Local Image Factory E2E runs on macOS are using an amd64 VM through QEMU on an arm64 host, which makes Pepr startup slow enough that the Watcher and Admission liveness probes fail before startup completes. The pods enter restart loops and the watcher ends up in CrashLoopBackOff.
Increasing initialDelaySeconds on the liveness probes from 60 to 180 seconds allows the tests to complete, but that changes normal liveness behavior rather than only relaxing startup. startupProbe overrides appear to be dropped today because Pepr does not expose a startupProbe field in the generated chart objects.
This should add startupProbe override support for both the Pepr Watcher and Admission deployments, similar to the earlier readiness/liveness override support.
Related context: issue 1085 and PR 1086 changes.
Local Image Factory E2E runs on macOS are using an amd64 VM through QEMU on an arm64 host, which makes Pepr startup slow enough that the Watcher and Admission liveness probes fail before startup completes. The pods enter restart loops and the watcher ends up in CrashLoopBackOff.
Increasing initialDelaySeconds on the liveness probes from 60 to 180 seconds allows the tests to complete, but that changes normal liveness behavior rather than only relaxing startup. startupProbe overrides appear to be dropped today because Pepr does not expose a startupProbe field in the generated chart objects.
This should add startupProbe override support for both the Pepr Watcher and Admission deployments, similar to the earlier readiness/liveness override support.
Related context: issue 1085 and PR 1086 changes.