Spark（五十）：使用JvisualVM监控Spark Executor JVM

引导

Windows环境下JvisulaVM一般存在于安装了JDK的目录${JAVA_HOME}/bin/JvisualVM.exe，它支持（本地和远程）jstatd和JMX两种方式连接远程JVM。

jstatd (Java Virtual Machine jstat Daemon)——监听远程服务器的CPU，内存，线程等信息

JMX（Java Management Extensions，即Java管理扩展）是一个为应用程序、设备、系统等植入管理功能的框架。JMX可以跨越一系列异构操作系统平台、系统体系结构和网络传输协议，灵活的开发无缝集成的系统、网络和服务管理应用。

备注：针对jstatd我尝试未成功，因此也不在这里误导别人。

JMX监控

正常配置：

-Dcom.sun.management.jmxremote
-Dcom.sun.management.jmxremote.ssl=false
-Dcom.sun.management.jmxremote.authenticate=false
-Djava.rmi.server.hostname=<ip>
-Dcom.sun.management.jmxremote.port=<port>

添加JMX配置：

在Spark中监控executor时，需要先配置jmx然后再启动spark应用程序，配置方式有三种：

1）在spark-defaults.conf中配置那三个参数

2）在spark-env.sh中配置：配置master,worker的JavaOptions

3）在spark-submit提交时配置

这里采用以下spark-submit提交时配置：

spark-submit 
--class myTest.KafkaWordCount 
--master yarn 
--deploy-mode cluster 
--conf "spark.executor.extraJavaOptions=-Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=0 -Dcom.sun.management.jmxremote.authenticate=false -Dcom.sun.management.jmxremote.ssl=false" 
--verbose 
--executor-memory 1G 
--total-executor-cores 6 
/hadoop/spark/app/spark/20151223/testSpark.jar *.*.*.*:* test3 wordcount 4 kafkawordcount3 checkpoint4

注意：

1）不能指定具体的 ip 和 port------因为spark中运行时，很可能一个节点上分配多个container进程，此时占用同一个端口，会导致spark应用程序通过spark-submit提交失败。

2）因为不指定具体的ip和port，所以在任务提交阶段会自动分配端口。

3）上边三种配置方式可能会导致监控级别不同（比如spark-submit只针对一个应用程序，spark-env.sh可能是全局一个节点所有executor监控【未验证】，请读者注意。）

查找JMX分配端口

通过yarn applicationattempt -list appicationId查找到applicationattemptid

[root@cdh-143 bin]# yarn applicationattempt -list application_1559203334026_0015
19/06/01 17:57:18 INFO client.RMProxy: Connecting to ResourceManager at CDH-143/10.dx.dx.143:8032
Total number of application attempts :1
         ApplicationAttempt-Id                 State                        AM-Container-Id                            Tracking-URL
appattempt_1559203334026_0015_000001                 RUNNING    container_1559203334026_0015_01_000001  http://CDH-143:8088/proxy/application_1559203334026_0015/

通过yarn container -list aaplicationattemptId查找container id list

[root@cdh-143 bin]# yarn container -list appattempt_1559203334026_0015_000001
19/06/01 17:57:52 INFO client.RMProxy: Connecting to ResourceManager at CDH-143/10.dx.dx.143:8032
Total number of containers :16
                  Container-Id            Start Time             Finish Time                   State                    Host                                LOG-URL
container_1559203334026_0015_01_000012  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-146:8041    http://CDH-146:8042/node/containerlogs/container_1559203334026_0015_01_000012/dx
container_1559203334026_0015_01_000013  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-146:8041    http://CDH-146:8042/node/containerlogs/container_1559203334026_0015_01_000013/dx
container_1559203334026_0015_01_000010  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-146:8041    http://CDH-146:8042/node/containerlogs/container_1559203334026_0015_01_000010/dx
container_1559203334026_0015_01_000011  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-146:8041    http://CDH-146:8042/node/containerlogs/container_1559203334026_0015_01_000011/dx
container_1559203334026_0015_01_000016  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-146:8041    http://CDH-146:8042/node/containerlogs/container_1559203334026_0015_01_000016/dx
container_1559203334026_0015_01_000014  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-146:8041    http://CDH-146:8042/node/containerlogs/container_1559203334026_0015_01_000014/dx
container_1559203334026_0015_01_000015  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-146:8041    http://CDH-146:8042/node/containerlogs/container_1559203334026_0015_01_000015/dx
container_1559203334026_0015_01_000004  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-142:8041    http://CDH-142:8042/node/containerlogs/container_1559203334026_0015_01_000004/dx
container_1559203334026_0015_01_000005  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-142:8041    http://CDH-142:8042/node/containerlogs/container_1559203334026_0015_01_000005/dx
container_1559203334026_0015_01_000002  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-142:8041    http://CDH-142:8042/node/containerlogs/container_1559203334026_0015_01_000002/dx
container_1559203334026_0015_01_000003  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-142:8041    http://CDH-142:8042/node/containerlogs/container_1559203334026_0015_01_000003/dx
container_1559203334026_0015_01_000008  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-142:8041    http://CDH-142:8042/node/containerlogs/container_1559203334026_0015_01_000008/dx
container_1559203334026_0015_01_000009  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-142:8041    http://CDH-142:8042/node/containerlogs/container_1559203334026_0015_01_000009/dx
container_1559203334026_0015_01_000006  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-142:8041    http://CDH-142:8042/node/containerlogs/container_1559203334026_0015_01_000006/dx
container_1559203334026_0015_01_000007  Sat Jun 01 13:27:52 +0800 2019                   N/A                 RUNNING            CDH-142:8041    http://CDH-142:8042/node/containerlogs/container_1559203334026_0015_01_000007/dx
container_1559203334026_0015_01_000001  Sat Jun 01 13:27:38 +0800 2019                   N/A                 RUNNING            CDH-142:8041    http://CDH-142:8042/node/containerlogs/container_1559203334026_0015_01_000001/dx

