables let you start at that stage.

        All Access

        $29.99/month

        Every stage in every format, plus optional spoken readouts using dumbbells.

        Your Work Gym account

        Save your progress and pick up on another device. Your browser's progress comes with you.

        or

        You can keep practicing without an account.

        Help and feedback

        Ask a question, tell us what is not working, or share an idea. We reply by email.

        What is it about?

        We use it only to reply to you.