[YUNIKORN-387] Use Tracing to Improve YuniKorn's Observability

• Type: New Feature
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: core - scheduler, shim - kubernetes
• Assignee: Weihao Zheng
• Reporter: Weihao Zheng
• Created: 28/Aug/20 06:50
• Updated: 19/Jan/22 04:52

Description

We can use existing tracing framework to collect tracing information in a standardized format for scheduling and resource management. It will improve YuniKorn's observability significantly with less work. Here are our design ideas: https://docs.google.com/document/d/1MKL9SfTH8Pjw6kBM0vRnyv_ctnxBHAz-iuA7Zbux60E/edit?usp=sharing

Sub-Tasks

• Inject Tracing Logic in the Core's Scheduling Procedures: Open
• Add Tracing's Configuration in YAML File: Open
• Add Documents for Tracing's Usage and Configuration: Open