spark集群安装并集成到hadoop集群

https://www.scala-lang.org/files/archive/scala-2.11.12.tgz
tar zxf scala-2.11.12.tgz

chown hduser:hduser -R scala-2.11.11
mv /root/scala-2.11.11 /usr/local/scala

vim .bashrc
#scala var
export SCALA_HOME=/usr/local/scala
export PATH=$PATH:$SCALA_HOME/bin

注意:Spark与hadoop版本必须互相匹配,因为Spark会读取Hadoop HDFS 并且必须能在Hadoop YARN执行程序,所以必须要按照我们目前安装的Hadoop版本来选择
笔者这里用的是hadoop2.7.7 所以我选择的是Pre-built for Apache Hadoop 2.7 and later

http://mirror.bit.edu.cn/apache/spark/spark-2.3.3/spark-2.3.3-bin-hadoop2.7.tgz
tar zxf spark-2.3.3-bin-hadoop2.7.tgz

chown hduser:hduser spark-2.3.3-bin-hadoop2.7
mv spark-2.3.3-bin-hadoop2.7 /usr/local/spark

vim .bashrc
#spark var
export SPARK_HOME=/usr/local/spark
export PATH=$PATH:$SPARK_HOME/bin

修改master下的spark-env.sh  #没有这个文件可以cp spark-env.sh.template spark-env.sh
在最后一行添加如下
export PYSPARK_PYTHON=/usr/bin/python3
修改master下的spark bin目录下pyspark
将文本中
    PYSPARK_PYTHON=python
改为
    PYSPARK_PYTHON=python3

#取消INFO信息打印
复制conf目录下的log4j模本文件到log4j.properties
将文本中
    log4j.rootCategory=INFO, console
改为
    log4j.rootCategory=WARN, console

pyspark  --master local[4]

    spark 读取本地文件,所有节点都必须存在该文件
    textFile=sc.textFile("file:/usr/local/spark/README.md")
    spark 读取hdfs文件
    textFile2=sc.textFile("hdfs://hadoop-master-001:9000/wordcount/input/LICENSE.txt")

HADOOP_CONF_DIR=/usr/local/hadoop/etc/hadoop pyspark --master yarn --deploy-mode client
    textFile = sc.textFile("hdfs://hadoop-master-001:9000/wordcount/input/LICENSE.txt")
    textFile.count()

编辑spark-env.sh #spark_home/conf
    export SPARK_MASTER=hadoop-master-001            //设置master的ip或域名
    export SPARK_WORKER_CORES=1                        //设置每个worker使用的CPU核心
    export SPARK_WORKER_MEMORY=512m                    //设置每个worker使用的内存
    export SPARK_WORKER_INSTANCES=4                    //设置实例数

设置spark Standalone Cluster 服务器(master环境)
    vim /usr/local/spark/conf/slaves  添加ip或域名
    hadoop-data-001
    hadoop-data-002
    hadoop-data-003

pyspark --master spark://hadoop-master-001:7077 --num-executors 1 --total-executor-cores 3 --executor-memory 512m
    textFile = sc.textFile("file:/usr/local/spark/README.md")
    textFile.count()
    注意 当在cluster模式下,如yarn-client或spark standalone 读取本地文件时,因为程序是分不到不同的服务器,所以必须确认所有机器都有该文件,否则会发生错误.
    建议 最好在cluster读取hdfs文件,这样不会出现文件
    text2=sc.textFile("hdfs://hadoop-master-001:9000/wordcount/input/LICENSE.txt")
     text2.count() 

hadoop yarn运行pyspark时异常信息:
ERROR SparkContext: Error initializing SparkContext. org.apache.spark.SparkException: Yarn application has already ended! It might have been killed or unable to launch application master

解决方式
查看http://hadoop-master-001:8088/cluster/app/ 最新任务点击history 查看信息
"Diagnostics: Container [pid=29708,containerID=container_1563435447194_0007_02_000001] is running beyond virtual memory limits. Current usage: 55.6 MB of 1 GB physical memory used; 2.2 GB of 2.1 GB virtual memory used. Killing container."

修改所有节点的yarn-site.xml,添加如下
    <property>
        <name>yarn.nodemanager.pmem-check-enabled</name>
        <value>false</value>
    </property>
    <property>
        <name>yarn.nodemanager.vmem-check-enabled</name>
        <value>false</value>
    </property>
主节点执行stop-yarn.sh, start-yarn.sh 重启所有节点yarn