Goals

• Provide additional capabilities to administrator to influence collector pod scheduling

Non-Goals

Motivation

Service delivery and HCP teams that provide clusters for customers have requirements to forward logs from multiple tenants on the control plane to their respective log stores. They can accomplish this by defining multiple ClusterLogForwarders but this can result in idle collector pods running on nodes where they are not needed. The intent of this enhancement allows them to spec affinity/anti-affinity rules for the collector to allow them to target specific nodes in the control plane.

Alternatives

Acceptance Criteria

• Verify collectors only run on nodes which meet the spec'd affinity or anti-affinity for their respective ClusterLogForwarder
• Verify affinity/anti-affinity rules are not required to deploy a ClusterLogForwarder

Risk and Assumptions

Documentation Considerations

• Update the scheduling documentation to include the new options

Open Questions

Additional Notes