Spark on yarn resource optimization ideas notes

yarn.nodemanager.resource.memory-mb 
// Every NodeManager You can supply yarn Dispatch （ Assigned to container） Physical memory , Company MB

yarn.nodemanager.resource.cpu-vcores 
// Every NodeManager You can supply yarn Dispatch （ Assigned to container） Of vcore Number 
 
yarn.scheduler.maximum-allocation-mb 
// Every container The maximum memory that can be applied 

yarn.scheduler.minimum-allocation-mb 
// Every container The minimum memory that can be applied , If the value set is smaller than this value , Default is the value 

yarn.scheduler.increment-allocation-mb 
//container How much memory is added at a time when the memory is not enough   Company MB. Default 512M

yarn.scheduler.minimum-allocation-vcores 
// Every container The smallest you can apply for vcore Number , If the value set is smaller than this value , Default is the value 

yarn.scheduler.maximum-allocation-vcores 
// Every container The biggest... You can apply for vcore Number .
 
yarn.nodemanager.pmem-check-enabled 
// Whether the contanier Implement physical memory limits , It will be monitored through a thread container Memory usage , More than the container After the memory limit of , Will be kill fall .

yarn.nodemanager.vmem-check-enabled 
// Whether the container Implement virtual memory limits