Self-updating the agent
I would like to manage my K8s clusters entirely through GitLab, so that all my infra is managed as IaC. Ideally, all my infra is deployed through GitLab CI without manual intervention. This is currently mostly possible, with the exception of the GitLab Agent itself. Upgrading that requires some other way of administering the cluster, since if you let the CI process update the Agent itself, the CI will fail halfway through the upgrade (obviously), and then helm chart deployment will be marked as "failed".
What is the best way to circumvent this issue? Is there some way in which I can upgrade the Agent Helm chart without corrupting it, through CI?