How is the ephemeral flag supposed to work?

[wilinger]

I have a RunnerDeployment below and after an action is ran once, Github shows the runner as Idle status and the runner-deployment pod is still running on my cluster. How do I make the runners terminate and only spin up on a new job?

apiVersion: actions.summerwind.dev/v1alpha1
kind: RunnerDeployment
metadata:
  name: runner-deployment
spec:
  template:
    spec:
      repository: wilinger/test-self-hosted-runner
      env:
        - name: RUNNER_FEATURE_FLAG_EPHEMERAL
          value: "true"

[mumoshu]

How do I make the runners terminate and only spin up on a new job?

@wilinger See https://github.com/actions-runner-controller/actions-runner-controller#autoscaling-tofrom-0. I recommend webhook-based autoscaler which is more responsive than other options 😄

[achebel]

@mumoshu,
Actually we tested the ephemeral mode with GHES 3.3.1 :

• When runner is started it registers itself to the organization => Runner was in idle status
• Then Github looked for an online and idle runner that matched the job's runs-on label => Status of the runner switched to from idle to active
• Then after running the job, the runner is automatically unregistered from our GitHub instance
  and the runner was automatically shutdown.
• However after a few seconds later a new runner is provisonned based on the capacity reservation. Unfortunately It seems that we are not able to disabled it? is it possible to only spin up a new runner only when a new job is trigerred?

Thanks in advance
Regards