PipeBench 0.4.4

PipeBench 0.4.4

Cross-compiled harness CLI for orchestrating data-pipeline benchmark runs.
Bench-* container images for the same revision live on Docker Hub:

• vmetric/bench-generator:sha-ddba337
• vmetric/bench-receiver:sha-ddba337
• vmetric/bench-collector:sha-ddba337
• vmetric/bench-verifier:sha-ddba337

Install

Download the archive that matches your OS/arch below, extract, and put
harness on your PATH. Verify the checksum against checksums.txt.

What's changed

Pull requests in this release

The bench cases in virtualmetric-bench depend on case keys that v0.4.3 either rejected (the $TODAY seed-prefix token) or silently ignored (forbidden_substring, delay_seconds, max_received, subject_entrypoint, subject_disk). These three PRs add them:

• #58 feat(harness): S3 seed/verdict knobs for windowed polls
• #55 feat(harness): add subject entrypoint overrides
• #48 Add subject_disk tmpfs mounts and disk_pressure_correctness cases

Also included:

• #56 feat(runner): add resource limits and assertions
• #59 feat(runner): derive the expected total for database cases
• #60, #61 chore(deps): bump google.golang.org/grpc to 1.83.1 in the generator and receiver containers

Changelog

• 8d72dab: Add subject_disk tmpfs mounts and disk_pressure_correctness cases (@namles)
• 3d07568: Address review: shared byte-size parser, type guards, du without a shell (@namles)
• c85e74d: Apply the shared correctness checks in the disk-pressure driver (@namles)
• 66e0548: Harvest resource metrics in the disk-pressure driver (@namles)
• 243e8d3: chore(deps): bump google.golang.org/grpc in /containers/generator (@dependabot[bot])
• adf68a7: chore(deps): bump google.golang.org/grpc in /containers/receiver (@dependabot[bot])
• 7560fa5: feat(harness): S3 seed/verdict knobs for windowed and poison-object poll (@erenaslandev)
• c065c97: feat(harness): add subject entrypoint overrides (@erenaslandev)
• 0f48b18: feat(runner): add resource limits and assertions (@erenaslandev)
• 120a4e0: feat(runner): derive the expected total for database cases (@erenaslandev)
• 9267ef8: feat(runner): stamp resource limits in saveResult (@erenaslandev)
• 51ab8a0: fix(harness): review follow-ups for seed/verdict knobs (@erenaslandev)
• 8a776dc: fix(runner): clamp metric collection to run budget (@erenaslandev)
• 501c510: fix(runner): count loss over unique lines in mid-delivery cases (@erenaslandev)
• b1a793c: fix(runner): enforce expect_max ceiling for endpoint sources (@erenaslandev)
• c454bda: fix(runner): improve endpoint source ceiling logic (@erenaslandev)
• e98302d: fix(runner): one predicate for unique tracking, and a log-anchored crash (@erenaslandev)
• c488dfe: fix(runner): restart the subject without touching its dependencies (@erenaslandev)
• dbbc8a3: fix(runner): use persisted result for verdict in runner tests (@erenaslandev)
• 4c04f9e: fix(runner): use pre-resolved limits and return errors (@erenaslandev)
• 9aca132: fix(runner): use the database total in the mid-delivery verdict (@erenaslandev)

---

Full Changelog: v0.4.3...v0.4.4