Hello Google Cloud community,
We are troubleshooting authenticated communication from a Cloud Run Job to a Cloud Run Gen2 service using internal-only ingress, in the same Google Cloud project and region.
Configuration:
- Cloud Run target requires authentication and internal ingress.
- The Job service account has roles/run.invoker.
- Google-signed OIDC ID token with the target service URL as audience.
- Direct VPC egress configured for all traffic; subnet has Private Google Access.
- We also tested a dedicated Private Service Connect endpoint for Google APIs, with private DNS resolution.
- Requests use the official run.app hostname without overriding the Host header.
Observed results:
- The target container starts and passes startup/liveness probes.
- Ten authenticated requests using Direct VPC egress returned HTTP 404.
- Six authenticated requests via PSC also returned HTTP 404.
- The application logs a request identifier for requests reaching /healthz and responds HTTP 200, but none of these test request identifiers appeared in the collected target revision request or application logs.
- Requests from outside the permitted private network were also rejected.
We have not established whether the requests were rejected by Cloud Run ingress classification, IAM, or another frontend routing condition.
Questions:
- What specific requirements must a Cloud Run Job meet for its request to be classified as internal by a Cloud Run service?
- Can Cloud Run return 404 before reaching the target revision even with a valid OIDC token and run.invoker permission?
- How can we determine the precise rejection stage and inspect the effective ingress decision?
- What minimal working configuration would you recommend for an authenticated Job calling an internal-only Cloud Run service through Direct VPC egress or PSC?
Thank you for any guidance.