v16: CI observability + lifecycle via the new Actions APIs #78

Nytt ärende

Stängd

öppnade 2026-07-17 11:52:21 +00:00 av supernaut · 1 kommentar

supernaut kommenterad

2026-07-17 11:52:21 +00:00

Ägare

Tier 2 v16 follow-up (gated on #73 landing). v16 adds Actions APIs we lacked: list run jobs (PR 11915), download job logs (PR 12666), list/download/delete artifacts (PR 12140), POST .../runs/{id}/cancel (PR 12957), plus workflow-run deletion (PR 12478).

Evaluate three uses (re-verify endpoints against /api/swagger post-upgrade):

CI log export → Loki (ADR 0020): ship job logs to the monitoring VM so CI failures are searchable next to service logs; also mitigates LOG_RETENTION_DAYS = 30.
Artifact retention/cleanup: programmatic deletion beyond the blanket ARTIFACT_RETENTION_DAYS = 30 (e.g. keep release artifacts, purge PR builds early) — quota/disk relief.
runner-controller cancel path (ADR 0021): wire the cancel API into scale-down/orphan cleanup so a stuck job's VM can be reclaimed deliberately instead of waiting for timeout.

Note: v16 also evaluates job-level if: server-side (PR 13030), so fully-skipped jobs never dispatch → ephemeral VMs no longer boot for skipped jobs. Verify post-upgrade and note the saving in ADR 0021 docs (part of #73 verification, not this issue).

**Tier 2 v16 follow-up** (gated on #73 landing). v16 adds Actions APIs we lacked: list run jobs (PR 11915), download **job logs** (PR 12666), list/download/delete **artifacts** (PR 12140), `POST .../runs/{id}/cancel` (PR 12957), plus workflow-run **deletion** (PR 12478). Evaluate three uses (re-verify endpoints against `/api/swagger` post-upgrade): - [ ] **CI log export → Loki** (ADR 0020): ship job logs to the monitoring VM so CI failures are searchable next to service logs; also mitigates `LOG_RETENTION_DAYS = 30`. - [ ] **Artifact retention/cleanup**: programmatic deletion beyond the blanket `ARTIFACT_RETENTION_DAYS = 30` (e.g. keep release artifacts, purge PR builds early) — quota/disk relief. - [x] **runner-controller cancel path** (ADR 0021): wire the cancel API into scale-down/orphan cleanup so a stuck job's VM can be reclaimed deliberately instead of waiting for timeout. Note: v16 also evaluates job-level `if:` server-side (PR 13030), so fully-skipped jobs never dispatch → ephemeral VMs no longer boot for skipped jobs. Verify post-upgrade and note the saving in ADR 0021 docs (part of #73 verification, not this issue).

supernaut lade till etiketterna

2026-07-17 11:52:31 +00:00

supernaut lade till detta till projektet Bitborg Infrastructure

2026-07-17 11:55:48 +00:00

supernaut refererade till detta ärende

2026-07-17 14:07:19 +00:00

upgrade forgejo 15.0.4 → 16.0.0 #73

supernaut lade till