Summary
Introduce an option for customers to automatically schedule active single node health checks as low-priority, preemptible Kubernetes jobs, leveraging Kubernetes pod priority and preemption features (reference).
Requirements
- Add a configuration toggle during the GPU scanner data plane plugin installation to enable/disable scheduling health checks as preemptible jobs.
- Implement health check jobs with low priority so they are preemptible within Kubernetes clusters.
- Provide clear documentation and sample configuration showing how to enable this toggle and set up pod priority/preemption for health check jobs.
- Ensure the default behavior is unchanged unless explicitly configured by the customer.
Acceptance Criteria
- Customers can enable automatic scheduling of health checks as preemptible jobs during installation.
- Health checks run as low-priority jobs when toggle is enabled and are subject to Kubernetes preemption.
- Documentation is updated with instructions and examples for configuring this feature.
Labels: enhancement, documentation
Issue Type: Feature
Summary
Introduce an option for customers to automatically schedule active single node health checks as low-priority, preemptible Kubernetes jobs, leveraging Kubernetes pod priority and preemption features (reference).
Requirements
Acceptance Criteria
Labels: enhancement, documentation
Issue Type: Feature