fix(openaev): align healthcheck on application-level /api/health endpoint

[RomuDeuxfois]

What

Aligns the openaev service healthcheck in docker-compose.yml with the pattern already used for opencti in the same file, and with OpenAEV-Platform/docker's own compose file.

     healthcheck:
-      test: [ "CMD", "wget", "-qO-", "http://openaev:8080/actuator/health/ping" ]
+      test: [ "CMD", "wget", "-qO-", "http://openaev:8080/api/health?health_access_key=${OPENAEV_HEALTHCHECK_KEY}" ]
       interval: 10s
       timeout: 5s
-      retries: 30
-      start_period: 120s
+      retries: 20
+      start_period: 120s

Why

• /actuator/health/ping is a generic Spring Boot Actuator liveness probe — it only proves the JVM process responds, with no visibility into whether OpenAEV's actual dependencies (Postgres, MinIO, RabbitMQ, Elasticsearch) are reachable.
• /api/health?health_access_key= is OpenAEV's application-level health endpoint — a real readiness check, same mechanism OpenCTI already uses in this same file (/health?health_access_key=${OPENCTI_HEALTHCHECK_ACCESS_KEY}) and the one used in OpenAEV-Platform/docker.
• Several services depend on openaev: condition: service_healthy (xtm-composer, all collector-* and injector-* services). A liveness-only check means those consumers can start against an OpenAEV instance that's "up" but not actually ready to serve traffic (e.g. DB migration still running).
• OPENAEV_HEALTHCHECK_KEY is already defined and injected in the openaev service environment (OPENAEV_HEALTHCHECK_KEY=${OPENAEV_HEALTHCHECK_KEY:-ChangeMe}), so no new secret/env var is introduced.

Notes

• retries lowered from 30 → 20 to match OpenAEV-Platform/docker.
• start_period: 120s kept: this is now a real app-level readiness check (DB/ES/MinIO init), which can reasonably take longer to go green than a bare actuator ping — happy to drop it if we'd rather match OpenAEV-Platform/docker exactly (no start_period there).

Scope / impact

• Infra-only change (docker-compose.yml), no application code touched.
• No CE/EE impact, no public API breaking change, no multi-tenancy impact.
• Slightly stricter startup gating for xtm-composer / collectors / injectors that wait on openaev: service_healthy — expected and desired.

Related discussion: Teams thread on healthcheck inconsistency between docker and xtm-docker repos.

[Copilot]

Pull request overview

Updates the openaev service healthcheck in docker-compose.yml to use OpenAEV’s application-level readiness endpoint (/api/health) instead of the Spring Actuator liveness ping, ensuring dependent services only start once OpenAEV is actually ready.

Changes:

• Switched openaev healthcheck from /actuator/health/ping to /api/health?health_access_key=....
• Reduced healthcheck.retries from 30 to 20 (kept start_period: 120s).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.