flink部署总结

部署

local模式

一，安装

1.下载安装包

https://archive.apache.org/dist/flink/

2.上传flink-1.12.0-bin-scala_2.12.tgz到node1的指定目录

3.解压

tar -zxvf flink-1.12.0-bin-scala_2.12.tgz

4.如果出现权限问题，需要修改权限

chown -R root:root /export/server/flink-1.12.0

5.改名或创建软链接

mv flink-1.12.0 flink
ln -s /export/server/flink-1.12.0 /export/server/flink

二，测试

1.准备文件/root/words.txt

vim /root/words.txt
hello me you her
hello me you
hello me
hello

2.启动Flink本地“集群”

/export/server/flink/bin/start-cluster.sh

3.使用jps可以查看到下面两个进程

 - TaskManagerRunner
- StandaloneSessionClusterEntrypoint

4.访问Flink的Web UI

http://node1:8081/#/overview

5.执行官方示例

/export/server/flink/bin/flink run
/export/server/flink/examples/batch/WordCount.jar --input /root/words.txt
--output /root/out

6.停止Flink

/export/server/flink/bin/stop-cluster.sh

standalone模式

独立模式，不依赖第三方资源管理框架

一，部署

1.集群规划:

- 服务器: node1(Master + Slave): JobManager + TaskManager
- 服务器: node2(Slave): TaskManager
- 服务器: node3(Slave): TaskManager

2.修改flink-conf.yaml

vim /export/server/flink/conf/flink-conf.yaml
jobmanager.rpc.address: node1
taskmanager.numberOfTaskSlots: 2
web.submit.enable: true

#历史服务器
jobmanager.archive.fs.dir: hdfs://node1:8020/flink/completed-jobs/
historyserver.web.address: node1
historyserver.web.port: 8082
historyserver.archive.fs.dir: hdfs://node1:8020/flink/completed-jobs/

2.修改masters

vim /export/server/flink/conf/masters
node1:8081

3.修改slaves

vim /export/server/flink/conf/workers
node1
node2
node3

4.添加HADOOP_CONF_DIR环境变量

vim /etc/profile
export HADOOP_CONF_DIR=/export/server/hadoop/etc/hadoop

5.分发

scp -r /export/server/flink node2:/export/server/flink
scp -r /export/server/flink node3:/export/server/flink
scp /etc/profile node2:/etc/profile
scp /etc/profile node3:/etc/profile

或

for i in {2..3}; do scp -r flink node$i:$PWD; done

6.source

source /etc/profile

二，测试

1.启动集群，在node1上执行如下命令

/export/server/flink/bin/start-cluster.sh

或者单独启动

flink/bin/jobmanager.sh ((start|start-foreground) cluster)|stop|stop-all

/export/server/flink/bin/taskmanager.sh start|start-foreground|stop|stop-all

2.启动历史服务器

/export/server/flink/bin/historyserver.sh start

3.访问Flink UI界面或使用jps查看

http://node1:8081/#/overview
http://node1:8082/#/overview

TaskManager界面：可以查看到当前Flink集群中有多少个TaskManager，每个TaskMa
nager的slots、内存、CPU Core是多少

4.执行官方测试案例

/export/server/flink/bin/flink run 
/export/server/flink/examples/batch/WordCount.jar --input
hdfs://node1:8020/wordcount/input/words.txt --output
hdfs://node1:8020/wordcount/output/result.txt --parallelism 2

5.查看历史日志

http://node1:50070/explorer.html#/flink/completed-jobs
http://node1:8082/#/overview

6.停止Flink集群

/export/server/flink/bin/stop-cluster.sh

standalone 高可用

1.集群规划

- 服务器: node1(Master + Slave): JobManager + TaskManager
- 服务器: node2(Master + Slave): JobManager + TaskManager
- 服务器: node3(Slave): TaskManager

2.启动ZooKeeper

zkServer.sh status
zkServer.sh stop
zkServer.sh start

3.启动HDFS

/export/serves/hadoop/sbin/start-dfs.sh

4.停止Flink集群

/export/server/flink/bin/stop-cluster.sh

5.修改flink-conf.yaml

vim /export/server/flink/conf/flink-conf.yaml

增加如下内容G

#开启HA，使用文件系统作为快照存储
state.backend: filesystem
#启用检查点，可以将快照保存到HDFS
state.backend.fs.checkpointdir: hdfs://node1:8020/flink-checkpoints
#使用zookeeper搭建高可用
high-availability: zookeeper
# 存储JobManager的元数据到HDFS
high-availability.storageDir: hdfs://node1:8020/flink/ha/
# 配置ZK集群地址
high-availability.zookeeper.quorum: node1:2181,node2:2181,node3:2181

6.修改masters

vim /export/server/flink/conf/masters
node1:8081
node2:8081

7.同步

scp -r /export/server/flink/conf/flink-conf.yaml
node2:/export/server/flink/conf/
scp -r /export/server/flink/conf/flink-conf.yaml
node3:/export/server/flink/conf/
scp -r /export/server/flink/conf/masters
node2:/export/server/flink/conf/
scp -r /export/server/flink/conf/masters
node3:/export/server/flink/conf/

