[YARN-10690] GPU related improvement for better usage.

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Target Version/s: 3.5.0
• Assignee: Qi Zhu
• Reporter: Qi Zhu
• Created: 12/Mar/21 09:51
• Updated: 04/Jan/24 08:31

Description

This Jira will improve GPU for better usage.

 cc bibinchundatt pbacsko ebadger ztang  epayne gandras  bteke

Sub-Tasks

• Support queue capacity in terms of absolute resources with custom resourceType.: Resolved
• ClusterMetrics should support GPU capacity related metrics.: Resolved
• ClusterMetrics should support custom resource capacity related metrics.: Resolved
• Add Node GPU Utilization and apply to NodeMetrics.: Resolved
• Support custom resources in ResourceUtilization, and update Node GPU Utilization to use.: Resolved
• Nodemanagers cannot detect GPU failures: Open
• The CS effective capacity for absolute mode in UI should support GPU and other custom resources.: Patch Available
• Remove hardcoded resource values (e.g. GPU/FPGA) in code.: Resolved
• Change CS nodes page in UI to support custom resource.: Resolved