从EMR集群主机外部使用spark-submit

Question

我们有一个Hadoop集群，运行在带有Spark 1.6.1的AWS弹性MapReduce(EMR)中。进入集群主机并提交Spark作业没有问题，但我们希望能够从另一个独立的EC2实例提交它们。

另一个“外部”EC2实例设置了安全组，以允许所有TCP流量进出EMR实例主从实例。它有一个直接从Apache网站下载的Spark的二进制安装。

$ /usr/local/spark/bin/spark-submit --master yarn --deploy-mode client --class org.apache.spark.examples.SparkPi /usr/local/spark/lib/spark-examples-1.6.1-hadoop2.6.0.jar 
16/06/22 13:58:52 INFO spark.SparkContext: Running Spark version 1.6.1
16/06/22 13:58:52 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
16/06/22 13:58:52 INFO spark.SecurityManager: Changing view acls to: jungd
16/06/22 13:58:52 INFO spark.SecurityManager: Changing modify acls to: jungd
16/06/22 13:58:52 INFO spark.SecurityManager: SecurityManager: authentication disabled; ui acls disabled; users with view permissions:     Set(jungd); users with modify permissions: Set(jungd)
16/06/22 13:58:52 INFO util.Utils: Successfully started service 'sparkDriver' on port 34757.
16/06/22 13:58:52 INFO slf4j.Slf4jLogger: Slf4jLogger started
16/06/22 13:58:52 INFO Remoting: Starting remoting
16/06/22 13:58:53 INFO Remoting: Remoting started; listening on addresses :[akka.tcp://sparkDriverActorSystem@172.31.61.189:39241]
16/06/22 13:58:53 INFO util.Utils: Successfully started service 'sparkDriverActorSystem' on port 39241.
16/06/22 13:58:53 INFO spark.SparkEnv: Registering MapOutputTracker
16/06/22 13:58:53 INFO spark.SparkEnv: Registering BlockManagerMaster
16/06/22 13:58:53 INFO storage.DiskBlockManager: Created local directory at /tmp/blockmgr-300d738e-d7e4-4ae9-9cfe-4e257a05d456
16/06/22 13:58:53 INFO storage.MemoryStore: MemoryStore started with capacity 511.1 MB
16/06/22 13:58:53 INFO spark.SparkEnv: Registering OutputCommitCoordinator
16/06/22 13:58:53 INFO server.Server: jetty-8.y.z-SNAPSHOT
16/06/22 13:58:53 INFO server.AbstractConnector: Started SelectChannelConnector@0.0.0.0:4040
16/06/22 13:58:53 INFO util.Utils: Successfully started service 'SparkUI' on port 4040.
16/06/22 13:58:53 INFO ui.SparkUI: Started SparkUI at http://172.31.61.189:4040
16/06/22 13:58:53 INFO spark.HttpFileServer: HTTP File server directory is /tmp/spark-5e332986-ae2a-4bde-9ae4-edb4fac5e1d7/httpd-e475fd1b-c5c8-4f31-9699-be89fff4a69c
16/06/22 13:58:53 INFO spark.HttpServer: Starting HTTP Server
16/06/22 13:58:53 INFO server.Server: jetty-8.y.z-SNAPSHOT
16/06/22 13:58:53 INFO server.AbstractConnector: Started SocketConnector@0.0.0.0:43525
16/06/22 13:58:53 INFO util.Utils: Successfully started service 'HTTP file server' on port 43525.
16/06/22 13:58:53 INFO spark.SparkContext: Added JAR file:/usr/local/spark/lib/spark-examples-1.6.1-hadoop2.6.0.jar at http://172.31.61.189:43525/jars/spark-examples-1.6.1-hadoop2.6.0.jar with timestamp 1466603933454
16/06/22 13:58:53 INFO client.RMProxy: Connecting to ResourceManager at ip-172-31-60-166.ec2.internal/172.31.60.166:8032
16/06/22 13:58:53 INFO yarn.Client: Requesting a new application from cluster with 2 NodeManagers
16/06/22 13:58:53 INFO yarn.Client: Verifying our application has not requested more than the maximum memory capability of the cluster (11520 MB per container)
16/06/22 13:58:53 INFO yarn.Client: Will allocate AM container, with 896 MB memory including 384 MB overhead
16/06/22 13:58:53 INFO yarn.Client: Setting up container launch context for our AM
16/06/22 13:58:53 INFO yarn.Client: Setting up the launch environment for our AM container
16/06/22 13:58:53 INFO yarn.Client: Preparing resources for our AM container
16/06/22 13:58:54 ERROR spark.SparkContext: Error initializing SparkContext.
org.apache.hadoop.security.AccessControlException: Permission denied: user=jungd, access=WRITE, inode="/user/jungd/.sparkStaging/application_1466437015320_0014":hdfs:hadoop:drwxr-xr-x
at         org.apache.hadoop.hdfs.server.namenode.FSPermissionChecker.check(FSPermissionChecker.java:319)
at     org.apache.hadoop.hdfs.server.namenode.FSPermissionChecker.check(FSPermissionChecker.java:292)
at     org.apache.hadoop.hdfs.server.namenode.FSPermissionChecker.checkPermission(FSPermissionChecker.java:213)

我不知道是怎么回事。非常感谢您的任何帮助。

Answer 1

从主机以外的机器运行spark-submit需要几件事：

• 需要在HDFS中创建与提交用户匹配的用户
  • 例如，使用色调控制台，或者直接创建/user/name文件夹，并使用主服务器上的Hadoop FS命令行工具设置权限
  • 如果在AWS EC2 EMR环境中，计算机、主服务器和从服务器的安全组可以显式地允许来自其他组。

  还可能需要在主服务器上创建用户作为Linux帐户。