Runners Controller 0.11.0 improves observability for Linux runners that fail mid-job. When a runner exits silently, you can now diagnose the cause from both the runner and the host side.
- Runner-side logs: the actions/runner diagnostic log is now printed to stdout. This means the runner’s own reason for exiting is captured by the pod-log pipeline and survives after the Pod is removed.
- Host-side logs: the log collector now ships node journald logs, including containerd, kubelet, and kernel messages. This captures host-level teardown events, such as microVM or kernel-level kills, that the runner itself cannot see.
Both instruments are enabled by default in this release to help identify the root cause of unexpected Linux runner deaths.