java 中带有 spark 文件流的检查点

Question

我想用 spark 文件流应用程序实现检查点，以处理来自 hadoop 的所有未处理文件，如果在任何情况下我的 spark 流应用程序 stop/terminates。我正在关注：streaming programming guide，但未找到 JavaStreamingContextFactory。请帮帮我，我该怎么做。

我的代码是

public class StartAppWithCheckPoint {

    public static void main(String[] args) {
        
        try {
            
            String filePath = "hdfs://Master:9000/mmi_traffic/listenerTransaction/2020/*/*/*/"; 
            String checkpointDirectory = "hdfs://Mongo1:9000/probeAnalysis/checkpoint";
            SparkSession sparkSession = JavaSparkSessionSingleton.getInstance();

            JavaStreamingContextFactory contextFactory = new JavaStreamingContextFactory() {
                  @Override public JavaStreamingContext create() {
                      
                    SparkConf sparkConf = new SparkConf().setAppName("ProbeAnalysis");
                    JavaSparkContext sc = new JavaSparkContext(sparkConf);  
                    JavaStreamingContext jssc = new JavaStreamingContext(sc, Durations.seconds(300));
                    JavaDStream<String> lines = jssc.textFileStream(filePath).cache();
                    
                    jssc.checkpoint(checkpointDirectory);
                    return jssc;
                  }
                };
                
            JavaStreamingContext context = JavaStreamingContext.getOrCreate(checkpointDirectory, contextFactory);
            
            context.start();
            context.awaitTermination();
            context.close();
            sparkSession.close();
            
        } catch(Exception e) {
            e.printStackTrace();
        }   
    }
}

Answer 1

你必须使用Checkpointing

对于检查点，使用 有状态 转换 updateStateByKey 或 reduceByKeyAndWindow。 spark-examples provided along with prebuild spark and spark source in git-hub. For your specific, see JavaStatefulNetworkWordCount.java;

中有很多示例

java 中带有 spark 文件流的检查点

Checkpoint with spark file streaming in java

java

hadoop

spark-streaming