Engineering Decisions · Middle · Question 11 of 14
Checkout's p99 latency doubled. A request passes through six services, and logs don't show which one is slow. What do you reach for?
- 1More detailed logs in each of the six services
- 2Distributed tracing across the request's path
- 3CPU and memory dashboards for all the hosts
Show the answer
Correct answer: Distributed tracing across the request's path
Explanation
A trace shows one request's time in every service, so the slow step is visible at once. More logs fit a failure inside one known service, but matching them across six needs a shared ID and much more work. Host dashboards fit resource problems, while a slow query or a lock can hide behind perfectly normal CPU.