Improving Bpftrace Reliability
summary
bpftrace reliability work focuses on concrete fixes: reproducible CI built with NYX flake to ship identical builds and AppImage-style static bundles, a VM test harness that maps the host userland for real kernel testing, and runtime/unit test suites plus code K queries to catch serialization and IR regressions. Ongoing engineering includes off‑stack big strings via per‑CPU maps, reporting of dropped probes, and prototypes to detect and handle partially inlined functions so tracing doesn’t silently miss call sites.
AI-written summary · about
chapters (8)
- 0:00Introduction to bpftrace
- 1:09Defining reliability
- 2:57Challenges and holistic approach
- 8:14Investing in CI
- 10:06Nix and distribution
- 13:43Testing techniques
- 20:52Current reliability focus
- 25:34Q&A and conclusion
related talks
46mbpftrace internals
35mbpftrace: Evolution from a command line tool to a complex observability framework
25mKernel tracing with BPF
35mPerformance Insights into eBPF Step by Step
21mPerformance Wins with BPF: Getting Started
36mObservability and performance analysis with BPF
32mBPF: Introduction, Programming Tips and Tricks
38meBPF for Observability: Hard Truths and Practical Wins