Spark on Yarn job failed with ExitCode:1 and stderr says "Can't find main class"

Question

We tried to submit a simple SparkPI example onto Spark on Yarn. The bat is written as below:

./bin/spark-submit --class org.apache.spark.examples.SparkPi --master yarn-cluster --num-executors 3 --driver-memory 4g --executor-memory 1g --executor-cores 1 .\examples\target\spark-examples_2.10-1.4.0.jar 10
pause

Our HDFS and Yarn works well. We are using Hadoop 2.7.0 and Spark 1.4.1. We have only 1 node that acts as both NameNode and DataNode.

When we execute it, it fails with log says the following:

2015-08-21 11:07:22,044 DEBUG [main] | ===============================================================================
2015-08-21 11:07:22,044 DEBUG [main] | Yarn AM launch context:
2015-08-21 11:07:22,044 DEBUG [main] |     user class: org.apache.spark.examples.SparkPi
2015-08-21 11:07:22,044 DEBUG [main] |     env:
2015-08-21 11:07:22,044 DEBUG [main] |         CLASSPATH -> {{PWD}}<CPS>{{PWD}}/__hadoop_conf__<CPS>{{PWD}}/__spark__.jar<CPS>%HADOOP_HOME%\etc\hadoop<CPS>%HADOOP_HOME%\share\hadoop\common\*<CPS>%HADOOP_HOME%\share\hadoop\common\lib\*<CPS>%HADOOP_HOME%\share\hadoop\mapreduce\*<CPS>%HADOOP_HOME%\share\hadoop\mapreduce\lib\*<CPS>%HADOOP_HOME%\share\hadoop\hdfs\*<CPS>%HADOOP_HOME%\share\hadoop\hdfs\lib\*<CPS>%HADOOP_HOME%\share\hadoop\yarn\*<CPS>%HADOOP_HOME%\share\hadoop\yarn\lib\*<CPS>%HADOOP_MAPRED_HOME%\share\hadoop\mapreduce\*<CPS>%HADOOP_MAPRED_HOME%\share\hadoop\mapreduce\lib\*
2015-08-21 11:07:22,060 DEBUG [main] |         SPARK_YARN_CACHE_FILES_FILE_SIZES -> 165181064,1420218
2015-08-21 11:07:22,060 DEBUG [main] |         SPARK_YARN_STAGING_DIR -> .sparkStaging/application_1440062075415_0026
2015-08-21 11:07:22,060 DEBUG [main] |         SPARK_YARN_CACHE_FILES_VISIBILITIES -> PRIVATE,PRIVATE
2015-08-21 11:07:22,060 DEBUG [main] |         SPARK_USER -> msrabi
2015-08-21 11:07:22,060 DEBUG [main] |         SPARK_YARN_MODE -> true
2015-08-21 11:07:22,060 DEBUG [main] |         SPARK_YARN_CACHE_FILES_TIME_STAMPS -> 1440126441200,1440126441575
2015-08-21 11:07:22,060 DEBUG [main] |         SPARK_YARN_CACHE_FILES -> hdfs://msra-sa-44:9000/user/msrabi/.sparkStaging/application_1440062075415_0026/spark-assembly-1.4.0-hadoop2.7.0.jar#__spark__.jar,hdfs://msra-sa-44:9000/user/msrabi/.sparkStaging/application_1440062075415_0026/spark-examples_2.10-1.4.0.jar#__app__.jar
2015-08-21 11:07:22,060 DEBUG [main] |     resources:
2015-08-21 11:07:22,060 DEBUG [main] |         __app__.jar -> resource { scheme: "hdfs" host: "msra-sa-44" port: 9000 file: "/user/msrabi/.sparkStaging/application_1440062075415_0026/spark-examples_2.10-1.4.0.jar" } size: 1420218 timestamp: 1440126441575 type: FILE visibility: PRIVATE
2015-08-21 11:07:22,060 DEBUG [main] |         __spark__.jar -> resource { scheme: "hdfs" host: "msra-sa-44" port: 9000 file: "/user/msrabi/.sparkStaging/application_1440062075415_0026/spark-assembly-1.4.0-hadoop2.7.0.jar" } size: 165181064 timestamp: 1440126441200 type: FILE visibility: PRIVATE
2015-08-21 11:07:22,060 DEBUG [main] |         __hadoop_conf__ -> resource { scheme: "hdfs" host: "msra-sa-44" port: 9000 file: "/user/msrabi/.sparkStaging/application_1440062075415_0026/__hadoop_conf__7908628615251032149.zip" } size: 82888 timestamp: 1440126441794 type: ARCHIVE visibility: PRIVATE
2015-08-21 11:07:22,060 DEBUG [main] |     command:
2015-08-21 11:07:22,075 DEBUG [main] |         {{JAVA_HOME}}/bin/java -server -Xmx4096m -Djava.io.tmpdir={{PWD}}/tmp '-Dspark.app.name=org.apache.spark.examples.SparkPi' '-Dspark.executor.memory=1g' '-Dspark.driver.memory=4g' '-Dspark.master=yarn-cluster' -Dspark.yarn.app.container.log.dir=<LOG_DIR> org.apache.spark.deploy.yarn.ApplicationMaster --class 'org.apache.spark.examples.SparkPi' --jar file:/D:/sp/./examples/target/spark-examples_2.10-1.4.0.jar --arg '10' --executor-memory 1024m --executor-cores 1 --num-executors  3 1> <LOG_DIR>/stdout 2> <LOG_DIR>/stderr
2015-08-21 11:07:22,075 DEBUG [main] | ===============================================================================

...........(omitting some lines)......

