EMR的fair-scheduler.xml
cat /etc/ecm/hadoop-conf/fair-scheduler.xml
<?xml version="1.0"?>
<allocations>
<aclSubmitApps>*</aclSubmitApps>
<weight>2</weight>
<minResources>10000 mb, 10vcores</minResources>
<maxChildResources>34000 mb,24 vcores</maxChildResources>
<maxRunningApps>50</maxRunningApps>
<maxAMShare>1</maxAMShare>
<maxResources>400000 mb, 200vcores</maxResources> #限制队列最大使用资源
<aclAdministerApps>*</aclAdministerApps>
<schedulingPolicy>fair</schedulingPolicy>
<queue name="default">
<aclSubmitApps>*</aclSubmitApps>
<minResources>10000 mb, 10vcores</minResources>
<aclAdministerApps>*</aclAdministerApps>
<weight>1</weight>
<maxRunningApps>10</maxRunningApps>
<maxAMShare>0.5</maxAMShare>
<maxResources>200000 mb, 100vcores</maxResources>
</queue>
<queue name="collects">
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
<weight>8</weight>
<maxAMShare>0.8</maxAMShare>
<minResources>50 mb, 2vcores</minResources>
<maxResources>400000 mb, 200vcores</maxResources>
<maxRunningApps>50</maxRunningApps>
</queue>
<queue name="data_bi">
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
<weight>4</weight>
<minResources>100 mb, 1vcores</minResources>
<maxResources>30000 mb, 50vcores</maxResources>
<maxRunningApps>5</maxRunningApps>
</queue>
<queue name="opay_collects">
<weight>20</weight>
<minResources>10 mb, 1vcores</minResources>
<maxResources>400000 mb, 200vcores</maxResources>
<maxRunningApps>20</maxRunningApps>
<maxAMShare>0.5</maxAMShare>
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
</queue>
<queue name="opos_collects">
<weight>5</weight>
<minResources>10 mb, 1vcores</minResources>
<maxResources>80000 mb, 50vcores</maxResources>
<maxRunningApps>10</maxRunningApps>
</queue>
<queue name="users" type="parent">
<weight>5</weight>
<minResources>10 mb, 1vcores</minResources>
<maxResources>10000 mb, 150vcores</maxResources>
<maxRunningApps>30</maxRunningApps>
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
</queue>
<queue name="airflow">
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
<weight>8</weight>
<minResources>10 mb, 2vcores</minResources>
<maxResources>200000 mb, 150vcores</maxResources>
<maxRunningApps>30</maxRunningApps>
</queue>
<defaultQueueSchedulingPolicy>fair</defaultQueueSchedulingPolicy>
<userMaxAppsDefault>50</userMaxAppsDefault>
<queueMaxAppsDefault>50</queueMaxAppsDefault>
<queueMaxAMShareDefault>0.5</queueMaxAMShareDefault>
<defaultFairSharePreemptionThreshold>0.5</defaultFairSharePreemptionThreshold>
<queueMaxResourcesDefault>34000 mb,24vcores</queueMaxResourcesDefault>
<defaultFairSharePreemptionTimeout>9223372036854775807</defaultFairSharePreemptionTimeout>
<defaultMinSharePreemptionTimeout>9223372036854775807</defaultMinSharePreemptionTimeout>
</allocations>
#新的xml, 不带root限制: 放emr的yarn-配置-fair-scheduler
<?xml version="1.0" encoding="utf-8"?>
<allocations>
<queue name="root">
<queue name="default">
<aclSubmitApps>*</aclSubmitApps>
<minResources>10000 mb, 10vcores</minResources>
<aclAdministerApps>*</aclAdministerApps>
<weight>1</weight>
<maxRunningApps>10</maxRunningApps>
<maxAMShare>0.5</maxAMShare>
<maxResources>200000 mb, 100vcores</maxResources>
</queue>
<queue name="collects">
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
<weight>8</weight>
<maxAMShare>0.8</maxAMShare>
<minResources>50 mb, 2vcores</minResources>
<maxResources>400000 mb, 200vcores</maxResources>
<maxRunningApps>50</maxRunningApps>
</queue>
<queue name="data_bi">
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
<weight>4</weight>
<minResources>100 mb, 1vcores</minResources>
<maxResources>30000 mb, 50vcores</maxResources>
<maxRunningApps>5</maxRunningApps>
</queue>
<queue name="opay_collects">
<weight>20</weight>
<minResources>10 mb, 1vcores</minResources>
<maxResources>400000 mb, 200vcores</maxResources>
<maxRunningApps>20</maxRunningApps>
<maxAMShare>0.5</maxAMShare>
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
</queue>
<queue name="opos_collects">
<weight>5</weight>
<minResources>10 mb, 1vcores</minResources>
<maxResources>80000 mb, 50vcores</maxResources>
<maxRunningApps>10</maxRunningApps>
</queue>
<queue name="users" type="parent">
<weight>5</weight>
<minResources>10 mb, 1vcores</minResources>
<maxResources>10000 mb, 150vcores</maxResources>
<maxRunningApps>30</maxRunningApps>
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
</queue>
<queue name="airflow">
<schedulingPolicy>fair</schedulingPolicy>
<aclSubmitApps>*</aclSubmitApps>
<aclAdministerApps>*</aclAdministerApps>
<weight>8</weight>
<minResources>10 mb, 2vcores</minResources>
<maxResources>200000 mb, 150vcores</maxResources>
<maxRunningApps>30</maxRunningApps>
</queue>
</queue>
<defaultQueueSchedulingPolicy>fair</defaultQueueSchedulingPolicy>
<userMaxAppsDefault>50</userMaxAppsDefault>
<queueMaxAppsDefault>50</queueMaxAppsDefault>
<queueMaxAMShareDefault>0.5</queueMaxAMShareDefault>
<defaultFairSharePreemptionThreshold>0.5</defaultFairSharePreemptionThreshold>
<queueMaxResourcesDefault>34000 mb,24vcores</queueMaxResourcesDefault>
<defaultFairSharePreemptionTimeout>9223372036854775807</defaultFairSharePreemptionTimeout>
<defaultMinSharePreemptionTimeout>9223372036854775807</defaultMinSharePreemptionTimeout>
</allocations>
EMR的fair-scheduler.xml的更多相关文章
- Fair Scheduler 队列设置经验总结
Fair Scheduler 队列设置经验总结 由于公司的hadoop集群的计算资源不是很充足,需要开启yarn资源队列的资源抢占.在使用过程中,才明白资源抢占的一些特点.在这里总结一下. 只有一个队 ...
