FreeExam

Internal platforms

A platform engineer interview on building internal developer platforms that other teams choose to use, from golden paths and Kubernetes basics to migrations and cost visibility.

These prompts are original practice. They are not questions from a named employer or a live hiring loop.

  1. 1.At a bike-share company, the platform team built a new deployment tool. Six months later, only two of thirty service teams use it. What should the platform team do first?

  2. 2.What does it mean to treat an internal platform as a product?

  3. 3.A platform team at a payroll service has 60 open requests from product teams. How should it decide what to build next?

  4. 4.A platform team at a county library system plans a large new feature. Which step best reduces the risk of building something nobody uses?

  5. 5.What is the main risk when a platform team measures its success only by the number of features it ships?

  6. 6.A product team at a grain co-op says the platform's deploy tool is confusing. The platform team believes it is well designed. What is the most useful next step?

  7. 7.Why do many platform teams start with a thin platform, such as a documented set of existing tools and conventions, rather than a large custom system?

  8. 8.At a ferry booking company, the largest internal customer asks the platform team for a custom feature that no other team needs. What is a reasonable response?

  9. 9.What is a golden path in platform engineering?

  10. 10.At a pet insurance company, the platform team created a service template. A year later, services made from it have drifted and miss security fixes. What design would have reduced this?

  11. 11.A team at a recipe site needs a streaming workload that the golden path does not support. What is the best platform stance?

  12. 12.Which signal best shows that a golden path is working?

  13. 13.A new hire at a solar installer takes two weeks to get a simple service into production. Which platform investment most directly targets this?

  14. 14.A golden path template at a parking app includes a logging library, health checks, and a pipeline. A team asks why it cannot pick its own logging library. What is the best explanation?

  15. 15.When a platform team updates a golden path template, how should it handle services already created from older versions?

  16. 16.A platform team at a dental chain finds that 80 percent of services are simple HTTP APIs, and the rest vary widely. Where should the first golden path focus?

  17. 17.At a ticket resale company, developers wait three days on a ticket to get a new database. What is the most effective platform fix?

  18. 18.What is the main benefit of self-service over ticket-based requests?

  19. 19.Security at a credit union worries that self-service cloud accounts will let developers create costly or insecure resources. What approach keeps self-service while handling this?

  20. 20.A platform team is choosing which ticket types to automate first. Which is the best candidate?

  21. 21.After a museum's platform team launched a self-service DNS tool, the number of DNS change tickets did not drop. What should the team check first?

  22. 22.Which is a good reason to keep a human approval step for some platform requests?

  23. 23.A platform team offers a web portal, a command-line tool, and an API that all create new services. The three often behave differently. What is the best fix?

  24. 24.During an incident at a food delivery company, responders spend 40 minutes finding who owns a failing service. What platform capability addresses this most directly?

  25. 25.Service catalog entries at a school district are often out of date. What makes ownership data most likely to stay accurate?

  26. 26.Why should a service be owned by a team rather than a named individual?

  27. 27.After a reorganization at a newspaper, the catalog shows 30 services with no owning team. What should the platform team do?

  28. 28.Besides the owner, which service catalog field is most useful during an incident?

  29. 29.A platform team at a travel agency adds scorecards to its catalog that rate services on things like having an on-call rotation and current dependencies. What is the best way to use them?

  30. 30.A team at a theater ticketing site plans to shut down an old internal API. Which catalog data helps most when planning this?

  31. 31.In Kubernetes, what does a container's CPU request mainly affect?

  32. 32.A pod at a hotel booking site keeps restarting with an out of memory kill. Its memory limit is 256 MiB, and the app uses about 300 MiB at peak. What is the most likely fix?

  33. 33.A latency-sensitive API at a ride-share company has low average CPU use, but its 99th percentile latency spikes. Metrics show heavy CPU throttling. What is the most likely cause?

  34. 34.What is a Kubernetes namespace mainly used for on a shared cluster?

