Summary
Add a narrow, fast deploy/uptime health probe at GET /rest/v1/health, off by default behind the CRE_ENABLE_HEALTH feature flag.
Scope (v1 — thin)
On prod we only need the DB for serving, so the check is DB + data-sanity only:
- No Neo4j / Redis in the health check.
- 200 = app up + DB reachable + minimal data sanity (CRE count > 0, standards non-empty).
- 503 = DB unreachable or obviously broken/empty dataset.
Out of scope (by design)
Full GA / mapping completeness does not belong in a deploy health check — too slow and flaky, and it can fail deploys for the wrong reason. Deeper checks stay in ops land: verify_ga_completeness, monitor_ga_health, the weekly automation, etc.
If we want a heavier "deep health" later, we can add it separately. Ship the thin version first.
Summary
Add a narrow, fast deploy/uptime health probe at
GET /rest/v1/health, off by default behind theCRE_ENABLE_HEALTHfeature flag.Scope (v1 — thin)
On prod we only need the DB for serving, so the check is DB + data-sanity only:
Out of scope (by design)
Full GA / mapping completeness does not belong in a deploy health check — too slow and flaky, and it can fail deploys for the wrong reason. Deeper checks stay in ops land:
verify_ga_completeness,monitor_ga_health, the weekly automation, etc.If we want a heavier "deep health" later, we can add it separately. Ship the thin version first.