Java提交Flink作业（二）

接上篇，上一篇说到java提交flink有四种方式，第三种方式最适合做大数据平台，因为作业提交完毕可以立马拿到ClusterClient对象，通过此对象可以立马拿到ApplicationId，jobmanager的host和port，有了这些信息就可以很容易的通过flink restapi或yarnClient获取作业的状态信息，从而实现监控作业的目的，话不多说，今儿就实现第三种提交任务的方式: 在上一篇FlinkClient的基础上新增一个重载的方法:

public void startJobV1(Application app){
    String flinkHome = app.getFlinkHome();
    String flinkConf=flinkHome+"/conf";
    Configuration flinkConfig = GlobalConfiguration.loadConfiguration(flinkConf);
    List<CustomCommandLine> customCommandLines = CliFrontend.loadCustomCommandLines(flinkConfig, flinkConf);
    setYarnConfig(flinkConfig);
    String flinkDistJar = getFlinkDistJar(flinkHome);
    if(flinkDistJar!=null&&!flinkDistJar.isEmpty()){
        flinkConfig.setString(YarnConfigOptions.FLINK_DIST_JAR.key(),flinkDistJar);
    }
    //如果系统里配了FLINK_HOME的环境变量，不用此项配置，如果没有配置 则需要把flink lib下的jar包上传至hdfs，然后此项配置设置成hdfs相应的目录
    flinkConfig.setString(YarnConfigOptions.PROVIDED_LIB_DIRS.key(),"hdfs://hadoop202:8020/flink-1.14.2/lib");
    log.info("-----------------------------flink config--------------------------");
    flinkConfig.toMap().forEach((k,v)->{
        log.info("{}:{}",k,v);
    });
    try {
        CliFrontend cli = new CliFrontend(flinkConfig, customCommandLines);
        SecurityUtils.install(new SecurityConfiguration(flinkConfig));
        String[] customArgs = toArgs(app);
        SecurityUtils.getInstalledContext().runSecured(() -> {
            //采用第三种提交任务的方式，从yarnDescriptor切入
            String[] params = (String[])Arrays.copyOfRange(customArgs, 1, customArgs.length);
            Options commandOptions = CliFrontendParser.getRunCommandOptions();
            CommandLine commandLine = cli.getCommandLine(commandOptions, params, true);
            CustomCommandLine activeCommandLine = cli.validateAndGetActiveCommandLine((CommandLine) Preconditions.checkNotNull(commandLine));
            ProgramOptions programOptions = ProgramOptions.create(commandLine);
            programOptions.validate();
            URI uri = PackagedProgramUtils.resolveURI(programOptions.getJarFilePath());
            //get effective commandline
            Configuration effectiveConfiguration = this.getEffectiveConfiguration(activeCommandLine, commandLine, programOptions, Collections.singletonList(uri.toString()),flinkConfig);
            ApplicationConfiguration applicationConfiguration = new ApplicationConfiguration(programOptions.getProgramArgs(), programOptions.getEntryPointClassName());
            //直接new一个yarnClusterFactory
            YarnClusterClientFactory clusterFactory = new YarnClusterClientFactory();
            YarnClusterDescriptor clusterDescriptor = clusterFactory.createClusterDescriptor(effectiveConfiguration);
            ClusterSpecification clusterSpecification = clusterFactory.getClusterSpecification(effectiveConfiguration);
            //部署集群
            ClusterClientProvider<ApplicationId> clusterClient = clusterDescriptor.deployApplicationCluster(clusterSpecification, applicationConfiguration);
            ClusterClient<ApplicationId> realClient = clusterClient.getClusterClient();
            int num=10;
            while(num>0){
                ApplicationId appId = realClient.getClusterId();
                String jmUrl = realClient.getWebInterfaceURL();
                if(jmUrl!=null && !jmUrl.isEmpty()){
                    log.info("appId:{},jm url:{}",appId,jmUrl);
                    break;
                }
                num--;
                Thread.sleep(5000);
            }

            cli.parseAndRun(customArgs);
            return 0;
        });
    } catch (Throwable var10) {
        Throwable strippedThrowable = ExceptionUtils.stripException(var10, UndeclaredThrowableException.class);
        log.error("Fatal error while running command line interface.", strippedThrowable);
        strippedThrowable.printStackTrace();
    }

}

使用这种方式提交的作业就可以很容易获取applicationId和jobmanager url，下面就是提交完成后打印的日志信息