Load Testing Middle interview questions and answers

76 questions for a Load Testing (Middle) interview - with correct answers, explanations and links to the docs.

Take this quiz
  1. 1.What is the key difference between a closed and an open workload model?
  2. 2.Why can a closed model hide a slowdown?
  3. 3.Which k6 executor keeps a fixed request rate regardless of response times?
  4. 4.What load does this scenario generate?
  5. 5.What does a non-zero dropped_iterations metric tell you in an arrival-rate test?
  6. 6.You need 100 RPS, and one iteration takes about 1.5 s. Roughly how many VUs must be available?
  7. 7.What is coordinated omission?
  8. 8.What is the difference between shared-iterations and per-vu-iterations?
  9. 9.What does startTime do in a k6 scenario?
  10. 10.What does exec set in a k6 scenario?
  11. 11.What does this threshold do?
  12. 12.How do you set a threshold for only one endpoint?
  13. 13.Which custom k6 metric type fits measuring a duration with percentiles?
  14. 14.Why load a big CSV of users with SharedArray?
  15. 15.Where in a k6 script can you call open() to read a file?
  16. 16.What does http.batch() do?
  17. 17.A login endpoint should return 401 for wrong passwords in your test. How do you keep it out of http_req_failed?
  18. 18.Why set discardResponseBodies: true in a large k6 test?
  19. 19.How does a protocol-level k6 test differ from a browser-level test?
  20. 20.When does a hybrid test with a few browser VUs and many protocol VUs make sense?
  21. 21.Two generators report p95 = 200 ms and p95 = 400 ms. Is the overall p95 300 ms?
  22. 22.A test sent 200 requests and reports p99.9. How reliable is that value?
  23. 23.What does the USE method check for each resource?
  24. 24.Why does latency grow sharply as a resource approaches 100% utilization?
  25. 25.What does Amdahl's law say about scaling?
  26. 26.Under load, timeouts trigger client retries, and traffic doubles. What is this called?
  27. 27.What is load shedding?
  28. 28.During a spike test, autoscaling adds servers, but errors happen in the first minutes. Why?
  29. 29.Why run a test both with a cold and a warm cache?
  30. 30.The test hits the site through a CDN. What do you measure for static assets?
  31. 31.In a soak test, logins start failing after 1 hour. What is a typical test-side cause?
  32. 32.What does k6 do with HTTP connections by default?
  33. 33.All load comes from one IP address through a load balancer with sticky sessions. What can go wrong?
  34. 34.Which JMeter element helps keep a target throughput instead of a fixed number of threads?
  35. 35.What is the CSV Data Set Config in JMeter for?
  36. 36.How do you pass a token from a login response to the next request in JMeter?
  37. 37.In Gatling, what is the difference between constantUsersPerSec and constantConcurrentUsers?
  38. 38.What is a feeder in Gatling?
  39. 39.In Locust, what does wait_time = between(1, 3) do?
  40. 40.In a soak test, memory grows steadily and GC pauses get longer. What does it suggest?
  41. 41.The new build's p95 is 8% higher than the baseline. What do you check first?
  42. 42.How does distributed tracing help during a load test?
  43. 43.What does an Apdex score express?
  44. 44.Why tie load test targets to the SLO and error budget?
  45. 45.Under load, requests wait for a free worker thread while CPU stays low. What is saturated?
  46. 46.Why is it hard to load test an asynchronous flow (API accepts, a worker processes later)?
  47. 47.A soak test creates 1 million orders. Why can this distort later runs?
  48. 48.What does pacing mean in a load test?
  49. 49.What happens to concurrency if you cut think time from 5 s to 1 s at the same RPS?
  50. 50.Why may a load test against a service hit its own rate limiter first?
  51. 51.The generator logs "cannot assign requested address" at high load. What is the likely cause?
  52. 52.Why is TLS handshake cost often noticeable in load tests without connection reuse?
  53. 53.What changes when a test uses HTTP/2 instead of HTTP/1.1?
  54. 54.The average latency is 200 ms, but the histogram shows two peaks: 50 ms and 800 ms. What does it suggest?
  55. 55.Why is standard deviation a weak summary for latency?
  56. 56.The system calls a payment provider. Should the load test hit the real provider?
  57. 57.Does a k6 HTTP test download images and CSS referenced by an HTML page?
  58. 58.How do you make two runs of the same test comparable?
  59. 59.In JMeter distributed mode, what does the controller do?
  60. 60.What does a Synchronizing Timer in JMeter do?
  61. 61.How can k6 results be shown in Grafana dashboards in real time?
  62. 62.What does handleSummary() in k6 let you do?
  63. 63.Why run a test at 120-150% of the expected peak?
  64. 64.During a test of 10 minutes, one endpoint's p95 is fine, but errors cluster in one minute. What do you do?
  65. 65.When testing WebSockets, what should you measure besides connection time?
  66. 66.How do you test a cache stampede scenario?
  67. 67.Why are load tests often written as code (k6, Gatling, Locust) rather than recorded?
  68. 68.What is a bad sign in a stress test even if the system recovers after load drops?
  69. 69.What is the main risk of running k6 with thousands of VUs on one laptop?
  70. 70.Why should load test reports include the configuration, not just numbers?
  71. 71.What does a rising http_req_blocked in k6 indicate?
  72. 72.A test of a search endpoint uses the same 10 queries. What is the risk?
  73. 73.What does a load test of the database layer alone not tell you?
  74. 74.Which phases of a request does k6 include in http_req_duration?
  75. 75.Why is a test that only hits GET /products a weak capacity test for a shop?
  76. 76.What is the difference between ramping-vus and ramping-arrival-rate?