ExChain: Exception Dependency Analysis for Root Cause Diagnosis
Talkat 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24) · 2024-04-16..04-18·
summary
ExChain is an online monitor that instruments services with low overhead and links silent, state-changing exceptions to distant failures using affected/responsible state analysis and a hybrid taint tracking approach. It claims to find root causes missed by logs and tracing, identifies the true culprit in 8 of 11 reproduced real-world exception-dependent failures, runs at about 8% overhead (static-only 6%, dynamic tracing >1000%), and outperforms statistical debugging and Slicer4J.
AI-written summary · about
chapters (10)
- 0:00Start
- 0:17Exception handling failures
- 2:22Exception dependency analysis
- 5:15ExChain design goals
- 7:04Affected state analysis
- 10:00Exception state propagation
- 12:13Hybrid taint analysis
- 13:16Key ideas of ExChain
- 13:39Evaluation on real failures
- 16:31Conclusion
related talks
02h 20mSuperchain Chats
38mChaos Engineering
23mAutomatically Detecting Error Handling Bugs Using Error Specifications
29mComplexity: The Crucial Ingredient in Your Kitchen
21mCorrectness and Performance for Stateful Chained Network Functions
34mAntics, Drift, and Chaos
27mArchitecting for Interconnected Global Ecosystems
49mFunctional Error and Optional-value Handling with STX