在从节点 (YARN) 上运行 Spark 不起作用答案

【问题标题】：Running Spark on the slave node (YARN) doesn't work在从节点 (YARN) 上运行 Spark 不起作用
【发布时间】：2015-07-31 04:33:00
【问题描述】：

我可以在主节点上运行 SparkPi 示例，但是当我尝试相同的命令时 “spark-submit --class SparkPi --master yarn-client sparkpi.jar 10” 在从节点上，出现错误：

2015-05-19 14:05:44,881 INFO  [main] spark.SecurityManager (Logging.scala:logInfo(59)) - Changing view acls to: maintainer
2015-05-19 14:05:44,886 INFO  [main] spark.SecurityManager (Logging.scala:logInfo(59)) - Changing modify acls to: maintainer
2015-05-19 14:05:44,887 INFO  [main] spark.SecurityManager (Logging.scala:logInfo(59)) - SecurityManager: authentication disabled; ui acls disabled; users with view permissions: Set(maintainer); users with modify permissions: Set(maintainer)
2015-05-19 14:05:45,389 INFO  [sparkDriver-akka.actor.default-dispatcher-4] slf4j.Slf4jLogger (Slf4jLogger.scala:applyOrElse(80)) - Slf4jLogger started
2015-05-19 14:05:45,443 INFO  [sparkDriver-akka.actor.default-dispatcher-4] Remoting (Slf4jLogger.scala:apply$mcV$sp(74)) - Starting remoting
2015-05-19 14:05:45,641 INFO  [sparkDriver-akka.actor.default-dispatcher-3] Remoting (Slf4jLogger.scala:apply$mcV$sp(74)) - Remoting started; listening on addresses :[akka.tcp://sparkDriver@slave2.com:33055]
2015-05-19 14:05:45,644 INFO  [sparkDriver-akka.actor.default-dispatcher-3] Remoting (Slf4jLogger.scala:apply$mcV$sp(74)) - Remoting now listens on addresses: [akka.tcp://sparkDriver@slave2.com:33055]
2015-05-19 14:05:45,653 INFO  [main] util.Utils (Logging.scala:logInfo(59)) - Successfully started service 'sparkDriver' on port 33055.
2015-05-19 14:05:45,674 INFO  [main] spark.SparkEnv (Logging.scala:logInfo(59)) - Registering MapOutputTracker
2015-05-19 14:05:45,688 INFO  [main] spark.SparkEnv (Logging.scala:logInfo(59)) - Registering BlockManagerMaster
2015-05-19 14:05:45,707 INFO  [main] storage.DiskBlockManager (Logging.scala:logInfo(59)) - Created local directory at /tmp/spark-local-20150519140545-c81b
2015-05-19 14:05:45,712 INFO  [main] storage.MemoryStore (Logging.scala:logInfo(59)) - MemoryStore started with capacity 265.4 MB
2015-05-19 14:05:46,205 WARN  [main] util.NativeCodeLoader (NativeCodeLoader.java:<clinit>(62)) - Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
2015-05-19 14:05:46,408 INFO  [main] spark.HttpFileServer (Logging.scala:logInfo(59)) - HTTP File server directory is /tmp/spark-e95a2b5b-efea-41eb-93b9-0a9f7d6f6701
2015-05-19 14:05:46,413 INFO  [main] spark.HttpServer (Logging.scala:logInfo(59)) - Starting HTTP Server
2015-05-19 14:05:46,477 INFO  [main] server.Server (Server.java:doStart(272)) - jetty-8.y.z-SNAPSHOT
2015-05-19 14:05:46,499 INFO  [main] server.AbstractConnector (AbstractConnector.java:doStart(338)) - Started SocketConnector@0.0.0.0:52737
2015-05-19 14:05:46,500 INFO  [main] util.Utils (Logging.scala:logInfo(59)) - Successfully started service 'HTTP file server' on port 52737.
2015-05-19 14:05:46,790 INFO  [main] server.Server (Server.java:doStart(272)) - jetty-8.y.z-SNAPSHOT
2015-05-19 14:05:46,805 INFO  [main] server.AbstractConnector (AbstractConnector.java:doStart(338)) - Started SelectChannelConnector@0.0.0.0:4040
2015-05-19 14:05:46,805 INFO  [main] util.Utils (Logging.scala:logInfo(59)) - Successfully started service 'SparkUI' on port 4040.
2015-05-19 14:05:46,808 INFO  [main] ui.SparkUI (Logging.scala:logInfo(59)) - Started SparkUI at http://slave2.com:4040
2015-05-19 14:05:47,058 INFO  [main] spark.SparkContext (Logging.scala:logInfo(59)) - Added JAR file:/home/maintainer/myjars/sparkpi.jar at http://[ip]:52737/jars/sparkpi.jar with timestamp 1432033547057
2015-05-19 14:05:47,190 INFO  [main] client.RMProxy (RMProxy.java:createRMProxy(98)) - Connecting to ResourceManager at /0.0.0.0:8032
2015-05-19 14:09:45,861 INFO  [main] client.RMProxy (RMProxy.java:createRMProxy(98)) - Connecting to ResourceManager at /0.0.0.0:8032
**2015-05-19 14:09:47,067 INFO  [main] ipc.Client (Client.java:handleConnectionFailure(842)) - Retrying connect to server: 0.0.0.0/0.0.0.0:8032. Already tried 0 time(s); retry policy is RetryUpToMaximumCountWithFixedSleep(maxRetries=10, sleepTime=1000 MILLISECONDS)
2015-05-19 14:09:48,068 INFO  [main] ipc.Client (Client.java:handleConnectionFailure(842)) - Retrying connect to server: 0.0.0.0/0.0.0.0:8032. Already tried 1 time(s); retry policy is RetryUpToMaximumCountWithFixedSleep(maxRetries=10, sleepTime=1000 MILLISECONDS)
...**