- 三：Fair Scheduler 公平调度器
参考资料: http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/FairScheduler.html http://h ...
- Fair Scheduler中的Delay Schedule分析
延迟调度的主要目的是提高数据本地性(data locality),减少数据在网络中的传输.对于那些输入数据不在本地的MapTask,调度器将会延迟调度他们,而把slot分配给那些具备本地性的MapTa ...
- Hadoop学习之--Fair Scheduler作业调度分析
Fair Scheduler调度器同步心跳分配任务的过程简单来讲会经历以下环节: 1. 对map/reduce是否已经达到资源上限的循环判断 2. 对pool队列根据Fair算法排序 3.然后循环po ...
- YARN的Fair Scheduler和Capacity Scheduler
关于Scheduler YARN有四种调度机制:Fair Schedule,Capacity Schedule,FIFO以及Priority: 其中Fair Scheduler是资源池机制,进入到里面 ...
- Hadoop的三种调度器FIFO、Capacity Scheduler、Fair Scheduler（转载）
目前Hadoop有三种比较流行的资源调度器:FIFO .Capacity Scheduler.Fair Scheduler.目前Hadoop2.7默认使用的是Capacity Scheduler容量调 ...
- fair scheduler配置
<property> <name>yarn.resourcemanager.scheduler.class</name> <value>or ...
- Yarn参数优化(Fair Scheduler版本)
YARN 自从hadoop2.0之后, 我们可以使用apache yarn 来对集群资源进行管理.yarn把可以把资源(内存,CPU)以Container的方式进行划分隔离.YARN会管理集群中所有机 ...
- Linux 2.6 完全公平调度算法CFS（Completely Fair Scheduler）
分析
转会http://www.ibm.com/developerworks/cn/linux/l-completely-fair-scheduler/index.html? ca=drs-cn-0125 ...
- 利用yarn capacity scheduler在EMR集群上实现大集群的多租户的集群资源隔离和quota限制
转自:https://m.aliyun.com/yunqi/articles/79700 背景 使用过hadoop的人基本都会考虑集群里面资源的调度和优先级的问题,假设你现在所在的公司有一个大hado ...
随机推荐
- iSCSI引入FC/SAN
由 cxemc 在 2013-9-24 上午9:10 上创建,最后由 cxemc 在 2013-9-24 上午9:10 上修改 版本 1 集成iSCSI 和FC SAN有五种常见的方法,各有优缺,适应 ...
- [转]C++ 类中的static成员的初始化和特点
在C++的类中有些成员变量初始化和一般数据类型的成员变量有所不同.以下测试编译环境为: ➜ g++ -v Using built-in specs. COLLECT_GCC=g++ Target: x ...
- javaee三层架构案例--简单学生管理系统
背景 学了jdbc.jsp等需要串起来,不然会忘记 项目环境 win10 jdk11 mysql8.0.13 jar包 c3p0-0.9.5.2 commons-dbutils-1.7 jstl mc ...
- dell如何安装Win10/Ubuntu双系统
原文:https://www.cnblogs.com/askDing/p/10477345.html 测试环境: DELL PRECISION 7510: CPU:Intel Core i5-6300 ...
- 搭建vue-cli
https://www.cnblogs.com/wisewrong/p/8570309.html https://www.jianshu.com/p/1ee1c410dc67
- MySQL 中视图和表的区别以及联系是什么？
两者的区别: (1)视图是已经编译好的 SQL 语句,是基于 SQL 语句的结果集的可视化的表,而表不是. (2)视图没有实际的物理记录,而基本表有. (3)表是内容,视图是窗口. (4)表占用物理空 ...
- oracle利用触发器实现主键字段自增
我们都知道oracle主键自增利用的是序列sequence.我们先创建一个sequence: create sequence test_sequence start increment maxvalu ...
- React 高阶组件浅析
高阶组件的这种写法的诞生来自于社区的实践,目的是解决一些交叉问题(Cross-Cutting Concerns).而最早时候 React 官方给出的解决方案是使用 mixin .而 React 也在官 ...
- GoogleNet-ILSVRC-2014冠军
Going deeper with convolutions-22层 https://my.oschina.net/u/876354/blog/1637819 那么,GoogLeNet是如何进一步提升 ...
- It’s worth noting值得注意的是
It’s worth noting that in JavaScript applications the Model is often connected via Ajax to a back-en ...