How to connect with JMX remotely to Spark worker on Dataproc

问题

I can connect to the driver just fine by adding the following:

spark.driver.extraJavaOptions=-Dcom.sun.management.jmxremote \
                              -Dcom.sun.management.jmxremote.port=9178 \
                              -Dcom.sun.management.jmxremote.authenticate=false \
                              -Dcom.sun.management.jmxremote.ssl=false

But doing ...

spark.executor.extraJavaOptions=-Dcom.sun.management.jmxremote \
                                -Dcom.sun.management.jmxremote.port=9178 \
                                -Dcom.sun.management.jmxremote.authenticate=false \
                                -Dcom.sun.management.jmxremote.ssl=false

... only yield a bunch of errors on the driver ...

Container id: container_1501548048292_0024_01_000003
Exit code: 1
Stack trace: ExitCodeException exitCode=1: 
    at org.apache.hadoop.util.Shell.runCommand(Shell.java:972)
    at org.apache.hadoop.util.Shell.run(Shell.java:869)
    at org.apache.hadoop.util.Shell$ShellCommandExecutor.execute(Shell.java:1170)
    at org.apache.hadoop.yarn.server.nodemanager.DefaultContainerExecutor.launchContainer(DefaultContainerExecutor.java:236)
    at org.apache.hadoop.yarn.server.nodemanager.containermanager.launcher.ContainerLaunch.call(ContainerLaunch.java:305)
    at org.apache.hadoop.yarn.server.nodemanager.containermanager.launcher.ContainerLaunch.call(ContainerLaunch.java:84)
    at java.util.concurrent.FutureTask.run(FutureTask.java:266)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
    at java.lang.Thread.run(Thread.java:748)


Container exited with a non-zero exit code 1

... and finally crashes the job.

There are no errors on the workers, it simply exits with:

[org.apache.spark.util.ShutdownHookManager] - Shutdown hook called

Spark v2.2.0, and the cluster is a simple 1m-2w-configuration, and my jobs run without issues without the executor parameters.

回答1:

As Rick Mortiz pointed out, the issue was conflicting ports for the executor jmx.

Setting:

-Dcom.sun.management.jmxremote.port=0

yields a random port, and removed the errors from Spark. To figure out which port it ends up using do:

netstat -alp | grep LISTEN.*<executor-pid>/java

which lists the currently open ports for that process.

回答2:

Passing following configuration with spark-submit worked for me --conf "spark.executor.extraJavaOptions=-Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=9178 -Dcom.sun.management.jmxremote.authenticate=false -Dcom.sun.management.jmxremote.ssl=false"

来源：https://stackoverflow.com/questions/45435237/how-to-connect-with-jmx-remotely-to-spark-worker-on-dataproc

标签

apache-spark

yarn

google-cloud-dataproc