CloudWatch shows ReadLatency on an RDS for MySQL instance climbing from about 2 ms to 40 ms while CPU stays moderate. Queries that use indexes are also getting slower. Which cause should be investigated first?
Show answer and explanation
Answer: A
ReadLatency measures the time the engine waits for read requests against the volume, so a sustained rise points at storage: exhausted burst balance, an IOPS or throughput ceiling, or a queue of pending I/Os. Comparing the volume's achieved IOPS and throughput with its configured limits confirms it, and moving to gp3 with provisioned IOPS or to io2 raises the ceiling.
Why the other options are wrong
- B. Cross-AZ traffic adds network round-trip time to query responses, but it does not inflate the engine's storage read latency metric.
- C. Connection pressure appears in DatabaseConnections. It can slow the instance through context switching but does not specifically raise read latency against the volume.
- D. Memory pressure shows in FreeableMemory and swap. It usually affects both reads and writes and would not leave CPU moderate with a pure read-latency spike.
Exam tip. ReadLatency is a storage metric. Match it to volume IOPS and throughput.