What problem does it solve?
Optimizing rsyslog without a disciplined method leads to noisy benchmarks, unsafe changes, and regressions in durability or concurrency. This Skill enforces an evidence-driven workflow so performance changes to inputs, parsers, queues, and outputs are measured, validated, and reversible.
Core Features & Use Cases
- Controlled benchmarking: Build isolated baseline and candidate worktrees, run calibration plus at least eleven paired trials, and compare medians with median absolute deviation to resist host noise.
- Safety-first change discipline: Define invariants for security, durability, memory, ownership, and concurrency before coding, and stop at an evidence-backed proposal before weakening any of them.
- Risk-matched validation: Route changes through focused tests, sanitizers (ASAN/UBSAN for buffers, TSAN for locking and queues), static analysis, and container validation.
- Use Case: When optimizing the imtcp input path or a disk queue drain routine, use this Skill to profile the path, implement one candidate at a time, and retain it only if two independent sessions meet the declared thresholds.
Quick Start
Use the rsyslog-performance skill to plan and validate a safe optimization for the rsyslog queue subsystem with representative workloads and paired benchmarks.
