You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Over the last 100 runs, only 2 are non-success, and both are action_required (runs awaiting approval that never execute). Fleet Monitor counts action_required as a failure (fleet_monitor.sh:140), so these two drive the 10.5% rate.
Inspecting the two flagged runs by head SHA:
SHA 1d366770: twopull_request runs at the same second β one successand one action_required. A classic duplicate/superseded run with no concurrency control.
SHA f1c0172: a lone gated run (pure approval artifact, not fixable in the workflow file).
The retry mechanism (added in #290/#383) already eliminated the transient-CDN real failures β last 100 runs contain zero failure conclusions. The residual noise is duplicate pull_request runs. sonarcloud.yml currently has no concurrency block, which the org ci-standards.md prescribes (SHA-scoped) for exactly this reason.
Scope
Add a SHA-scoped concurrency block to .github/workflows/sonarcloud.yml so duplicate/superseded runs for the same commit are cancelled (counted separately, not as failures) instead of lingering as action_required.
Implementation Checklist
Add concurrency block (group: sonarcloud-${{ github.ref }}-${{ github.sha }}, cancel-in-progress: true) to .github/workflows/sonarcloud.yml, matching the org ci-standards SHA-scoped pattern (workflow-name-prefixed group so it never collides with ci.yml).
Tests to Write
No app-level test harness exists for workflow YAML. Per the Phase-2 exception (behavior-preserving CI configuration change), validate via YAML parse + actionlint/lint instead of unit tests. Noted in the completion report.
Over the last 100 sonarcloud.yml runs, zero were failure β the retry added in #290/#383 already eliminated the transient scanner-CDN failures. The only 2 non-success runs were both action_required, which fleet_monitor.sh:140 counts as failures. Inspecting them by head SHA:
SHA 1d366770: twopull_request runs for one commit at the same second β one success + one action_required. A duplicate/superseded run with no concurrency control to cancel it.
SHA f1c0172: a lone approval-gated run (not fixable from a workflow file).
sonarcloud.yml had no concurrency block, which the org ci-standards.md prescribes (SHA-scoped) precisely to keep duplicate runs from lingering.
Fix
Added a SHA-scoped concurrency block. Duplicate runs for the same commit now supersede each other and resolve as cancelled (a separately-counted, non-failure conclusion) instead of lingering as action_required. SHA-scoping (vs per-ref) means distinct commits keep their own slots, so an in-flight scan for a still-relevant commit is never cancelled β matching the org rationale. The sonarcloud- prefix keeps the group distinct from ci.yml's (concurrency groups are repo-global).
Plan Execution
Root-caused the failure rate to action_required duplicate runs
Added SHA-scoped concurrency block to .github/workflows/sonarcloud.yml
Validated YAML parses; concurrency present
bash .dev-lead/scripts/dev-lead-lint.sh β all checks passed
npx prettier --check β passes
Test Results
$ python3 -c "import yaml; yaml.safe_load(open('.github/workflows/sonarcloud.yml'))"
YAML OK
concurrency: {'group': 'sonarcloud-${{ github.ref }}-${{ github.sha }}', 'cancel-in-progress': True}
$ bash .dev-lead/scripts/dev-lead-lint.sh
[lint] shellcheck on 5 file(s)...
[lint] validating CODEOWNERS...
OK: CODEOWNERS
[lint] all checks passed
$ npx prettier --check .github/workflows/sonarcloud.yml
Checking formatting...
All matched files use Prettier code style!
Files Changed
.github/workflows/sonarcloud.yml: added a SHA-scoped concurrency block (group sonarcloud-${{ github.ref }}-${{ github.sha }}, cancel-in-progress: true) between the on: triggers and jobs:, with an explanatory comment.
Notes
Phase-2 (TDD) exception: behavior-preserving CI-config change. The repo's Vitest workspaces are scoped to src/**, so no app test can exercise a .github/workflows/*.yml file. Verified via YAML parse + dev-lead-lint + Prettier instead.
Residual: the lone action_required run (f1c0172, no superseding sibling) stems from GitHub approval-gating of a bot-authored PR and cannot be resolved in the workflow file. The concurrency block removes the duplicate class, which is what pushed the rate over the 10% threshold β expected to drop the warning back under threshold.
