Flink 1.10 on yarn集群搭建(hadoop 3.1.3)

2020双十阿里云服务器ECS团购特惠链接
前提条件：需先安装zookeeper与hadoop
zookeeper安装
 hadoop安装

一、系统配置

3台虚拟机
hadoop1: 4g内存 2核 80g硬盘
hadoop2 1g内存 1核 8g硬盘
hadoop3: 1g内存 1核 8g硬盘

二、基础概念

1、Flink Client

2、JobManager

3、TaskManager

三、flink集群搭建(on yarn)

集群搭建的主要内容为JobManager高可用配置

1、下载与解压

(1)下载安装包flink-1.10.0-bin-scala_2.12.tgz至hadoop1节点

curl -O http://apache.mirrors.hoobly.com/flink/flink-1.10.0/flink-1.10.0-bin-scala_2.12.tgz

(2)远程传输到hadoop2、hadoop3节点

scp flink-1.10.0-bin-scala_2.12.tgz root@hadoop2:`pwd`
scp flink-1.10.0-bin-scala_2.12.tgz root@hadoop3:`pwd`

(3)解压安装包

tar zxvf flink-1.10.0-bin-scala_2.12.tgz

安装结束

2、JobManager高可用配置

(1)yarn-site.xml
文件目录：/usr/local/softwareinstall/hadoop-3.1.3/etc/hadoop/yarn-site.xml(hadoop安装目录下的yarn集群配置文件）
添加如下配置

<property>
  <name>yarn.resourcemanager.am.max-attempts</name>
  <value>4</value>
  <description>
    The maximum number of application master execution attempts.
  </description>
</property>

(2)flink配置文件flink-conf.yaml
文件目录：(/usr/local/softwareinstall/flink-1.10.0/conf/flink-conf.yaml)
高可用设置

high-availability: zookeeper
high-availability.zookeeper.quorum: hadoop1:2181,hadoop2:2181,hadoop3:2181
high-availability.storageDir: hdfs:///flink/recovery
high-availability.zookeeper.path.root: /flink
yarn.application-attempts: 10

JobManager与TaskManager配置

(3)添加flink环境变量/etc/profile

vim /etc/profile

#flink环境变量配置
export FLINK_HOME=/usr/local/softwareinstall/flink-1.10.0
export PATH=$PATH:$FLINK_HOME/bin

#flink on yarn配置
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
export HADOOP_CLASSPATH=`hadoop classpath`

注：如果hadoop是集成在CDH中，配置HADOOP_CONF_DIR=/etc/hadoop/conf

source /etc/profile   #使环境变量生效

3、向yarn集群申请资源并提交任务

3.1 申请资源
(1)申请资源

bin/yarn-session.sh -nm wordCount  -n 2

(2)查看申请的资源
使用命令查看

yarn application --list

yarn集群ui界面查看

yarn.webapp.ui2界面查看

3.2 提交wordCount计算作业
(1)提交作业到application id为application_1589204773282_0003的yarn session中

cd /usr/local/softwareinstall/flink-1.10.0

./bin/flink run -yid application_1589204773282_0003 ./examples/batch/WordCount.jar

(2)flink job ui界面观察作业执行信息