KEP-6247keps/sig-node/6247-kubelet-watchdog-diagnostics
Kubelet Systemd Watchdog Diagnostic Guardrails
On Linux nodes, kubelet can use the systemd watchdog. It periodically runs health checkers and sends a heartbeat through SdNotify(). If heartbeats stop, systemd can restart kubelet. This KEP makes existing watchdog failures easier to diagnose. It adds default-visible structured logs for checker errors and SdNotify() errors, plus one summary when existing no… Full summary
On Linux nodes, kubelet can use the systemd watchdog. It periodically runs health checkers and sends a heartbeat through SdNotify(). If heartbeats stop, systemd can restart kubelet. This KEP makes existing watchdog failures easier to diagnose. It adds default-visible structured logs for checker errors and SdNotify() errors, plus one summary when existing notification retries are exhausted. The new diagnostics are controlled by the Alpha KubeletWatchdogDiagnostics feature gate, which is disabled by default. The feature does not change watchdog execution, timing, retry count, socket behavior, or cancellation.
Facts
- Authors
- @googs1025
- Approvers
- @dchen1107
- Reviewers
- @sergeykanzhelev
- PRR approvers
- @jpbetzalpha
- Tracking issue
- #6247(opens GitHub in a new tab) openmilestone v1.38sig/nodestage/alphakind/featuretracked/yeslead-opted-in
- Milestones
- alpha v1.38
- Participating SIGs
- none
- Feature gates
- KubeletWatchdogDiagnostics
- Created / last updated
- 2026-07-19 / none
Loading decisions and status history