- This topic is empty.
-
Every weekday around 3 PM, our order-processing service starts taking 8-12 seconds to respond instead of the usual 200 milliseconds. The dashboards show high CPU, but only on two application nodes. We first blamed a traffic surge, yet request volume stayed flat. The real clue came from database metrics: connection wait time spiked at exactly 3:02 PM. A nightly reporting job was grabbing connection pool slots and holding them longer than expected. Once we moved the report to a replica and capped its pool usage, the slowdown disappeared. This saga taught us to look beyond CPU and memory. Latency often hides in queues, locks, and shared resources. If you have chased a recurring slowdown, what unexpected bottleneck did you find, and how did you prove it was the cause?
- You must be logged in to reply to this topic.