Activity
don-petry commentedon Jul 22, 2026
Dev-Lead Implementation Plan
Issue: #401 β [Fleet Monitor] petry-projects/TalkTerm β .github/workflows/sonarcloud.yml
Root-cause analysis
Over the last 100 runs, only 2 are non-
success, and both areaction_required(runs awaiting approval that never execute). Fleet Monitor countsaction_requiredas a failure (fleet_monitor.sh:140), so these two drive the 10.5% rate.Inspecting the two flagged runs by head SHA:
1d366770: twopull_requestruns at the same second β onesuccessand oneaction_required. A classic duplicate/superseded run with no concurrency control.f1c0172: a lone gated run (pure approval artifact, not fixable in the workflow file).The retry mechanism (added in #290/#383) already eliminated the transient-CDN real failures β last 100 runs contain zero
failureconclusions. The residual noise is duplicatepull_requestruns.sonarcloud.ymlcurrently has noconcurrencyblock, which the orgci-standards.mdprescribes (SHA-scoped) for exactly this reason.Scope
Add a SHA-scoped
concurrencyblock to.github/workflows/sonarcloud.ymlso duplicate/superseded runs for the same commit are cancelled (counted separately, not as failures) instead of lingering asaction_required.Implementation Checklist
concurrencyblock (group: sonarcloud-${{ github.ref }}-${{ github.sha }},cancel-in-progress: true) to.github/workflows/sonarcloud.yml, matching the org ci-standards SHA-scoped pattern (workflow-name-prefixed group so it never collides withci.yml).Tests to Write
actionlint/lint instead of unit tests. Noted in the completion report.Test Command
bash .dev-lead/scripts/dev-lead-lint.sh+python3 -c "import yaml,sys; yaml.safe_load(open('.github/workflows/sonarcloud.yml'))"don-petry commentedon Jul 22, 2026
Dev-Lead: Implementation Complete
Root cause
Over the last 100
sonarcloud.ymlruns, zero werefailureβ the retry added in #290/#383 already eliminated the transient scanner-CDN failures. The only 2 non-successruns were bothaction_required, whichfleet_monitor.sh:140counts as failures. Inspecting them by head SHA:1d366770: twopull_requestruns for one commit at the same second β onesuccess+ oneaction_required. A duplicate/superseded run with no concurrency control to cancel it.f1c0172: a lone approval-gated run (not fixable from a workflow file).sonarcloud.ymlhad noconcurrencyblock, which the orgci-standards.mdprescribes (SHA-scoped) precisely to keep duplicate runs from lingering.Fix
Added a SHA-scoped
concurrencyblock. Duplicate runs for the same commit now supersede each other and resolve ascancelled(a separately-counted, non-failure conclusion) instead of lingering asaction_required. SHA-scoping (vs per-ref) means distinct commits keep their own slots, so an in-flight scan for a still-relevant commit is never cancelled β matching the org rationale. Thesonarcloud-prefix keeps the group distinct fromci.yml's (concurrency groups are repo-global).Plan Execution
action_requiredduplicate runsconcurrencyblock to.github/workflows/sonarcloud.ymlconcurrencypresentbash .dev-lead/scripts/dev-lead-lint.shβ all checks passednpx prettier --checkβ passesTest Results
Files Changed
.github/workflows/sonarcloud.yml: added a SHA-scopedconcurrencyblock (groupsonarcloud-${{ github.ref }}-${{ github.sha }},cancel-in-progress: true) between theon:triggers andjobs:, with an explanatory comment.Notes
src/**, so no app test can exercise a.github/workflows/*.ymlfile. Verified via YAML parse + dev-lead-lint + Prettier instead.action_requiredrun (f1c0172, no superseding sibling) stems from GitHub approval-gating of a bot-authored PR and cannot be resolved in the workflow file. The concurrency block removes the duplicate class, which is what pushed the rate over the 10% threshold β expected to drop the warning back under threshold.feat: implement issue #401 β [Fleet Monitor] petry-projects/TalkTerm β¦
feat: implement issue #401 β [Fleet Monitor] petry-projects/TalkTerm β¦
feat: implement issue #401 β [Fleet Monitor] petry-projects/TalkTerm β¦
feat: implement issue #401 β [Fleet Monitor] petry-projects/TalkTerm β¦