Summary

Investigate options for addressing collectors potentially consuming too much node disk

Acceptance Criteria

• Written Plan of options
• Agreement of a path forward by logging team
• Feasibility of adding into 5.9.z stream
• Create cards for implementing a solution

Notes

• Fluentd singleton introduced a cap of 15% evaluated at runtime to address a 3.x bug where it consumed too much node disk and caused the API node to crash
• OCP 4.x world with multi CLF is more difficult because nodes can vary in size and, there can multipl CLF competing for the same aggregate total disk, unable to make calculations at runtime

Initial Option Ideas

• Alert that triggers when disk buffer usage exceeds percentage of a node's available disk