Question 360
After an incident where bigtable read errors increase for one table while node CPU is not high, which recurring control would best prevent the same defect?
Explanation
Correct Answer
Review row-key query shapes and timeout budgets before releasing new analytics access. It targets the actual operating weakness: The client is requesting an inefficient scan or malformed key range and timing out. The immediate repair was: Trace read RPCs and key-range access patterns then use supported bounded row-key queries.
Incorrect Answers
For the situation in which Spanner latency rises for one tenant while aggregate database CPU appears healthy, it makes sense to examine skew and hot key ranges in multitenant Spanner workloads. Here, the client is requesting an inefficient scan or malformed key range and timing out, so the safeguard misses the recurring fault.
The failure in the question arises because the client is requesting an inefficient scan or malformed key range and timing out. A safeguard to review UTC versus local reporting boundaries for all scheduled database jobs serves situations where a Cloud SQL scheduled export runs at midnight UTC but the reporting team expects the prior local business day rather than this one.
For the situation in which AlloyDB reports elevated CPU shortly after an index is added and application writes become slower, it makes sense to assess read benefit versus write cost during every index review. Here, the client is requesting an inefficient scan or malformed key range and timing out, so the safeguard misses the recurring fault.