使用java远程提交flink任务到yarn集群

使用java远程提交flink任务到yarn集群

背景

由于业务需要，使用命令行的方式提交flink任务比较麻烦，要么将后端任务部署到大数据集群，要么弄一个提交机，感觉都不是很离线。经过一些调研，发现可以实现远程的任务发布。接下来就记录一下实现过程。这里用flink on yarn 的Application模式实现

环境准备

• 大数据集群，只要有hadoop就行
• 后端服务器，linux mac都行，windows不行

正式开始

1. 上传flink jar包到hdfs

去flink官网下载你需要的版本，我这里用的是flink-1.18.1，把flink lib目录下的jar包传到hdfs中。

其中flink-yarn-1.18.1.jar需要大家自己去maven仓库下载。

2. 编写一段flink代码

随便写一段flink代码就行，我们目的是测试

package com.azt;

import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.source.SourceFunction;

import java.util.Random;
import java.util.concurrent.TimeUnit;

public class WordCount {
    public static void main(String[] args) throws Exception {
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        DataStreamSource<String> source = env.addSource(new SourceFunction<String>() {
            @Override
            public void run(SourceContext<String> ctx) throws Exception {
                String[] words = {"spark", "flink", "hadoop", "hdfs", "yarn"};
                Random random = new Random();
                while (true) {
                    ctx.collect(words[random.nextInt(words.length)]);
                    TimeUnit.SECONDS.sleep(1);
                }
            }
            @Override
            public void cancel() {

            }
        });

        source.print();
        env.execute();
    }
}

3. 打包第二步的代码，上传到hdfs

4. 拷贝配置文件

• 拷贝flink conf下的所有文件到java项目的resource中
• 拷贝hadoop配置文件到到java项目的resource中

具体看截图

5. 编写java远程提交任务的程序

这一步有个注意的地方就是，如果你跟我一样是windows电脑，那么本地用idea提交会报错；如果你是mac或者linux，那么可以直接在idea中提交任务。

package com.test;


import org.apache.flink.client.deployment.ClusterDeploymentException;
import org.apache.flink.client.deployment.ClusterSpecification;
import org.apache.flink.client.deployment.application.ApplicationConfiguration;
import org.apache.flink.client.program.ClusterClient;
import org.apache.flink.client.program.ClusterClientProvider;
import org.apache.flink.configuration.*;
import org.apache.flink.runtime.client.JobStatusMessage;
import org.apache.flink.yarn.YarnClientYarnClusterInformationRetriever;
import org.apache.flink.yarn.YarnClusterDescriptor;
import org.apache.flink.yarn.YarnClusterInformationRetriever;
import org.apache.flink.yarn.configuration.YarnConfigOptions;
import org.apache.flink.yarn.configuration.YarnDeploymentTarget;

import org.apache.flink.yarn.configuration.YarnLogConfigUtil;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.yarn.api.records.ApplicationId;
import org.apache.hadoop.yarn.client.api.YarnClient;
import org.apache.hadoop.yarn.conf.YarnConfiguration;

import java.util.ArrayList;
import java.util.Collection;
import java.util.Collections;
import java.util.List;
import java.util.concurrent.CompletableFuture;

import static org.apache.flink.configuration.MemorySize.MemoryUnit.MEGA_BYTES;

/**
 * @date ：2021/5/12 7:16 下午
 */
public class Main {
    public static void main(String[] args) throws Exception {
        ///home/root/flink/lib/lib
        System.setProperty("HADOOP_USER_NAME","root");
//        String configurationDirectory = "C:\\project\\test_flink_mode\\src\\main\\resources\\conf";
        String configurationDirectory = "/export/server/flink-1.18.1/conf";
        org.apache.hadoop.conf.Configuration conf = new org.apache.hadoop.conf.Configuration();
        conf.set("fs.hdfs.impl","org.apache.hadoop.hdfs.DistributedFileSystem");
        conf.set("fs.file.impl", org.apache.hadoop.fs.LocalFileSystem.class.getName());
        String flinkLibs = "hdfs://node1.itcast.cn/flink/lib";
        String userJarPath = "hdfs://node1.itcast.cn/flink/user-lib/original.jar";
        String flinkDistJar = "hdfs://node1.itcast.cn/flink/lib/flink-yarn-1.18.1.jar";

