Topic 16.4
Senior and Staff Questions (51–63)
In one line
Thirteen probing questions about what really happens in failures and at scale: acknowledged-write loss, lagging replicas, hot and huge keys, CPU and latency incidents, memory growth, AOF growth, restart loss, double lock holders, safer locks, Redis vs consensus, and when not to use Redis.
Think of it like this
A senior pilot's check ride. The examiner doesn't ask how to take off; they fail an engine at the worst moment and watch how you reason. Staff-level Redis questions are failure drills.
Key ideas
- 01
Answer shape: state precisely what happens (mechanism), what's lost or wrong, how you'd detect it, and what design or configuration changes prevent it. Avoid absolutes; say which conditions matter.
Explain it without notes
51. The Redis primary fails immediately after acknowledging a write. What happens?
52. A replica is 2 seconds behind. Should the application read from it?
53. Redis Cluster has one extremely hot key. How do you solve it?
54. One Redis key contains 50 GB. What do you do?
55. Redis CPU is 100%. How do you debug it?
56. Redis latency jumps from 1 ms to 500 ms. Build your diagnosis.
57. Redis memory keeps increasing. How do you find the cause?
58. The AOF becomes huge. What happens?
59. Redis restarts. How much data can be lost?
60. Two services believe they hold the same lock. How did it happen?
61. How do you make a distributed lock safer?
62. Why isn't a Redis lock equivalent to a consensus system?
63. When should Redis NOT be used?
Practice
Answer questions 51, 60 and 63 in writing, then check which conditions and mechanisms you left out.
Done when you can
I can answer all 13 staff-level questions with mechanisms, conditions, detection and prevention.