Un-breaking production bugs with non-breaking breakpoints
summary
Explains a concrete system that debugs live services without pausing them: non-breaking breakpoints are injected via runtime bytecode edits that call Apache SkyWalking’s manual tracing, capturing full variable context, logs and counters and streaming hits back to an IDE. Implementation uses SourceMarker plus the new Source++ live-coding plugin with JetBrains IDEs, limits scope and redacts secrets, adds live logs and metrics, installs instantly to production with negligible overhead, and is released as open source.
AI-written summary · about
chapters (13)
- 0:00Introduction to talk
- 1:10Production vs local debugging
- 3:55Blind vs heuristic search
- 8:32Limitations of logging
- 9:57Non-breaking breakpoints explained
- 12:01Source Marker framework
- 15:08Introducing Source++
- 16:52Live coding platform
- 18:39How live breakpoints work
- 20:01Live control bar demo
- 21:49Live logs and metrics
- 23:16Live metrics in action
- 25:22Future of live coding
related talks
45mDebugging Production: No Re-deploys, No Downtime, No Risk
25mDebugging at Scale in Production with kubectl debug
52mApplication-Level Debugging and Profiling: Gaps in the Tool Ecosystem
34mChaos Debugging: Finding and Fixing Microservice Weak Points
15mDebugging debugging; why we must approach debugging differently
30mCloud Native Debugging in Production - Dig Deep into your agents
31mConsole.loggers Anonymous: A Debugging Roadmap from Local to Production
17mHow to crash your app in production and still delight your users