CI Visibility Dashboard

Generated 2026-07-08 05:15:48 UTC · lookback 30 days · total runs sampled: 1000 ·

CI success rate (strict)
75%
success / (success + failure) — excludes cancellations
Cancellations
2%
17 of 997 completed — see breakdown below
Time-to-green (p50)
0.6 min
Time-to-green (p95)
6.6 min
Why cancellations are not failures: most cancels are by design — concurrency cancel-in-progress (concurrency-cancel pattern, old run yields to newer), PR merged/closed (auto-cancel), or operator workflow disable (e.g. 2026-05-08 marathon mass-disable during zombie incident). Only timeout-cancellations indicate a quality problem. Strict success rate: 75% · Loose rate (incl. cancellations): 43%.

Root cause health (FMEA — ADR-209)

One row per root cause from #908 deep CI fragility RCA / ADR-209. Color = current band (healthy · watch · action). Closes the C5 visibility residual from #908 (issue #920).

CauseSymptomMetricStatusNotes
FMEA metrics unavailable.

Queue Depth Trend

max: 32026-07-08 (3)

Daily count of runs whose status was queued or in_progress at fetch time. A historical proxy — not the live queue depth.

Failure Rate by Workflow (top 10, strict)

Failure rate = failed / (success + failed). Cancellations excluded from denominator. Sorted by failure rate descending.

WorkflowTotalFailedCancelledFailure rate (strict)Cancel rate
.github/workflows/dep-vuln-scan.yml1001000
100.0%
0%
vitest nightly (full suite)330
100.0%
0%
light-touch-tier7715
33.3%
6%
ci94211
22.6%
1%
pr-gate80173
22.1%
4%
Architecture Validation3600
0.0%
0%
Auto-merge pipeline10200
0.0%
0%
Daily digest200
0.0%
0%
ci-docs-stubs1300
0.0%
0%
ci-lite-gate7701
0.0%
1%

Time-to-Green

MetricValue
p50 (median run duration)0.6 min
p956.6 min
Sample size429

Successful runs only. Duration = updated_at − created_at.

Top 5 Workflows by Failure Rate (strict)

WorkflowTotalFailedCancelledFailure rate (strict)
.github/workflows/dep-vuln-scan.yml1001000100.0%
ci9421122.6%
pr-gate8017322.1%

Strict failure rate = failed / (success + failed). Workflows with <5 non-cancelled runs excluded. Cancelled column shown for context only — not included in failure rate.

Stuck Jobs (>60 min in_progress)

No stuck jobs — all in-progress runs are within the 60-minute threshold.

PR Cycle Time (from pr-events.jsonl)

MetricValue
PRs merged (sample)200
Average merge time122.9 min
p5021.7 min
p95340.8 min

Source: claudedocs/dispatches/pr-events.jsonl (#583). Only merged PRs counted.