[YARN-8550] YARN root queue exceeds 100%

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.7.3
• Component/s: yarn
• Labels: capacityscheduler
• Assignee: Unassigned
• Reporter: Prabhu Joseph
• Created: 18/Jul/18 20:22
• Updated: 21/Aug/18 08:10

Description

YARN root queue usage is more than 100% which is misleading. (attached screenshot) This happens when there is a container reserved and so used + reserved exceeds Total. Cluster is configured with CPU Scheduling.

2018-07-17 13:27:59,569 INFO  capacity.ParentQueue (ParentQueue.java:assignContainers(475)) - assignedContainer queue=root usedCapacity=0.9713542 absoluteUsedCapacity=0.9713542 used=<memory:763904, vCores:83> cluster=<memory:786432, vCores:384>

2018-07-17 13:27:59,627 INFO  rmcontainer.RMContainerImpl (RMContainerImpl.java:handle(422)) - container_e56_1531419441577_2045_01_000003 Container Transitioned from NEW to RESERVED
2018-07-17 13:27:59,627 INFO  allocator.AbstractContainerAllocator (AbstractContainerAllocator.java:getCSAssignmentFromAllocateResult(77)) - Reserved container  application=application_1531419441577_2045 resource=<memory:53248, vCores:2>

 queue=org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.allocator.RegularContainerAllocator@2a1563f4 cluster=<memory:786432, vCores:384>
2018-07-17 13:27:59,627 INFO  capacity.ParentQueue (ParentQueue.java:assignContainers(475)) - assignedContainer queue=root usedCapacity=1.0390625 absoluteUsedCapacity=1.0390625 used=<memory:817152, vCores:85> cluster=<memory:786432, vCores:384>