Practicing Linux Crash/Panic Issue on Production and Cloud Server: Using K...
summary
Describes a hands‑on workflow that deploys kdump + crash to capture vmcores from production and cloud servers, reproduces six focused experiments (a null pointer panic, accessing a deleted linked list, a driver crash, extracting local variables from call stacks, a multi‑thread deadlock, and manual call‑stack recovery), and shows running analysis on x86_64 and arm64 stack layout. Demonstrates locating exact parameter values, identifying lock holders in deadlocks, and resolving hugepage/IPI migration hangs; the tools and steps are validated on real servers.
AI-written summary · about
chapters (8)
- 0:00Start
- 16:51企业级虚拟化计算节点服务器性能问题 实验目的
- 18:16Case Description
- 20:20Investigation on the VM side
- 22:15Host Machine - Hung task Backtrace
- 29:36KSM merge list extraction
- 30:30Introduction to the KSM Stable Tree (Stable/Unstable tree)
- 34:34FlameGraph of the performance problem
related talks
26mStateless Linux Kernel Crashdumps
22mScaling Kernel Production Monitoring @ Meta
27mWhen kdump is way too much
20mProactive and crash-time data collection on Steam Deck
34mChaos Debugging: Finding and Fixing Microservice Weak Points
57mTroubleshooting Node.js Applications Under the Hood
50mPost-Mortem Debugging in Node.js
43mProd Detectives: Solving the Case Before the Crash