Hacker News Reader

Centralized logging, with a track and trace ID that is passed around and logged. I have seen this work to great effect with a dynamic (force directed) based graph been built in real-time showing the flow of calls/messages between systems. Also provided timings and drilldowns to find processing issues and performance problems.

View on HN

People I've worked with kinda-sorta understand the advantages of having a trace ID (aka context ID, correlation ID, etc), but don't code it into their systems. Or they do something like replace an incoming ID with their own, which negates it's value.

For folks who aren't familiar with the concept - at the initiation of any action (user button click, timer firing, automated job starting, etc.) a new ID is created then passed to every single method and external call. This allows you to trace that action as it wends it's way through your distributed systems.

this helps a lot, particularly if your event system spans multiple nodes. its invaluable for correctness, but you can learn a lot about performance too

Reply on news.ycombinator.com