Bug information
Steps to reproduce
Run the scheduled noncloud functional tests, including Test_FirstApplicationSample in test/functional-portable/samples/noncloud/tutorial_test.go. The failure is intermittent and occurs when the tutorial application takes longer to start listening than the post-deployment verification allows.
Observed in the September 6, 2026 workflow run 34003702190, samples-noncloud job, originally reported by #12913.
Observed behavior
Deployment and Kubernetes validation succeed, but the post-deployment HTTP verification exhausts three immediate retries before the application starts listening on port 3000:
[rad] Deployment Complete
Pod is ready tutorial:demo-tutorial-58f5675c5f-9fxcw
running post-deploy verification
01:35:39.530092 ... localhost:3000 ... connection refused
01:35:39.566572 ... localhost:3000 ... connection refused
01:35:39.594423 ... localhost:3000 ... connection refused
Get "http://localhost:34267/": EOF
tests failed after 3 retries
The container-log artifact records:
2026-09-06T01:35:40.846732302Z ... Server is running at http://localhost:3000
All three requests fail within approximately 64 milliseconds; the application starts listening approximately 1.25 seconds after the final attempt. Kubernetes pod readiness here does not establish that the application's HTTP endpoint is accepting connections.
Desired behavior
Post-deployment verification should allow a bounded application-startup interval, using context-aware readiness polling or retry backoff rather than exhausting all attempts immediately. Failed port-forward sessions should be recreated as needed. A genuinely unavailable application must still fail within a clear timeout and retain actionable diagnostics.
Consider an application HTTP readiness probe as well, rather than treating pod readiness alone as sufficient for the HTTP check.
Workaround
Rerun the failed job. The next scheduled September 6, 13:07 UTC run passed the samples job at the same Radius commit. This demonstrates intermittency, not a code fix.
System information
rad Version
CLI built from Radius commit c8ad9211a25699c377c45268890e4f67070aa114 in the failing workflow; released version output was not captured in this triage.
Operating system
Linux GitHub Actions runner with a local Kubernetes test cluster; see the linked job for its runner configuration.
Additional context
Bug information
Steps to reproduce
Run the scheduled noncloud functional tests, including
Test_FirstApplicationSampleintest/functional-portable/samples/noncloud/tutorial_test.go. The failure is intermittent and occurs when the tutorial application takes longer to start listening than the post-deployment verification allows.Observed in the September 6, 2026 workflow run 34003702190, samples-noncloud job, originally reported by #12913.
Observed behavior
Deployment and Kubernetes validation succeed, but the post-deployment HTTP verification exhausts three immediate retries before the application starts listening on port 3000:
The container-log artifact records:
All three requests fail within approximately 64 milliseconds; the application starts listening approximately 1.25 seconds after the final attempt. Kubernetes pod readiness here does not establish that the application's HTTP endpoint is accepting connections.
Desired behavior
Post-deployment verification should allow a bounded application-startup interval, using context-aware readiness polling or retry backoff rather than exhausting all attempts immediately. Failed port-forward sessions should be recreated as needed. A genuinely unavailable application must still fail within a clear timeout and retain actionable diagnostics.
Consider an application HTTP readiness probe as well, rather than treating pod readiness alone as sufficient for the HTTP check.
Workaround
Rerun the failed job. The next scheduled September 6, 13:07 UTC run passed the samples job at the same Radius commit. This demonstrates intermittency, not a code fix.
System information
rad Version
CLI built from Radius commit
c8ad9211a25699c377c45268890e4f67070aa114in the failing workflow; released version output was not captured in this triage.Operating system
Linux GitHub Actions runner with a local Kubernetes test cluster; see the linked job for its runner configuration.
Additional context