        YarnClient yarnClient = YarnClient.createYarnClient();
        YarnConfiguration yarnConfiguration = new YarnConfiguration();
        yarnClient.init(yarnConfiguration);
        yarnClient.start();

        YarnClusterInformationRetriever clusterInformationRetriever = YarnClientYarnClusterInformationRetriever
                .create(yarnClient);
        //获取flink的配置
        Configuration flinkConfiguration = GlobalConfiguration.loadConfiguration(
                configurationDirectory);
        flinkConfiguration.set(CheckpointingOptions.INCREMENTAL_CHECKPOINTS, true);
        flinkConfiguration.set(
                PipelineOptions.JARS,
                Collections.singletonList(
                        userJarPath));
        YarnLogConfigUtil.setLogConfigFileInConfig(flinkConfiguration,configurationDirectory);
        Path remoteLib = new Path(flinkLibs);
        flinkConfiguration.set(
                YarnConfigOptions.PROVIDED_LIB_DIRS,
                Collections.singletonList(remoteLib.toString()));

        flinkConfiguration.set(
                YarnConfigOptions.FLINK_DIST_JAR,
                flinkDistJar);
        //设置为application模式
        flinkConfiguration.set(
                DeploymentOptions.TARGET,
                YarnDeploymentTarget.APPLICATION.getName());
        //yarn application name
        flinkConfiguration.set(YarnConfigOptions.APPLICATION_NAME, "jobname");
        //设置配置，可以设置很多
        flinkConfiguration.set(JobManagerOptions.TOTAL_PROCESS_MEMORY, MemorySize.parse("1024",MEGA_BYTES));
        flinkConfiguration.set(TaskManagerOptions.TOTAL_PROCESS_MEMORY, MemorySize.parse("1024",MEGA_BYTES));
        flinkConfiguration.set(TaskManagerOptions.NUM_TASK_SLOTS, 4);
        flinkConfiguration.setInteger("parallelism.default", 4);

        ClusterSpecification clusterSpecification = new ClusterSpecification.ClusterSpecificationBuilder()
                .createClusterSpecification();

//		设置用户jar的参数和主类
        ApplicationConfiguration appConfig = new ApplicationConfiguration(args,"com.azt.WordCount");


        YarnClusterDescriptor yarnClusterDescriptor = new YarnClusterDescriptor(
                flinkConfiguration,
                yarnConfiguration,
                yarnClient,
                clusterInformationRetriever,
                true);
        ClusterClientProvider<ApplicationId> clusterClientProvider = null;
        try {
            clusterClientProvider = yarnClusterDescriptor.deployApplicationCluster(
                    clusterSpecification,
                    appConfig);
        } catch (ClusterDeploymentException e){
            e.printStackTrace();
        }

        ClusterClient<ApplicationId> clusterClient = clusterClientProvider.getClusterClient();
        System.out.println(clusterClient.getWebInterfaceURL());
        ApplicationId applicationId = clusterClient.getClusterId();

        System.out.println(applicationId);

        Collection<JobStatusMessage> jobStatusMessages = clusterClient.listJobs().get();
        int counts = 30;
        while (jobStatusMessages.size() == 0 && counts > 0) {
            Thread.sleep(1000);
            counts--;
            jobStatusMessages = clusterClient.listJobs().get();
            if (jobStatusMessages.size() > 0) {
                break;
            }
        }
        if (jobStatusMessages.size() > 0) {
            List<String> jids = new ArrayList<>();
            for (JobStatusMessage jobStatusMessage : jobStatusMessages) {
                jids.add(jobStatusMessage.getJobId().toHexString());
            }
            System.out.println(String.join(",",jids));
        }

    }
}

由于我这里是windows电脑，所以我打包放到服务器上去运行
执行命令 ：

java -cp test_flink_mode-1.0-SNAPSHOT.jar com.test.Main

不出以外的话，会打印如下日志

log4j:WARN No appenders could be found for logger (org.apache.hadoop.util.Shell).
log4j:WARN Please initialize the log4j system properly.
log4j:WARN See http://logging.apache.org/log4j/1.2/faq.html#noconfig for more info.
http://node2:33811
application_1715418089838_0017
6d4d6ed5277a62fc9a3a274c4f34a468

复制打印的url连接，就可以打开flink的webui了，在yarn的前端页面中也可以看到flink任务。