8.修改node2上的flink-conf.yaml

vim /export/server/flink/conf/flink-conf.yaml
jobmanager.rpc.address: node2

9.重新启动Flink集群,node1上执行

/export/server/flink/bin/stop-cluster.sh
/export/server/flink/bin/start-cluster.sh

10.使用jps命令查看
发现没有Flink相关进程被启动
11.查看日志

cat /export/server/flink/log/flink-root-standalonesession-0-node1.log

发现如下错误
因为在Flink1.8版本后,Flink官方提供的安装包里没有整合HDFS的jar

12.下载jar包并在Flink的lib目录下放入该jar包并分发使Flink能够支持对Hadoop的操作

# 放入lib目录
cd /export/server/flink/lib
分发
for i in {2..3}; do scp -r flink-shaded-hadoop-2-uber-2.7.5-10.0.jar node$i:$PWD; done

13.重新启动Flink集群,node1上执行

/export/server/flink/bin/start-cluster.sh

14.使用jps命令查看,发现三台机器已经ok

测试

1.访问WebUI

http://node1:8081/#/job-manager/config
http://node2:8081/#/job-manager/config

2.执行wc

/export/server/flink/bin/flink run /export/server/flink/examples/batch/WordCount.jar

3.kill掉其中一个master
4.重新执行wc,还是可以正常执行

/export/server/flink/bin/flink run /export/server/flink/examples/batch/WordCount.jar

5.停止集群

/export/server/flink/bin/stop-cluster.sh

flink on yarn

flink和yarn的整合：

export FLINK_HOME=/usr/share/flink/flink-1.14.2
export PATH=:$FLINK_HOME/bin:$PATH
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
export HADOOP_CLASSPATH=`hadoop classpath`

1.关闭yarn的内存检查

vim /export/server/hadoop/etc/hadoop/yarn-site.xml



<!-- 关闭yarn内存检查 -->
<property>
<name>yarn.nodemanager.pmem-check-enabled</name>
 <value>false</value>
</property>
<property>
 <name>yarn.nodemanager.vmem-check-enabled</name>
 <value>false</value>
</property>

说明: 是否启动一个线程检查每个任务正使用的虚拟内存量，如果任务超出分配值，则直 接将其杀掉，默认是true。
在这里面我们需要关闭，因为对于flink使用yarn模式下，很容易内存超标，这个时 候yarn会自动杀掉job

2.同步

scp -r /export/server/hadoop/etc/hadoop/yarn-site.xml
node2:/export/server/hadoop/etc/hadoop/yarn-site.xml
scp -r /export/server/hadoop/etc/hadoop/yarn-site.xml
node3:/export/server/hadoop/etc/hadoop/yarn-site.xml

3.重启yarn

/export/server/hadoop/sbin/stop-yarn.sh
/export/server/hadoop/sbin/start-yarn.sh

测试

1，Session模式

yarn-session.sh(开辟资源) + flink run(提交任务)

在yarn上启动一个Flink会话，node1上执行以下命令

/export/server/flink/bin/yarn-session.sh -n 2 -tm 800 -s 1 -d

# 说明: 申请2个CPU、1600M内存
# -n 表示申请2个容器，这里指的就是多少个taskmanager
# -tm 表示每个TaskManager的内存大小
# -s 表示每个TaskManager的slots数量
# -d 表示以后台程序方式运行

2.查看UI界面

http://node1:8088/cluster

3.使用flink run提交任务：

/export/server/flink/bin/flink run /export/server/flink/examples/batch/WordCount.jar

运行完之后可以继续运行其他的小任务

/export/server/flink/bin/flink run /export/server/flink/examples/batch/WordCount.jar

4.通过上方的ApplicationMaster可以进入Flink的管理界面

5.关闭yarn-session

yarn application -kill application_1599402747874_0001
rm -rf /tmp/.yarn-properties-root

3 Per-Job分离模式

1.直接提交job

/export/server/flink/bin/flink run -m yarn-cluster -yjm 1024 -ytm 1024
/export/server/flink/examples/batch/WordCount.jar

# -m jobmanager的地址
# -yjm 1024 指定jobmanager的内存信息
# -ytm 1024 指定taskmanager的内存信息

用自己的jar包，带队列名称：

/export/server/flink/bin/flink run -m yarn-cluster -yjm 1024 -ytm 1024 -yqu prod -c flink.FlinkDataStreamWordCount /export/data/wc.jar --output hdfs://node1:8020/flink/output/wc_1

2.查看UI界面

http://node1:8088/cluster

3.注意：

在之前版本中如果使用的是flink on
yarn方式，想切换回standalone模式的话，如果报错需要删除：【/tmp/.yarn-properties-root】 rm -rf
/tmp/.yarn-properties-root
因为默认查找当前yarn集群中已有的yarn-session信息中的jobmanager