Upgrading Gitlab Runner via Helm from 0.56.0 to 0.59.2 Breaks the Runner Token

We’ve been using gitlab-runner Helm chart version 0.56.0, which corresponds to app version 16.3.0. I attempted to upgrade the chart to 0.59.2, app version 16.6.1, which caused the new gitlab-runner pod to fail the liveness/readiness probes with the following message in the logs:

PANIC: The registration-token needs to be entered

We have a Kubernetes secret that contains the registration token, but for some reason the upgraded pod doesn’t detect it. Reverting to the previous chart version fixes the issue.

I know that the registration token has been deprecated in 15.6 in favor of runner-token, but since we’re already on version 16.*, I’d imagine that either a) the upgraded pod would use the registration token just like the old one did, or b) the pod with the older version would fail to use the token as well.

Does anyone know why this issue came up between versions 16.3 and 16.6?

Also, is it possible to just cut/paste the token inside the Kube secret from “runner-registration-token” to “runner-token” and have the new pod use that?

For posterity, the issue wasn’t with the name of the token field, but rather with the fact that starting with app version 16.1, there’s a way to specify the name of the Kube secret in values.yaml for the pod to use for retrieving that value. Apparently, a pod using version 16.6 won’t attempt to mount the secret by default unless it’s specified in values.yaml.