UAT defect — blocker
PR under test: #2376 (chore: offload projects to nestjs)
Preview: https://console-pr-2376.dso.cpin-hp.numerique-interieur.fr/
Tester: debugger (automated UAT)
Date: 2026-07-28
Severity: Blocker — the primary feature of the PR (project reprovisioning offloaded to server-nestjs) is unreachable.
Symptom
Every route the strangler now proxies to server-nestjs returns HTTP 502 Bad Gateway:
/api/v1/projects → 502/api/v1/projects-bulk → 502/api/v1/service-chains → 502/api/v1/projects/<id>/deployments → 502/api/v2/ (nestjs-only, does not exist on legacy) → 502
Legacy backend is healthy and unaffected:
/api/v1/healthz → {"status":"OK"} (legacy format, still served by legacy)/api/v1/users → 302 (legacy auth redirect)/health (nginx-strangler self) → 200
Reproduction
- Open https://console-pr-2376.dso.cpin-hp.numerique-interieur.fr/api/v1/projects
- Observe
502 Bad Gateway (nginx 1.31.3).
- Re-poll 3x — result is persistent, not transient.
Root cause (consistent with prior task t_3cb173ad)
The server-nestjs container crashes at startup in the preview environment (process crash ~5.5s, not a slow healthcheck), so nginx cannot reach the upstream. CI passes (build/unit/lint/integration) because it does not run the container against the live preview env. The crash must be fixed before the offloaded routes can be exercised.
Acceptance impact
UAT cannot be completed: the offloaded behavior cannot be exercised, so new-feature correctness, edge cases, and regression scope are all unverifiable. Recommend blocking merge until server-nestjs starts cleanly in preview and migrated routes return non-502.
Evidence
curl -k https://console-pr-2376.dso.cpin-hp.numerique-interieur.fr/api/v2/ → 502 (repeated)curl -k https://console-pr-2376.dso.cpin-hp.numerique-interieur.fr/api/v1/healthz → {"status":"OK"} (legacy)
UAT defect — blocker
PR under test: #2376 (chore: offload projects to nestjs)
Preview: https://console-pr-2376.dso.cpin-hp.numerique-interieur.fr/
Tester: debugger (automated UAT)
Date: 2026-07-28
Severity: Blocker — the primary feature of the PR (project reprovisioning offloaded to server-nestjs) is unreachable.
Symptom
Every route the strangler now proxies to
server-nestjsreturns HTTP 502 Bad Gateway:/api/v1/projects→ 502/api/v1/projects-bulk→ 502/api/v1/service-chains→ 502/api/v1/projects/<id>/deployments→ 502/api/v2/(nestjs-only, does not exist on legacy) → 502Legacy backend is healthy and unaffected:
/api/v1/healthz→{"status":"OK"}(legacy format, still served by legacy)/api/v1/users→ 302 (legacy auth redirect)/health(nginx-strangler self) → 200Reproduction
502 Bad Gateway(nginx 1.31.3).Root cause (consistent with prior task t_3cb173ad)
The
server-nestjscontainer crashes at startup in the preview environment (process crash ~5.5s, not a slow healthcheck), so nginx cannot reach the upstream. CI passes (build/unit/lint/integration) because it does not run the container against the live preview env. The crash must be fixed before the offloaded routes can be exercised.Acceptance impact
UAT cannot be completed: the offloaded behavior cannot be exercised, so new-feature correctness, edge cases, and regression scope are all unverifiable. Recommend blocking merge until
server-nestjsstarts cleanly in preview and migrated routes return non-502.Evidence
curl -k https://console-pr-2376.dso.cpin-hp.numerique-interieur.fr/api/v2/→ 502 (repeated)curl -k https://console-pr-2376.dso.cpin-hp.numerique-interieur.fr/api/v1/healthz→{"status":"OK"}(legacy)