关于Spark on Yarn的一些经历

1、ERROR spark.SparkContext: Error initializing SparkContext 

org.apache.spark.SparkException: Yarn application has already ended! It might have been killed or unable to launch application master.

这是因为jdk为java 8的原因。

在所有节点上的yarn-site.xml上加上：

<property>
    <name>yarn.nodemanager.pmem-check-enabled</name>
    <value>false</value>
</property>
<property>
    <name>yarn.nodemanager.vmem-check-enabled</name>
    <value>false</value>
</property>

重启一下start-all.sh就可以

2、Name node is in safe mode

WARN Client: Failed to cleanup staging dir

The reported blocks 38 has reached the threshold 0.9990 of total blocks 38. The number of live datanodes 3 has reached the minimum number 0. In safe mode extension. Safe mode will be turned off automatically in 9 seconds.

解决方法：https://blog.csdn.net/xw13106209/article/details/6866072    

输入hadoop dfsadmin -safemode leave

3、Neither spark.yarn.jars nor spark.yarn.archive is set

解决方法：

命令行输入：

hadoop fs -mkdir   /spark_jars

hadoop fs -put /spark/jars/*  /spark_jars

然后在spark的conf的spark-default.conf

添加配置 ：spark.yarn.jars   hdfs://master:9000/spark_jars/*

4、Spark on YARN配置日志Web UI

1）在spark的conf的spark-default.conf

添加配置：（见http://blog.51cto.com/beyond3518/1787513、http://www.cnblogs.com/luogankun/p/3981645.html）

spark.eventLog.enabled             true
spark.eventLog.dir                 hdfs://master:9000/spark_logs
spark.yarn.historyServer.address   master:18080

2）在spark的conf的spark-env.sh

添加配置：

export SPARK_HISTORY_OPTS="-Dspark.history.ui.port=18080 -Dspark.history.fs.logDirectory=hdfs://master:9000/spark_logs"

注意，history.fs.logDirectory和上面的eventLog需要同一个文件。（见https://www.cnblogs.com/langfanyun/p/7788784.html）

3）在hadoop的yarn-site.xml配置加入：

<property>
  <name>yarn.resourcemanager.webapp.address</name>
  <value>bigdata01:8088</value>
</property>

这样，就可以直接从hadoop界面跳进spark web（见https://www.jianshu.com/p/ea85d074a494）

4）重启hadoop，然后启动./sbin/start-history-server.sh，就可以。