UpFuzz: Detecting Data Format Incompatibility Bugs during Distributed Storage System...
summary
UPFUZZ builds an automated upgrade fuzzer plus a likely-invariant analysis engine that instruments sync I/O and does backward analysis to capture transitively persistent (TP) state before serialization, infers unary/binary/conjunctive properties with temporal and spatial locality, and prioritizes only tests that create new format-relevant states. It finds 15 previously unknown cross-version data-format bugs (8 confirmed, 7 only found by its data-format analysis), with 6 cluster crashes and 4 cases of data loss/corruption, and shows a 47% detection improvement within the same 24-hour budget.
AI-written summary · about
chapters (9)
- 0:00UpFuzz introduction
- 1:21Cross-version migration risks
- 3:09Traditional fuzzing bottlenecks
- 4:33Interesting tests before upgrade
- 6:40Persistent state as input
- 8:04Low-overhead state collection
- 9:27Upgrade-relevant data properties
- 11:14Bugs found and feedback
- 12:37Conclusion and future
related talks
12mUnderstanding and Finding Crash-Consistency Bugs in Parallel File Systems
19mCAFault: Enhance Fault Injection Technique in Practical Distributed Systems via...
13mTCP-Fuzz: Detecting Memory and Semantic Bugs in TCP Stacks with Fuzzing
18mRandomized Testing of Byzantine Fault Tolerant Algorithms
31m“Fuzz or lose..."
25mHow fuzzing helps to find bugs
39mMaking Fuzzing Part of Your Software Development Lifecycle
01h 00mClusterFuzz: Fuzzing at Google Scale