actions-runner-controller: Runner Container fails to terminate when runner fails to register with github
Starting Runner listener with startup type: service
Started listener process
An error occurred: Not configured
Runner listener exited with error code 2
Runner listener exit with retryable error, re-launch runner in 5 seconds.
This prevents the pod from ever recycling and re-registering.
The above can come about when the pod runner resource has been created, but fails to create the containers because the control plane fails to provision the pod to a node due to scheduling errors.
About this issue
- Original URL
- State: closed
- Created 4 years ago
- Comments: 20 (6 by maintainers)
Hey - circling back here. I ran into this issue again and it turned out to be a different flavor of what @arjunmadan has noted above. In my case, the ec2 machine we were running on did not have enough available memory to support the number of runner containers + infrastructure containers i was running on the github actions nodes. Once i bumped the memory of the ec2 node, the problem disappeared.
Interesting error though… seems to encapsulate a lot of failure cases. I think we should look back into limiting the number of restarts available.
I’ve been seeing this failure a lot, it seems to happen due to an error in the ./config.sh step that the service assumes will be corrected with manual intervention. So far I’ve seen the config fail due to expired github tokens and the github api being unresponsive. There’s an upstream request to change the behaviour so that the service eventually exits if it is not corrected (https://github.com/actions/runner/issues/879), but pending that, an option might be to exit the entrypoint script if the config step does not report success?