2015-08-21 11:07:23,231 INFO [main] | Application report for application_1440062075415_0026 (state: ACCEPTED)
2015-08-21 11:07:23,247 DEBUG [main] | 
     client token: N/A
     diagnostics: N/A
     ApplicationMaster host: N/A
     ApplicationMaster RPC port: -1
     queue: default
     start time: 1440126442169
     final status: UNDEFINED
     tracking URL: http://msra-sa-44:8088/proxy/application_1440062075415_0026/
     user: msrabi
2015-08-21 11:07:24,263 TRACE [main] | 1: Call -> MSRA-SA-44/10.190.173.181:8032: getApplicationReport {application_id { id: 26 cluster_timestamp: 1440062075415 }}
2015-08-21 11:07:24,263 DEBUG [IPC Parameter Sending Thread #0] | IPC Client (443384617) connection to MSRA-SA-44/10.190.173.181:8032 from msrabi sending #37
2015-08-21 11:07:24,263 DEBUG [IPC Client (443384617) connection to MSRA-SA-44/10.190.173.181:8032 from msrabi] | IPC Client (443384617) connection to MSRA-SA-44/10.190.173.181:8032 from msrabi got value #37
2015-08-21 11:07:24,263 DEBUG [main] | Call: getApplicationReport took 0ms
2015-08-21 11:07:24,263 TRACE [main] | 1: Response <- MSRA-SA-44/10.190.173.181:8032: getApplicationReport {application_report { applicationId { id: 26 cluster_timestamp: 1440062075415 } user: "msrabi" queue: "default" name: "org.apache.spark.examples.SparkPi" host: "N/A" rpc_port: -1 yarn_application_state: ACCEPTED trackingUrl: "http://msra-sa-44:8088/proxy/application_1440062075415_0026/" diagnostics: "" startTime: 1440126442169 finishTime: 0 final_application_status: APP_UNDEFINED app_resource_Usage { num_used_containers: 1 num_reserved_containers: 0 used_resources { memory: 4608 virtual_cores: 1 } reserved_resources { memory: 0 virtual_cores: 0 } needed_resources { memory: 4608 virtual_cores: 1 } memory_seconds: 0 vcore_seconds: 0 } originalTrackingUrl: "N/A" currentApplicationAttemptId { application_id { id: 26 cluster_timestamp: 1440062075415 } attemptId: 1 } progress: 0.0 applicationType: "SPARK" }}
2015-08-21 11:07:24,263 INFO [main] | Application report for application_1440062075415_0026 (state: ACCEPTED)

.......(omitting some lines where the state are all ACCEPTED and final status are all UNDEFINED).....

2015-08-21 11:07:30,359 INFO [main] | Application report for application_1440062075415_0026 (state: FAILED)
2015-08-21 11:07:30,359 DEBUG [main] | 
     client token: N/A
     diagnostics: Application application_1440062075415_0026 failed 2 times due to AM Container for appattempt_1440062075415_0026_000002 exited with  exitCode: 1
For more detailed output, check application tracking page:http://msra-sa-44:8088/cluster/app/application_1440062075415_0026Then, click on links to logs of each attempt.
Diagnostics: Exception from container-launch.
Container id: container_1440062075415_0026_02_000001
Exit code: 1
Stack trace: ExitCodeException exitCode=1: 
    at org.apache.hadoop.util.Shell.runCommand(Shell.java:545)
    at org.apache.hadoop.util.Shell.run(Shell.java:456)
    at org.apache.hadoop.util.Shell$ShellCommandExecutor.execute(Shell.java:722)
    at org.apache.hadoop.yarn.server.nodemanager.DefaultContainerExecutor.launchContainer(DefaultContainerExecutor.java:211)
    at org.apache.hadoop.yarn.server.nodemanager.containermanager.launcher.ContainerLaunch.call(ContainerLaunch.java:302)
    at org.apache.hadoop.yarn.server.nodemanager.containermanager.launcher.ContainerLaunch.call(ContainerLaunch.java:82)
    at java.util.concurrent.FutureTask.run(Unknown Source)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(Unknown Source)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(Unknown Source)
    at java.lang.Thread.run(Unknown Source)

Shell output:         1 file(s) moved.

And then we opened stderr, it says:

Error: Could not find or load main class 'Dspark.app.name=org.apache.spark.examples.SparkPi'

It's so strange, this should be a parameter passed to java, and it seems that java recognized it as the main class. There should be a main class parameter in the command section of the log, but there is not.

How can that happen? What should we do to know what's wrong with it?

Thank you!

Answer 1

We solved this problem.

The root cause is that when generating the java command line, our Spark uses single quote('-Dxxxx') to wrap the parameters. Single quote works only in Linux. On Windows, the parameters are either not wrapped, or wrapped with double quotes("-Dxxxx"). The only way to solve this is to edit the source code of Spark and re-compile it.

It seems that this is currently an issue of Spark. (https://issues.apache.org/jira/browse/SPARK-5754)