Benchmarks

KSail includes automated benchmark regression testing to detect performance changes in pull requests that modify Go code.

Performance Trends

Loading benchmark data…

Regression Detection

Setting	Value	Meaning
alert-threshold	150%	Marks benchmarks as regressed and posts a PR comment when ≥1.5× slower than baseline (never blocks CI on PRs)
fail-threshold	200%	Fails CI on non-PR runs (pushes to main, merge queue) when a benchmark is ≥2× slower than baseline

On pull requests where benchmarks run and benchmark functions are discovered, a comment is posted only when the alert threshold is exceeded, highlighting the regressed benchmarks.

Benchmark results are also recorded in every CI workflow run summary. On pushes to main, results are auto-pushed to the benchmark-data branch.

Running Benchmarks Locally

# Run all benchmarks
go test -bench=. -benchmem -run='^$' ./...

# Run a specific package
go test -bench=. -benchmem -run='^$' ./pkg/k8s/readiness/...

Writing Effective Benchmarks

• Call b.ReportAllocs() to track allocations
• Use b.ResetTimer() after expensive setup
• Use for b.Loop() style (Go 1.26+)
• Use b.TempDir() for reproducible temp files
• Use table-driven scenarios to cover multiple input sizes
• Fail fast with b.Fatalf on unexpected errors
• Avoid time.Sleep() inside benchmark loops — measure real CPU work, not timers
• For I/O-bound benchmarks (e.g. tarball creation), run a warmup iteration before b.ResetTimer() to prime the OS page cache