Request Timeout Cascade

A dependency at the bottom of a four-hop call chain (Gateway, Service A, Service B, and the Dependency itself) has developed a latency blip — it is slow, not failing outright, and customers depend on the capability it provides. The Gateway is configured to retry a request when its own call times out. No request chain is being explicitly cancelled anywhere along the way when a caller gives up on it. In-flight request counts and connection utilization are climbing at every layer, and customer-facing latency is degrading with them. Diagnose where in the chain pressure is actually accumulating — and why it keeps accumulating even after a fix lands on one hop — before connection capacity runs out anywhere in the chain.

Difficulty

Staff

Expected Time

~20 min

Category

API & Traffic