26469498063
95%

Ran 26 May 2026 07:18PM UTC

Jobs 1

Files 88

Run time 1min

Badge

Embed ▾

Committed 26 May 2026 07:12PM UTC coverage: 95.243%. Remained the same

Build # 26469498063

Build Type

push

github

Committed by

web-flow

Commit Message

test(kube-e2e): tighten soak error budgets to 0.05% (#507)

PR #505 shipped with `MAX_SIGKILL_SOAK_ERROR_RATE = 5%` explicitly
labelled as a placeholder to tune against baseline runs. A local kind
run produced both baselines on a single 3-pod openraft cluster:

  graceful soak:  1 / 37,410  errors = 0.0027%
  sigkill soak:   0 / 71,261  errors = 0.0000%

The SIGKILL result is the headline surprise: the client's retry +
leader-redirect masks the entire force-delete + re-election + rejoin
window, so the SIGKILL budget can sit at the same level as the graceful
soak rather than 10x looser. Both budgets now sit at 0.05% (5e-4),
~18x above the observed graceful baseline — tight enough that a real
regression in either the graceful or the crash-recovery path will
fail the lane, while leaving headroom for transport-layer jitter the
test cluster's network may not exhibit.

The MixedVersionSoak mode shares `MAX_SOAK_ERROR_RATE` with the
graceful soak, so it gets the same 0.05% budget; its docstring is
updated to match. No baseline for that lane yet, but it exercises a
strict superset of the graceful-soak teardown surface plus the
activation control plane, so the same budget is the right starting
point.

Coverage Stats

13414 of 14084 relevant lines covered (95.24%)

388694.96 hits per line

Jobs

ID	Job ID	Ran	Files	Coverage
1	26469498063.1	26 May 2026 07:18PM UTC	88	95.24	GitHub Action Run

prisma-risk / tsoracle / 26469498063
95%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 26469498063

prisma-risk / tsoracle / 26469498063 95%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 26469498063

prisma-risk / tsoracle / 26469498063
95%

README BADGES
x