[YARN-10483] yarn hang住卡死，任务无法提交，切换RM主节点或重启才能恢复

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Information Provided
• Affects Version/s: 3.1.1
• Component/s: capacity scheduler, capacityscheduler, resourcemanager, RM
• Release Note: Please create Jiras that makes it easy for other developers to search and understand.
• Assignee: Unassigned
• Reporter: jufeng li
• Created: 05/Nov/20 02:33
• Updated: 05/Nov/20 18:22
• Resolved: 05/Nov/20 17:09

Description

yarn不定期卡死，新任务无法提交，经排查jstack日志，capacity scheduler有线程在无限等待锁，rm的cpu内存网络磁盘均正常。问题基本可以确定是capacity scheduler内部的锁出了问题。正常状态下和卡住状态下rm的jstack日志已上传，希望有人可以解决一下，此bug比较严重，直接导致生产不可用。没人解答待会我再来问