  35. 35.A team at a tax prep firm sets a horizontal pod autoscaler to target 70 percent CPU, but its pods have no CPU requests. What happens?

  36. 36.At a ski resort app, the autoscaler adds pods during a traffic spike, but the new pods stay in the Pending state. What is the most likely cause?

  37. 37.What happens when a container tries to use more memory than its memory limit?

  38. 38.A platform team at a car rental company wants to stop one team from using all of a shared cluster's memory. Which Kubernetes feature fits best?

  39. 39.Many developers at a gym chain forget to set resource requests on their pods. What platform approach helps most?

  40. 40.Nodes at a moving company's cluster are 90 percent requested but only 20 percent used. What is the most likely cause?

  41. 41.What is the difference between a horizontal pod autoscaler and a cluster autoscaler?

  42. 42.A critical service sets its requests equal to its limits for both CPU and memory. Why might a platform team recommend this?

  43. 43.On a shared cluster at a podcast host, one team's nightly batch job slows other teams' APIs. What is the best first platform step?

  44. 44.What is a noisy neighbor in a multi-tenant platform?

  45. 45.A wedding planning app runs all teams in one cluster. A new team will process sensitive health data under strict legal rules. What is a reasonable platform choice?

  46. 46.Many teams at a farm market platform share one Postgres instance. One team's reporting queries cause timeouts for everyone. What fix best addresses the root issue?

  47. 47.On a shared cluster at a translation service, any pod can reach any other team's pods. What should the platform add?

  48. 48.A shared cluster runs customer-facing services and best-effort batch jobs. During a capacity crunch, the platform wants customer services to win. What helps?

  49. 49.The internal artifact service at a seed bank slows down when one team's pipeline downloads the same files thousands of times an hour. What is the best fix?

  50. 50.A continuous integration (CI) pipeline at a toy store takes 25 minutes, and 15 of those are spent downloading dependencies on every run. What is the most effective fix?

  51. 51.A Dockerfile at a flower shop copies the whole source tree before installing dependencies. Builds rarely reuse cached layers. What change helps most?

  52. 52.Why should a dependency cache key include a hash of the lockfile rather than only the branch name?

  53. 53.At a pharmacy chain, one in five pipeline runs fails on tests that pass on retry. Teams now rerun until the build is green. What should the platform team push for?

  54. 54.A monorepo at a game studio runs every test for every change, which takes 40 minutes. Which approach cuts this most while keeping it safe?

  55. 55.A vet clinic software company adds a shared remote build cache. Soon, some builds produce wrong outputs in CI that do not appear locally. What is the most likely cause?

  56. 56.The platform team at a laundromat chain wants faster pipelines for all teams. What should it do first?

  57. 57.A city water utility's test suite has 2,000 independent tests that run one after another in 30 minutes. What is a direct way to cut the wall-clock time?

  58. 58.What is a preview environment?

  59. 59.Preview environments at a bakery chain cost more each month, and many keep running for weeks. What is the best fix?

  60. 60.At a freight broker, ten teams share one staging environment. Teams often break it for each other and wait days to test. What change helps most?

  61. 61.Preview environments for a clinic scheduler need data to be useful. What is the safest data source?

  62. 62.A bug at a newspaper shows up only in production. Staging runs a different Postgres major version than production. What does this show?

  63. 63.Why should the same build artifact be promoted from staging to production instead of being rebuilt?

  64. 64.Preview environments at a moving company take 30 minutes to start, so developers stopped using them. What should the platform team look at first?

  65. 65.A team at a credit union stores a database password in a config file in its repo. What should the platform offer?

  66. 66.What is a common misunderstanding about Kubernetes Secret objects?

  67. 67.After an API key leaked at a parking app, the team had to rotate it, which took a day of coordination. What design makes future rotation easy?

  68. 68.Why do platform teams prefer short-lived, automatically issued credentials for services over long-lived static keys?

  69. 69.A config change at a grain co-op set a timeout to 0 and took down a service. Config is pushed instantly to all instances. What would most reduce this risk?

