kubernetes / kubernetes

Report event for the cases when probe returned Unknown result

Summary of Open Issue: Report event for cases when probe returned Unknown result

Issue Description: The Kubernetes project has identified a gap in the logging and event reporting for cases where container probes return an “Unknown” result. Currently, specific scenarios leading to an “Unknown” status exist (e.g., outlined in issue #106682). It is proposed that these cases be logged similarly to existing warnings, providing better visibility into probe statuses.

Proposed Implementation:

  1. Log and Event Reporting:
  • Introduce logging and event recording for “Unknown” probe results in the kubernetes/pkg/kubelet/prober/prober.go file, around line 115.
  • This should mirror the existing warning logging, but will utilize a new event type (e.g., v1.EventTypeUnknown instead of v1.EventTypeWarning).
  1. Testing:
  • Implement tests for any probe conditions that might return an “Unknown” result. This may involve both end-to-end (e2e) and unit tests.
  • Reference example e2e tests, such as those demonstrating how an invalid liveness probe command does not trigger container restarts (as seen in issue #106682).

Issue Management:

  • Multiple contributors have been assigned and have created pull requests addressing the issue, indicating a collaborative approach. However, as of the latest updates, clarity on whether the implementation is finalized or still active is required.

Labels:

  • good first issue: Suitable for newer developers.
  • help wanted: Indicates the community’s assistance in addressing the issue is needed.
  • priority/backlog: Suggests this issue is not high-priority but still on the radar for future development.
  • sig/node: Important for Kubernetes’ node management.

This summary should assist advanced developers in grasping the core objectives of the issue and understanding the current state of collaboration surrounding it.