• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

kubernetes-sigs / inference-perf / 43485826608
87%

Build:
DEFAULT BRANCH: main
Ran 13 Nov 2026 06:59PM UTC
Jobs 1
Files 142
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

13 Nov 2026 06:46PM UTC coverage: 86.498% (+0.001%) from 86.497%
43485826608

push

github

web-flow
Keep teardown-cancelled requests out of the throughput window (#1081)

Part of #1033

**Old behavior:** since #859, a request cancelled at stage teardown is
its own outcome, `cancelled`, and is kept out of successes, failures and
every latency distribution. The throughput window still spans it,
though. `benchmark_time_seconds` runs from the first send to the last
`end_time` over all requests, and a cancelled request's `end_time` is
when it was cancelled: up to `stage_teardown_grace_seconds` (120 s by
default) after the stage ended. A single straggler stretches the window
by up to the grace and lowers every `*_per_sec` figure, although nothing
was being served in that time. The BR v0.2 partial's throughput has the
same window.

**New behavior:** the window spans the answered requests (successes and
failures). If every request was cancelled, it falls back to all of them
so that it stays positive. Reports with no cancelled requests are
unchanged.

Changes:
- `reportgen/base.py`: `summarize_requests` computes `total_time` over
successes and failures.
- `reportgen/br/v0_2/adapter.py`: the partial's throughput uses the same
window.
- `tests/required/reportgen/test_failure_reporting.py`: two successes
within 1 s plus one request cancelled 120 s after it was sent, which
expects `benchmark_time_seconds == 1.0` and `requests_per_sec == 2.0`.
Also the all-cancelled fallback.
- `tests/reportgen/br/v0_2/test_partial_report.py`: the same scenario
for the partial's `request_rate`.

**Fails on main:**
```
FAILED tests/required/reportgen/test_failure_reporting.py::TestRequestFailuresByLabel::test_stage_teardown_cancellations_do_not_stretch_the_throughput_window
FAILED tests/reportgen/br/v0_2/test_partial_report.py::test_partial_throughput_window_excludes_a_teardown_cancel
```
On main the window is 120.9 s instead of 1 s (0.017 req/s instead of 2).

**Testing** (Linux, Python 3.13, rebased on `68730a4`):
- `pdm run validate` passes: format, license headers, lint, `mypy... (continued)

3 of 3 new or added lines in 2 files covered. (100.0%)

13838 of 15998 relevant lines covered (86.5%)

0.86 hits per line

Jobs
ID Job ID Ran Files Coverage
1 43485826608.1 13 Nov 2026 06:59PM UTC 142
86.5
GitHub Action Run
Source Files on build 43485826608
  • Tree
  • List 142
  • Changed 2
  • Source Changed 2
  • Coverage Changed 2
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #43485826608
  • 4e3bcc6f on github
  • Prev Build on main (#43355313171)
  • Next Build on main (#43930557778)
  • Delete
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc