概述
YARN命令是调用bin/yarn脚本文件,如果运行yarn脚本没有带任何参数,则会打印yarn所有命令的描述。
使用: yarn [–config confdir] COMMAND [–loglevel loglevel] [GENERIC_OPTIONS] [COMMAND_OPTIONS]
YARN有一个参数解析框架,采用解析泛型参数以及运行类。
命令参数
描述
--config confdir
指定一个默认的配置文件目录,默认值是: ${HADOOP_PREFIX}/conf
.
--loglevel loglevel
重载Log级别。有效的日志级别包含:FATAL, ERROR, WARN, INFO, DEBUG, and TRACE。默认是INFO。
GENERIC_OPTIONS
YARN支持表A的通用命令项。
COMMAND COMMAND_OPTIONS
YARN分为用户命令和管理员命令。
表A:
通用项
Description
-archives <comma separated list of archives>
用逗号分隔计算中未归档的文件。 仅仅针对JOB。
-conf <configuration file>
制定应用程序的配置文件。
-D <property>=<value>
使用给定的属性值。
-files <comma separated list of files>
用逗号分隔的文件,拷贝到Map reduce机器,仅仅针对JOB
-jt <local> or <resourcemanager:port>
指定一个ResourceManager. 仅仅针对JOB。
-libjars <comma seperated list of jars>
将用逗号分隔的jar路径包含到classpath中去,仅仅针对JOB。
用户命令:
对于Hadoop集群用户很有用的命令:
application
使用:yarn application [options]
命令选项
描述
-appStates
使用-list命令,基于应用程序的状态来过滤应用程序。如果应用程序的状态有多个,用逗号分隔。 有效的应用程序状态包含
如下: ALL, NEW, NEW_SAVING, SUBMITTED, ACCEPTED, RUNNING, FINISHED, FAILED, KILLED
-appTypes
使用-list命令,基于应用程序类型来过滤应用程序。如果应用程序的类型有多个,用逗号分隔。
-list
从RM返回的应用程序列表,使用-appTypes参数,支持基于应用程序类型的过滤,使用-appStates参数,支持对应用程序状态的过滤。
-kill
kill掉指定的应用程序。
-status
打印应用程序的状态。
示例1:
[hduser@hadoop0 bin]$ ./yarn application -list -appStates ACCEPTED
15/08/10 11:48:43 INFO client.RMProxy: Connecting to ResourceManager at hadoop1/10.0.1.41:8032
Total number of applications (application-types: [] and states: [ACCEPTED]):1
Application-Id Application-Name Application-Type User Queue State Final-State Progress Tracking-URL
application_1438998625140_1703 MAC_STATUS MAPREDUCE hduser default ACCEPTED UNDEFINED 0% N/A
示例2:
[hduser@hadoop0 bin]$ ./yarn application -list
15/08/10 11:43:01 INFO client.RMProxy: Connecting to ResourceManager at hadoop1/10.0.1.41:8032
Total number of applications (application-types: [] and states: [SUBMITTED, ACCEPTED, RUNNING]):1
Application-Id Application-Name Application-Type User Queue State Final-State Progress Tracking-URL
application_1438998625140_1701 MAC_STATUS MAPREDUCE hduser default ACCEPTED UNDEFINED 0% N/A
示例3:
[hduser@hadoop0 bin]$ ./yarn application -kill application_1438998625140_1705
15/08/10 11:57:41 INFO client.RMProxy: Connecting to ResourceManager at hadoop1/10.0.1.41:8032
Killing application application_1438998625140_1705
15/08/10 11:57:42 INFO impl.YarnClientImpl: Killed application application_1438998625140_1705
applicationattempt
使用:yarn applicationattempt [options]
命令选项
描述
-help
帮助
-list
获取到应用程序尝试的列表,其返回值ApplicationAttempt-Id 等于
-status
打印应用程序尝试的状态。
打印应用程序尝试的报告。
示例1:
[hadoop@hadoopcluster78 bin]$ yarn applicationattempt -list application_1437364567082_0106
15/08/10 20:58:28 INFO client.RMProxy: Connecting to ResourceManager at hadoopcluster79/10.0.1.79:8032
Total number of application attempts :1
ApplicationAttempt-Id State AM-Container-Id Tracking-URL
appattempt_1437364567082_0106_000001 RUNNING container_1437364567082_0106_01_000001 http://hadoopcluster79:8088/proxy/application_1437364567082_0106/
示例2:
[hadoop@hadoopcluster78 bin]$ yarn applicationattempt -status appattempt_1437364567082_0106_000001
15/08/10 21:01:41 INFO client.RMProxy: Connecting to ResourceManager at hadoopcluster79/10.0.1.79:8032
Application Attempt Report :
ApplicationAttempt-Id : appattempt_1437364567082_0106_000001
State : FINISHED
AMContainer : container_1437364567082_0106_01_000001
Tracking-URL : http://hadoopcluster79:8088/proxy/application_1437364567082_0106/jobhistory/job/job_1437364567082_0106
RPC Port : 51911
AM Host : hadoopcluster80
Diagnostics :
classpath
使用: yarn classpath
打印需要得到Hadoop的jar和所需要的lib包路径
[hadoop@hadoopcluster78 bin]$ yarn classpath
/home/hadoop/apache/hadoop-2.4.1/etc/hadoop:/home/hadoop/apache/hadoop-2.4.1/etc/hadoop:/home/hadoop/apache/hadoop-2.4.1/etc/hadoop:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/common/lib/*:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/common/*:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/hdfs:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/hdfs/lib/*:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/hdfs/*:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/yarn/lib/*:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/yarn/*:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/mapreduce/lib/*:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/mapreduce/*:/home/hadoop/apache/hadoop-2.4.1/contrib/capacity-scheduler/*.jar:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/yarn/*:/home/hadoop/apache/hadoop-2.4.1/share/hadoop/yarn/lib/*
container
使用: yarn container [options]
命令选项
描述
-help
帮助
-list
应用程序尝试的Containers列表
-status
打印Container的状态
打印container(s)的报告
示例1:
[hadoop@hadoopcluster78 bin]$ yarn container -list appattempt_1437364567082_0106_01
15/08/10 20:45:45 INFO client.RMProxy: Connecting to ResourceManager at hadoopcluster79/10.0.1.79:8032
Total number of containers :25
Container-Id Start Time Finish Time State Host LOG-URL
container_1437364567082_0106_01_000028 1439210458659 0 RUNNING hadoopcluster83:37140 //hadoopcluster83:8042/node/containerlogs/container_1437364567082_0106_01_000028/hadoop
container_1437364567082_0106_01_000016 1439210314436 0 RUNNING hadoopcluster84:43818 //hadoopcluster84:8042/node/containerlogs/container_1437364567082_0106_01_000016/hadoop
container_1437364567082_0106_01_000019 1439210338598 0 RUNNING hadoopcluster83:37140 //hadoopcluster83:8042/node/containerlogs/container_1437364567082_0106_01_000019/hadoop
container_1437364567082_0106_01_000004 1439210314130 0 RUNNING hadoopcluster82:48622 //hadoopcluster82:8042/node/containerlogs/container_1437364567082_0106_01_000004/hadoop
container_1437364567082_0106_01_000008 1439210314130 0 RUNNING hadoopcluster82:48622 //hadoopcluster82:8042/node/containerlogs/container_1437364567082_0106_01_000008/hadoop
container_1437364567082_0106_01_000031 1439210718604 0 RUNNING hadoopcluster83:37140 //hadoopcluster83:8042/node/containerlogs/container_1437364567082_0106_01_000031/hadoop
container_1437364567082_0106_01_000020 1439210339601 0 RUNNING hadoopcluster83:37140 //hadoopcluster83:8042/node/containerlogs/container_1437364567082_0106_01_000020/hadoop
container_1437364567082_0106_01_000005 1439210314130 0 RUNNING hadoopcluster82:48622 //hadoopcluster82:8042/node/containerlogs/container_1437364567082_0106_01_000005/hadoop
container_1437364567082_0106_01_000013 1439210314435 0 RUNNING hadoopcluster84:43818 //hadoopcluster84:8042/node/containerlogs/container_1437364567082_0106_01_000013/hadoop
container_1437364567082_0106_01_000022 1439210368679 0 RUNNING hadoopcluster84:43818 //hadoopcluster84:8042/node/containerlogs/container_1437364567082_0106_01_000022/hadoop
container_1437364567082_0106_01_000021 1439210353626 0 RUNNING hadoopcluster83:37140 //hadoopcluster83:8042/node/containerlogs/container_1437364567082_0106_01_000021/hadoop
container_1437364567082_0106_01_000014 1439210314435 0 RUNNING hadoopcluster84:43818 //hadoopcluster84:8042/node/containerlogs/container_1437364567082_0106_01_000014/hadoop
container_1437364567082_0106_01_000029 1439210473726 0 RUNNING hadoopcluster80:42366 //hadoopcluster80:8042/node/containerlogs/container_1437364567082_0106_01_000029/hadoop
container_1437364567082_0106_01_000006 1439210314130 0 RUNNING hadoopcluster82:48622 //hadoopcluster82:8042/node/containerlogs/container_1437364567082_0106_01_000006/hadoop
container_1437364567082_0106_01_000003 1439210314129 0 RUNNING hadoopcluster82:48622 //hadoopcluster82:8042/node/containerlogs/container_1437364567082_0106_01_000003/hadoop
container_1437364567082_0106_01_000015 1439210314436 0 RUNNING hadoopcluster84:43818 //hadoopcluster84:8042/node/containerlogs/container_1437364567082_0106_01_000015/hadoop
container_1437364567082_0106_01_000009 1439210314130 0 RUNNING hadoopcluster82:48622 //hadoopcluster82:8042/node/containerlogs/container_1437364567082_0106_01_000009/hadoop
container_1437364567082_0106_01_000030 1439210708467 0 RUNNING hadoopcluster83:37140 //hadoopcluster83:8042/node/containerlogs/container_1437364567082_0106_01_000030/hadoop
container_1437364567082_0106_01_000012 1439210314435 0 RUNNING hadoopcluster84:43818 //hadoopcluster84:8042/node/containerlogs/container_1437364567082_0106_01_000012/hadoop
container_1437364567082_0106_01_000027 1439210444354 0 RUNNING hadoopcluster84:43818 //hadoopcluster84:8042/node/containerlogs/container_1437364567082_0106_01_000027/hadoop
container_1437364567082_0106_01_000026 1439210428514 0 RUNNING hadoopcluster83:37140 //hadoopcluster83:8042/node/containerlogs/container_1437364567082_0106_01_000026/hadoop
container_1437364567082_0106_01_000017 1439210314436 0 RUNNING hadoopcluster84:43818 //hadoopcluster84:8042/node/containerlogs/container_1437364567082_0106_01_000017/hadoop
container_1437364567082_0106_01_000001 1439210306902 0 RUNNING hadoopcluster80:42366 //hadoopcluster80:8042/node/containerlogs/container_1437364567082_0106_01_000001/hadoop
container_1437364567082_0106_01_000002 1439210314129 0 RUNNING hadoopcluster82:48622 //hadoopcluster82:8042/node/containerlogs/container_1437364567082_0106_01_000002/hadoop
container_1437364567082_0106_01_000025 1439210414171 0 RUNNING hadoopcluster83:37140 //hadoopcluster83:8042/node/containerlogs/container_1437364567082_0106_01_000025/hadoop
示例2:
[hadoop@hadoopcluster78 bin]$ yarn container -status container_1437364567082_0105_01_000020
15/08/10 20:28:00 INFO client.RMProxy: Connecting to ResourceManager at hadoopcluster79/10.0.1.79:8032
Container Report :
Container-Id : container_1437364567082_0105_01_000020
Start-Time : 1439208779842
Finish-Time : 0
State : RUNNING
LOG-URL : //hadoopcluster83:8042/node/containerlogs/container_1437364567082_0105_01_000020/hadoop
Host : hadoopcluster83:37140
Diagnostics : null
jar
使用: yarn jar [mainClass] args…
运行jar文件,用户可以将写好的YARN代码打包成jar文件,用这个命令去运行它。
logs
使用: yarn logs -applicationId [options]
注:应用程序没有完成,该命令是不能打印日志的。
命令选项
描述
-applicationId
指定应用程序ID,应用程序的ID可以在yarn.resourcemanager.webapp.address配置的路径查看(即:ID)
-appOwner
应用的所有者(如果没有指定就是当前用户)应用程序的ID可以在yarn.resourcemanager.webapp.address配置的路径查看(即:User)
-containerId
Container Id
-help
帮助
-nodeAddress
节点地址的格式:nodename:port (端口是配置文件中:yarn.nodemanager.webapp.address参数指定)
转存container的日志。
示例:
[hadoop@hadoopcluster78 bin]$ yarn logs -applicationId application_1437364567082_0104 -appOwner hadoop
15/08/10 17:59:19 INFO client.RMProxy: Connecting to ResourceManager at hadoopcluster79/10.0.1.79:8032
Container: container_1437364567082_0104_01_000003 on hadoopcluster82_48622
============================================================================
LogType: stderr
LogLength: 0
Log Contents:
LogType: stdout
LogLength: 0
Log Contents:
LogType: syslog
LogLength: 3673
Log Contents:
2015-08-10 17:24:01,565 WARN [main] org.apache.hadoop.conf.Configuration: job.xml:an attempt to override final parameter: mapreduce.job.end-notification.max.retry.interval; Ignoring.
2015-08-10 17:24:01,580 WARN [main] org.apache.hadoop.conf.Configuration: job.xml:an attempt to override final parameter: mapreduce.job.end-notification.max.attempts; Ignoring.
。。。。。。此处省略N万个字符
// 下面的命令,根据APP的所有者查看LOG日志,因为application_1437364567082_0104任务我是用hadoop用户启动的,所以打印的是如下信息:
[hadoop@hadoopcluster78 bin]$ yarn logs -applicationId application_1437364567082_0104 -appOwner root
15/08/10 17:59:25 INFO client.RMProxy: Connecting to ResourceManager at hadoopcluster79/10.0.1.79:8032
Logs not available at /tmp/logs/root/logs/application_1437364567082_0104
Log aggregation has not completed or is not enabled.
node
使用: yarn node [options]
命令选项
描述
-all
所有的节点,不管是什么状态的。
-list
列出所有RUNNING状态的节点。支持-states选项过滤指定的状态,节点的状态包
含:NEW,RUNNING,UNHEALTHY,DECOMMISSIONED,LOST,REBOOTED。支持–all显示所有的节点。
-states
和-list配合使用,用逗号分隔节点状态,只显示这些状态的节点信息。
-status
打印指定节点的状态。
示例1:
[hadoop@hadoopcluster78 bin]$ ./yarn node -list -all
15/08/10 17:34:17 INFO client.RMProxy: Connecting to ResourceManager at hadoopcluster79/10.0.1.79:8032
Total Nodes:4
Node-Id Node-State Node-Http-Address Number-of-Running-Containers
hadoopcluster82:48622 RUNNING hadoopcluster82:8042 0
hadoopcluster84:43818 RUNNING hadoopcluster84:8042 0
hadoopcluster83:37140 RUNNING hadoopcluster83:8042 0
hadoopcluster80:42366 RUNNING hadoopcluster80:8042 0
示例2:
[hadoop@hadoopcluster78 bin]$ ./yarn node -list -states RUNNING
15/08/10 17:39:55 INFO client.RMProxy: Connecting to ResourceManager at hadoopcluster79/10.0.1.79:8032
Total Nodes:4
Node-Id Node-State Node-Http-Address Number-of-Running-Containers
hadoopcluster82:48622 RUNNING hadoopcluster82:8042 0
hadoopcluster84:43818 RUNNING hadoopcluster84:8042 0
hadoopcluster83:37140 RUNNING hadoopcluster83:8042 0
hadoopcluster80:42366 RUNNING hadoopcluster80:8042 0
示例3:
[hadoop@hadoopcluster78 bin]$ ./yarn node -status hadoopcluster82:48622
15/08/10 17:52:52 INFO client.RMProxy: Connecting to ResourceManager at hadoopcluster79/10.0.1.79:8032
Node Report :
Node-Id : hadoopcluster82:48622
Rack : /default-rack
Node-State : RUNNING
Node-Http-Address : hadoopcluster82:8042
Last-Health-Update : 星期一 10/八月/15 05:52:09:601CST
Health-Report :
Containers : 0
Memory-Used : 0MB
Memory-Capacity : 10240MB
CPU-Used : 0 vcores
CPU-Capacity : 8 vcores
打印节点的报告。
queue
使用: yarn queue [options]
命令选项
描述
-help
帮助
-status
打印队列的状态
打印队列信息。
version
使用: yarn version
打印hadoop的版本。
管理员命令:
下列这些命令对hadoop集群的管理员是非常有用的。
daemonlog
使用:
yarn daemonlog -getlevel <host:httpport> <classname>
yarn daemonlog -setlevel <host:httpport> <classname> <level>
参数选项
描述
-getlevel host:httpport
打印运行在host:port的守护进程的日志级别。这个命令内部会连接http://host:port/logLevellog=
-setlevel host:httpport
设置运行在host:port的守护进程的日志级别。这个命令内部会连接http://host:port/logLevellog=
针对指定的守护进程,获取/设置日志级别.
示例1:
[root@hadoopcluster78 ~]# hadoop daemonlog -getlevel hadoopcluster82:50075 org.apache.hadoop.hdfs.server.datanode.DataNode
Connecting to http://hadoopcluster82:50075/logLevel?log=org.apache.hadoop.hdfs.server.datanode.DataNode
Submitted Log Name: org.apache.hadoop.hdfs.server.datanode.DataNode
Log Class: org.apache.commons.logging.impl.Log4JLogger
Effective level: INFO
[root@hadoopcluster78 ~]# yarn daemonlog -getlevel hadoopcluster79:8088 org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppImpl
Connecting to http://hadoopcluster79:8088/logLevel?log=org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppImpl
Submitted Log Name: org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppImpl
Log Class: org.apache.commons.logging.impl.Log4JLogger
Effective level: INFO
[root@hadoopcluster78 ~]# yarn daemonlog -getlevel hadoopcluster78:19888 org.apache.hadoop.mapreduce.v2.hs.JobHistory
Connecting to http://hadoopcluster78:19888/logLevel?log=org.apache.hadoop.mapreduce.v2.hs.JobHistory
Submitted Log Name: org.apache.hadoop.mapreduce.v2.hs.JobHistory
Log Class: org.apache.commons.logging.impl.Log4JLogger
Effective level: INFO
nodemanager
使用: yarn nodemanager
启动NodeManager
proxyserver
使用: yarn proxyserver
启动web proxy server
resourcemanager
使用: yarn resourcemanager [-format-state-store]
参数选项
描述
-format-state-store
RMStateStore的格式. 如果过去的应用程序不再需要,则清理RMStateStore, RMStateStore仅仅在ResourceManager没有运行的时候,才运行RMStateStore
启动ResourceManager
rmadmin
使用:
yarn rmadmin [-refreshQueues]
[-refreshNodes]
[-refreshUserToGroupsMapping]
[-refreshSuperUserGroupsConfiguration]
[-refreshAdminAcls]
[-refreshServiceAcl]
[-getGroups [username]]
[-transitionToActive [--forceactive] [--forcemanual] <serviceId>]
[-transitionToStandby [--forcemanual] <serviceId>]
[-failover [--forcefence] [--forceactive] <serviceId1> <serviceId2>]
[-getServiceState <serviceId>]
[-checkHealth <serviceId>]
[-help [cmd]]
参数选项
描述
-refreshQueues
重载队列的ACL,状态和调度器特定的属性,ResourceManager将重载mapred-queues配置文件
-refreshNodes
动态刷新dfs.hosts和dfs.hosts.exclude配置,无需重启NameNode。
dfs.hosts:列出了允许连入NameNode的datanode清单(IP或者机器名)
dfs.hosts.exclude:列出了禁止连入NameNode的datanode清单(IP或者机器名)
重新读取hosts和exclude文件,更新允许连到Namenode的或那些需要退出或入编的Datanode的集合。
-refreshUserToGroupsMappings
刷新用户到组的映射。
-refreshSuperUserGroupsConfiguration
刷新用户组的配置
-refreshAdminAcls
刷新ResourceManager的ACL管理
-refreshServiceAcl
ResourceManager重载服务级别的授权文件。
-getGroups [username]
获取指定用户所属的组。
-transitionToActive [–forceactive] [–forcemanual]
尝试将目标服务转为 Active 状态。如果使用了–forceactive选项,不需要核对非Active节点。如果采用了自动故障转移,这个命令不能使用。虽然你可以重写–forcemanual选项,你需要谨慎。
-transitionToStandby [–forcemanual]
将服务转为 Standby 状态. 如果采用了自动故障转移,这个命令不能使用。虽然你可以重写–forcemanual选项,你需要谨慎。
-failover [–forceactive]
启动从serviceId1 到 serviceId2的故障转移。如果使用了-forceactive选项,即使服务没有准备,也会尝试故障转移到目标服务。如果采用了自动故障转移,这个命令不能使用。
-getServiceState
返回服务的状态。(注:ResourceManager不是HA的时候,时不能运行该命令的)
-checkHealth
请求服务器执行健康检查,如果检查失败,RMAdmin将用一个非零标示退出。(注:ResourceManager不是HA的时候,时不能运行该命令的)
-help [cmd]
显示指定命令的帮助,如果没有指定,则显示命令的帮助。
scmadmin
使用: yarn scmadmin [options]
参数选项
描述
-help
Help
-runCleanerTask
Runs the cleaner task
Runs Shared Cache Manager admin client
sharedcachemanager
使用: yarn sharedcachemanager
启动Shared Cache Manager
timelineserver
之前yarn运行框架只有Job history server,这是hadoop2.4版本之后加的通用Job History Server,命令为Application Timeline Server,详情请看:The YARN Timeline Server
使用: yarn timelineserver
启动TimeLineServer