If a --health-cmd is defined for a container, block until its status is healthy or unhealthy. The timeout is defined by the server internal logic based on associated --health-* defined delays. If it blocks indefinitely, the job timeout will eventually cancel it.
While waiting, the simplest solution would be to sleep 1 second until the container is healthy or unhealthy. To minimize log verbosity, the sleep interval is instead set to --health-interval and default to one second if it is not defined.
This logic does not apply to host containers as they do not support services. They are assumed to always be healthy.
If --health-cmd is set for the container running a job, the first step will start to run without waiting for the container to become healthy. There may be valid use cases for that but they are not the focus of this implementation.
<!--start release-notes-assistant-->
<!--URL:https://code.forgejo.org/forgejo/runner-->
- features
- [PR](https://code.forgejo.org/forgejo/runner/pulls/805): <!--number 805 --><!--line 0 --><!--description ZmVhdDogd2FpdCBmb3Igc2VydmljZXMgdG8gYmUgaGVhbHRoeSBiZWZvcmUgc3RhcnRpbmcgYSBqb2I=-->feat: wait for services to be healthy before starting a job<!--description-->
<!--end release-notes-assistant-->
Reviewed-on: https://code.forgejo.org/forgejo/runner/pulls/805
Co-authored-by: Earl Warren <contact@earl-warren.org>
Co-committed-by: Earl Warren <contact@earl-warren.org>
96891ab314