Hadoop3.x,Spark 2.4配置--我遇到的问题汇总

 

配置Spark 绝对是一个扎心的过程，配置到你怀疑人生， 打算把大多的问题都记录下来，就当作记录遇到的坑

Table of Contents

 

WARN scheduler.TaskSchedulerImpl: Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources

 A JNI error has occurred, please check your installation and try again

ubuntu18 删除JAVA11并安装JAVA8

Yarn启动失败 java.io.IOException: Failed to send RPC 8277242275361198650 to datanode-055: java.nio.channels.ClosedChannelException

Py4JError("Answer from Java side is empty")

Terminal 下 spark-submit 会去启动Jupyter notebook

 

---

---

1.WARN scheduler.TaskSchedulerImpl: Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources

2.A JNI error has occurred, please check your installation and try again

3.ubuntu18 删除JAVA11并安装JAVA8

4.Yarn启动失败 java.io.IOException: Failed to send RPC 8277242275361198650 to datanode-055: java.nio.channels.ClosedChannelException

5.Py4JError("Answer from Java side is empty")

6.terminal下 spark-submit 会去启动Jupyter notebook

 

 

---

WARN scheduler.TaskSchedulerImpl: Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources

standalone 模式： jps查看进程，每一个node里的woker进程也都存在。 查看master：8080，发现worker数为0， 关闭spark后，重启 spark 后 解决问题。

ps.一般来说，not accepted any resources 就是没有woker了， 网吧秘技，重启spark解决

 

 

 A JNI error has occurred, please check your installation and try again

Java版本问题，default-java 为java11，改为java8 问题解决

 

 

ubuntu18 删除JAVA11并安装JAVA8

spark2.4在JAVA11下问题多多，真是搞得令人头大

首先 sudo update-java-alternatives 没啥用，可能是有些dependency仍然有影响，所以决定完全删除

default-jdk的dependency如下：

1. Depends: default-jre
2. Depends: default-jdk-headless
3. Depends: openjdk-11-jdk

sudo apt remove openjdk-11-jre-headless openjdk-11-jre openjdk-11-jdk-headless openjdk-11-jdk

在安装JAVA8 就行了

sudo apt-get install openjdk-8-jdk openjdk-8-jre

 

Yarn启动失败 java.io.IOException: Failed to send RPC 8277242275361198650 to datanode-055: java.nio.channels.ClosedChannelException

这个问题最恶心，Yarn无法启动，啥都败给，无数研究后终于解决，而且这个问题运行环境不同解决方法可能差很多

环境：ubuntu18.04

spark：2.4.2

hadoop：3.1.2

解决方法：step1降级到JAVA8， step2 配置yarn-site.xml文件(所有的node点都要修改)

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

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

在我的环境下这个问题已经解决了， 百度找到了其他方法是spark2.0.2的

https://blog.youkuaiyun.com/qq_38038143/article/details/88430151

 

Py4JError("Answer from Java side is empty")

设置spark.executor.memory,   spark.driver.memory，根据电脑设置我的16G内存，executor =1g num-executor=5，driver 512M 这个基本就是我的极限了，再大 就报错了

 

 

Terminal 下 spark-submit 会去启动Jupyter notebook

我在系统变量下设置了 

export PYSPARK_DRIVER_PYTHON=jupyter
export PYSPARK_DRIVER_PYTHON_OPTS='notebook'

结果spark-submit就不行了，把这两个注释掉再 source ～/.bashrc也不行

解决：terminal里输入

unset PYSPARK_DRIVER_PYTHON

就可以注销掉了，但是启动jupyter就需要手动设置driver了