【问题讨论】：

yarn config 中的yarn.resourcemanager.hostname 是什么？
@RemusRusanu，你是说 yarn-site.xml 吗？没有这样的参数
将其设置为 RM 主机名，在 RM 节点和所有 NM 节点上也是如此。
@RemusRusanu，我做到了（在所有节点的 yarn-site.xml 中添加了一个值为 mymaster.com 的属性 yarn.resourcemanager.hostname），但没有任何改变:(我也试过 ip地址而不是名字，再次没有运气

标签： hadoop apache-spark hadoop-yarn master-slave

【解决方案1】：

除了在yarn-site.xml 中指定yarn.resourcemanager.hostname 属性外，还需要将配置文件传播给worker。

这行可以完成（在运行spark-submit之前）：

export SPARK_YARN_DIST_FILES=$(ls $HADOOP_CONF_DIR* | sed 's#^#file://#g' | tr '\n' ',' | sed 's/,$//')

如果一切配置正确，您将在此行中看到 RM 主机名而不是 0.0.0.0：

2015-05-19 14:05:47,190 INFO  [main] client.RMProxy (RMProxy.java:createRMProxy(98)) - Connecting to ResourceManager at /0.0.0.0:8032

【讨论】：

谢谢，它帮助处理了上面提到的错误，但现在我有另一个错误：2015-05-25 11:10:51,982 WARN [Timer-0] cluster.YarnClientClusterScheduler (Logging.scala: logWarning(71)) - 初始作业未接受任何资源；检查您的集群 UI 以确保工作人员已注册并有足够的内存我尝试将参数添加到我的命令“spark-submit --class SparkPi --master yarn-client --executor-memory 512m --driver-memory 512m sparkexamples. jar 10" 但它没有帮助。

【解决方案2】：

为HADOOP_CONF_DIR 导出正确的值解决了这个问题。

export HADOOP_CONF_DIR=/your-path/hadoop/conf

【讨论】：