Distributed Tracing

1. The Incident

The support ticket says that checkout is sometimes slow. Its p99, the threshold exceeded by the slowest 1% of requests, is 4.2 seconds. The flow has an 800 ms budget, although only some requests are affected, at unpredictable times, without an obvious pattern.

A checkout request reaches twelve services through more than forty calls. These include the gateway, BFF, checkout, inventory, payment, fraud, and pricing services. BFF means the backend-for-frontend layer serving the mobile application