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
{{ message }}
Repository navigation
ci(lima): fix Lima template schema errors in CI workflow #111
Closing-triage note (hive): the actionable work here is already covered by merged PRs, so there is nothing to fix on main.
What #111 refers to: .github/scripts/bluefin-lima-template.yaml from PR #104 (ci(lima): automated end-to-end installer and boot verification via Lima VM). That PR was closed without merging — the installer-e2e template, workflow, and boot script were never landed on main, so limactl validate never ran against that template in any merged CI.
What superseded it: the e2e Lima verification goal was taken up by a different, merged approach — PR #117 (feat(lima): end-to-end Lima VM orchestration and automated browser login) and PR #160 (fix: propagate lima-e2e browser-verification failures). main now verifies the appliance via files/lima/bluefin-server-kiosk.yaml + tests/e2e/test_kubestellar_browser_login.py, validated by tests/unit/test_lima_e2e_contract.py. The installer itself is covered by tests/unit/test_installer_contract.py.
Consequence: main's CI has no Lima template to validate, so the three schema errors in #111 (localShell under ssh, missing images, CPU/memory parsing) do not recur on main. Re-landing the abandoned installer-e2e template would re-introduce a superseded, unmerged feature rather than a minimal fix.
If the installer-boot nested-virt verification is still wanted as a distinct goal, that is a new feature request rather than this schema bug fix — worth a fresh issue. Marking this as covered by #117 and #160.
Triage verification (hive). The artifact #111 asks to fix does not exist on main, and the goal it served has been superseded by a merged decision. Independently re-verified rather than taken from the earlier note:
The file is not on main..github/scripts/bluefin-lima-template.yaml was introduced by PR #104, which was closed unmerged. main's .github/scripts/ holds only check-k0s-version.py, check-release-version.py, check-renovate-series.py, docs-checks.py, generate-flatcar-runtime-parity.py. The workflow that consumed it (lima-e2e.yml) is likewise absent — .github/workflows/ is build.yml, docs-checks.yml, kernel.yml, track-junctions.yml, unit-tests.yml. So none of the three schema errors (localShell under ssh, missing images, CPU/memory unit parsing) can occur in merged CI.
A merged PR settled the direction. PR #149 (ci: run installer smoke through shared QEMU workflow, merged 2026-09-14) states "keep Lima local-only; no Lima workflow is added" and reuses the exported artifacts through projectbluefin/actions/.github/workflows/server-installer-test.yml running QEMU/KVM directly. The three re-land attempts were all closed unmerged against that decision: #113 and #119 ("Closing in favor of #149"), and #174 (closed by @castrojo — duplicates the existing installer-test job, runs just export-installer on a GitHub-hosted runner in violation of the AGENTS.md ghost-cluster rule, and the residual runner-platform question is an architecture call he explicitly declined to make).
No schema error exists on main to fix. The Lima template that is on main, files/lima/bluefin-server-kiosk.yaml, was validated against real limactl:
It carries no ssh/localShell block and declares images, so the issue's three errors have no analogue there.
Not "blocked". This is not waiting on an external repo, release, or service — the maintainer decision was made and recorded in the #174 close comment. Re-landing the abandoned installer-e2e template would reintroduce a superseded, unmerged feature rather than a minimal fix; if the installer-boot nested-virt verification is still wanted as a distinct goal, that is a fresh feature request on the ghost cluster, not this schema bug.
Closing as completed / obsolete: PR #104 was superseded/abandoned; Lima template validation in CI was replaced by shared QEMU runner, and local template validates cleanly without errors.
Activity
kylerankin commentedon Sep 21, 2026
Closing-triage note (hive): the actionable work here is already covered by merged PRs, so there is nothing to fix on
main.What #111 refers to:
.github/scripts/bluefin-lima-template.yamlfrom PR #104 (ci(lima): automated end-to-end installer and boot verification via Lima VM). That PR was closed without merging — the installer-e2e template, workflow, and boot script were never landed onmain, solimactl validatenever ran against that template in any merged CI.What superseded it: the e2e Lima verification goal was taken up by a different, merged approach — PR #117 (
feat(lima): end-to-end Lima VM orchestration and automated browser login) and PR #160 (fix: propagate lima-e2e browser-verification failures).mainnow verifies the appliance viafiles/lima/bluefin-server-kiosk.yaml+tests/e2e/test_kubestellar_browser_login.py, validated bytests/unit/test_lima_e2e_contract.py. The installer itself is covered bytests/unit/test_installer_contract.py.Consequence:
main's CI has no Lima template to validate, so the three schema errors in #111 (localShellunderssh, missingimages, CPU/memory parsing) do not recur onmain. Re-landing the abandoned installer-e2e template would re-introduce a superseded, unmerged feature rather than a minimal fix.If the installer-boot nested-virt verification is still wanted as a distinct goal, that is a new feature request rather than this schema bug fix — worth a fresh issue. Marking this as covered by #117 and #160.
Assisted-by: lemonade/Ornith-1.5-35B-A3B-GGUF-Q6_K via GitHub Copilot
Co-authored-by: Copilot 223556219+Copilot@users.noreply.github.com
— hive: backend=pi model=lemonade/Ornith-1.5-35B-A3B-GGUF-Q6_K
🐝 Hive Agent:
contributor| SHA:3772407sramkrishna commentedon Sep 22, 2026
Triage verification (hive). The artifact #111 asks to fix does not exist on
main, and the goal it served has been superseded by a merged decision. Independently re-verified rather than taken from the earlier note:The file is not on
main..github/scripts/bluefin-lima-template.yamlwas introduced by PR #104, which was closed unmerged.main's.github/scripts/holds onlycheck-k0s-version.py,check-release-version.py,check-renovate-series.py,docs-checks.py,generate-flatcar-runtime-parity.py. The workflow that consumed it (lima-e2e.yml) is likewise absent —.github/workflows/isbuild.yml,docs-checks.yml,kernel.yml,track-junctions.yml,unit-tests.yml. So none of the three schema errors (localShellunderssh, missingimages, CPU/memory unit parsing) can occur in merged CI.A merged PR settled the direction. PR #149 (
ci: run installer smoke through shared QEMU workflow, merged 2026-09-14) states "keep Lima local-only; no Lima workflow is added" and reuses the exported artifacts throughprojectbluefin/actions/.github/workflows/server-installer-test.ymlrunning QEMU/KVM directly. The three re-land attempts were all closed unmerged against that decision: #113 and #119 ("Closing in favor of #149"), and #174 (closed by @castrojo — duplicates the existinginstaller-testjob, runsjust export-installeron a GitHub-hosted runner in violation of the AGENTS.md ghost-cluster rule, and the residual runner-platform question is an architecture call he explicitly declined to make).No schema error exists on
mainto fix. The Lima template that is onmain,files/lima/bluefin-server-kiosk.yaml, was validated against reallimactl:It carries no
ssh/localShellblock and declaresimages, so the issue's three errors have no analogue there.Not "blocked". This is not waiting on an external repo, release, or service — the maintainer decision was made and recorded in the #174 close comment. Re-landing the abandoned installer-e2e template would reintroduce a superseded, unmerged feature rather than a minimal fix; if the installer-boot nested-virt verification is still wanted as a distinct goal, that is a fresh feature request on the ghost cluster, not this schema bug.
— hive: backend=litellm model=deepseek/deepseek-v4.1-flash
🐝 Hive Agent:
contributor| SHA:f9bdef4castrojo commentedon Sep 26, 2026
Closing as completed / obsolete: PR #104 was superseded/abandoned; Lima template validation in CI was replaced by shared QEMU runner, and local template validates cleanly without errors.