到具体executor所在节点服务器上，使用如下命令找到运行的线程，和 pid

[root@cdh-146 ~]# ps -axu | grep container_1559203334026_0015_01_000013
yarn      8844  0.0  0.0 113144  1496 ?        S    13:27   0:00 bash /data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/default_container_executor.sh
yarn      8857  0.0  0.0 113280  1520 ?        Ss   13:27   0:00 /bin/bash -c /usr/java/jdk1.8.0_171-amd64/bin/java -server -Xmx6144m '-Dcom.sun.management.jmxremote' '-Dcom.sun.management.jmxremote.port=0' '-Dcom.sun.management.jmxremote.authenticate=false' '-Dcom.sun.management.jmxremote.ssl=false' -Djava.io.tmpdir=/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/tmp '-Dspark.network.timeout=10000000' '-Dspark.driver.port=47564' '-Dspark.port.maxRetries=32' -Dspark.yarn.app.container.log.dir=/data6/yarn/container-logs/application_1559203334026_0015/container_1559203334026_0015_01_000013 -XX:OnOutOfMemoryError='kill %p' org.apache.spark.executor.CoarseGrainedExecutorBackend --driver-url spark://CoarseGrainedScheduler@CDH-143:47564 --executor-id 12 --hostname CDH-146 --cores 2 --app-id application_1559203334026_0015 --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/__app__.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/streaming-dx-perf-3.0.0.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/dx-common-3.0.0.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/spark-sql-kafka-0-10_2.11-2.4.0.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/spark-avro_2.11-3.2.0.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/shc-core-1.1.2-2.2-s_2.11-SNAPSHOT.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/rocksdbjni-5.17.2.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/kafka-clients-0.10.0.1.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/elasticsearch-spark-20_2.11-6.4.1.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/dx_Spark_State_Store_Plugin-1.0-SNAPSHOT.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/bijection-core_2.11-0.9.5.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/bijection-avro_2.11-0.9.5.jar 1>/data6/yarn/container-logs/application_1559203334026_0015/container_1559203334026_0015_01_000013/stdout 2>/data6/yarn/container-logs/application_1559203334026_0015/container_1559203334026_0015_01_000013/stderr
yarn      9000  143  3.3 8736712 4379648 ?     Sl   13:27  24:35 /usr/java/jdk1.8.0_171-amd64/bin/java -server -Xmx6144m -Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=0 -Dcom.sun.management.jmxremote.authenticate=false -Dcom.sun.management.jmxremote.ssl=false -Djava.io.tmpdir=/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/tmp -Dspark.network.timeout=10000000 -Dspark.driver.port=47564 -Dspark.port.maxRetries=32 -Dspark.yarn.app.container.log.dir=/data6/yarn/container-logs/application_1559203334026_0015/container_1559203334026_0015_01_000013 -XX:OnOutOfMemoryError=kill %p org.apache.spark.executor.CoarseGrainedExecutorBackend --driver-url spark://CoarseGrainedScheduler@CDH-143:47564 --executor-id 12 --hostname CDH-146 --cores 2 --app-id application_1559203334026_0015 --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/__app__.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/dx-domain-perf-3.0.0.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/dx-common-3.0.0.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/spark-sql-kafka-0-10_2.11-2.4.0.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/spark-avro_2.11-3.2.0.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/shc-core-1.1.2-2.2-s_2.11-SNAPSHOT.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/rocksdbjni-5.17.2.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/kafka-clients-0.10.0.1.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/elasticsearch-spark-20_2.11-6.4.1.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/dx_Spark_State_Store_Plugin-1.0-SNAPSHOT.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/bijection-core_2.11-0.9.5.jar --user-class-path file:/data6/yarn/nm/usercache/dx/appcache/application_1559203334026_0015/container_1559203334026_0015_01_000013/bijection-avro_2.11-0.9.5.jar
root     25939  0.0  0.0 112780   956 pts/1    S+   13:45   0:00 grep --color=auto container_1559203334026_0015_01_000013

然后通过 pid 找到对应JMX的端口

[root@cdh-146 ~]# sudo netstat -antp | grep 9000
tcp        0      0 10.dx.dx.146:9000      0.0.0.0:*               LISTEN      2642/python2.7      
tcp6       0      0 :::48169                :::*                    LISTEN      9000/java           
tcp6       0      0 :::37692                :::*                    LISTEN      9000/java           
tcp6       0      0 10.dx.dx.146:52710     :::*                    LISTEN      9000/java           
tcp6       0      0 10.dx.dx.146:55535     10.dx.dx.142:38397     ESTABLISHED 9000/java                
tcp6   64088      0 10.dx.dx.146:45410     10.206.186.35:9092      ESTABLISHED 9000/java           
tcp6       0      0 10.dx.dx.146:60259     10.dx.dx.143:47564     ESTABLISHED 9000/java           

结果中看，疑似为48169或37692，稍微尝试一下即可连上对应的 spark executor

使用JvisulaVM.exe工具添加监控

在本地windows服务器上找到JDK的目录，找到文件${JAVA_HOME}/bin/JvisualVM.exe，并运行它。启动后选择“远程”右键，添加JMX监控

填写监控executor所在节点ip

然后就可以启动监控：