  70. 70.Which value belongs in a secrets store rather than in plain config?

  71. 71.A team at a translation service finds secrets printed in application logs at startup. What is the right first step?

  72. 72.Why keep environment-specific config out of the build artifact?

  73. 73.What is the difference between a guardrail and a gate in platform engineering?

  74. 74.A change review board at a dental chain meets weekly to approve every production deploy. Lead time is two weeks, yet incidents have not dropped. What should the platform team propose?

  75. 75.A platform team at a hotel group wants to stop public storage buckets. Which approach prevents the problem without adding a human wait?

  76. 76.The platform team at a toy store will start rejecting deploys without resource limits. What is the best rollout?

  77. 77.A new platform policy at a solar installer blocks a valid use case for one team. What is the healthiest way to handle it?

  78. 78.Where is the best place for a guardrail to report a problem to a developer?

  79. 79.At a pharmacy chain, the platform team wants only images built by the trusted pipeline to run in production. Which guardrail fits?

  80. 80.A platform team at a county library is moving 50 services from an old deploy system to a new one. Which migration order usually works best?

  81. 81.Eighty percent of teams at a bike-share company have moved to the new platform, but the last 20 percent are stuck. What usually helps most?

  82. 82.Why run the old and new platforms in parallel during a migration?

  83. 83.The platform team at a gym chain will retire an internal message queue. Which plan is best?

  84. 84.Before shutting down an old internal API at a ride-share company, how can the team be confident nobody still uses it?

  85. 85.What is a brownout when deprecating an internal service?

  86. 86.A migration to a new logging library at a seed bank needs changes in 300 repositories. What lowers the cost most?

  87. 87.Why might a platform team stop adding features to the old platform during a migration?

  88. 88.Which is the best measure of a platform migration's progress?

  89. 89.Which set of metrics best reflects how well software delivery is going across the teams a platform serves?

  90. 90.A platform team reports that 95 percent of services use its new pipeline. Developers still complain about slow delivery. What is missing?

  91. 91.A platform team at a theater ticketing site runs a developer survey. Which question gives the most actionable input?

  92. 92.Adoption of an internal tool jumped to 100 percent right after leadership made it mandatory. What does this number tell you?

  93. 93.Which metric best captures how easy it is for a new engineer to become productive on the platform?

  94. 94.Leadership sets a goal to raise deploy frequency. Some teams start splitting trivial changes into many deploys. What is the problem?

  95. 95.Pipeline data at a freight broker shows builds are fast, but surveys say developers find the build slow. What should the platform team do?

  96. 96.A bad change to a shared base image at a podcast host broke every service that rebuilt that day. What practice most limits this blast radius?

  97. 97.What does blast radius mean in platform engineering?

  98. 98.A car rental company runs every team's services on one large cluster. A failed control plane upgrade took everything down. What structural change helps most?

  99. 99.The platform on-call at a wedding planning app is paged for every app bug in every team's service. What should change?

  100. 100.During an outage of the internal deploy system at a bakery chain, what should the platform on-call do early?

  101. 101.A city water utility's deploy system runs on the same cluster it deploys to. During a cluster outage, teams cannot deploy fixes. What is the core issue?

  102. 102.How should a platform team set reliability targets for its internal services?

  103. 103.Cloud costs at a pet insurance company rose 40 percent, and nobody knows which team drove it. What should the platform team set up first?

  104. 104.How can a platform team fairly show each team its share of a shared Kubernetes cluster's cost?

  105. 105.What is the difference between showback and chargeback?

  106. 106.A cost report at a laundromat chain shows many test databases running all weekend with no traffic. What is a good platform fix?

  107. 107.When is cost information most likely to change a developer's design decisions?

  108. 108.A team's cloud cost at a food delivery company doubled, but so did its traffic. Which metric shows whether efficiency changed?

  109. 109.A platform team at a school district wants to cut cluster cost without hurting services. What is a safe first move?