Diving into the Details with DTrace
summary
DTrace proves able to flip on dynamic instrumentation of running processes and time kernel and user calls, so experiments trace and rule out CPU, GC and disk as causes. A small script that does getaddrinfo tracing shows intermittent 30 second DNS resolution for tests hitting someplace.com, reproducing the 30s pauses. The investigator reuses and tweaks those scripts, confirms the root cause and fixes the suite by avoiding or faking external lookups.
AI-written summary · about
chapters (10)
- 0:00Start
- 3:24DTrace is...
- 9:44statically-defined tracing
- 11:17Dynamic tracing
- 17:02Rule out the usual suspect(s)?
- 18:09Garbage collection?
- 19:42System resources
- 23:11Network connections
- 30:51Examples
- 32:11Extrapolate
related talks
11mFinding out what's really going on, with DTrace!
54mUnderstanding Performance w/DTrace (While The Customer Yells)...
30mDTrace at 21: Reflections on Fully-grown Software
36mDebugging Ruby
05mPerformance analysis using Systrace
55mRecent Advances in Linux Tracing
39mLinux Server Deep Dives
34mTrace-Profiling: A New Way About How to Track Application Behavior