Question

为了增加Spark Streaming Programming指南中建议的并行性，我设置了多个接收器并尝试将它们列表联合起来。此代码按预期工作：

    private JavaDStream<SparkFlumeEvent> getEventsWorking(JavaStreamingContext jssc, List<String> hosts, List<String> ports) {

        List<JavaReceiverInputDStream<SparkFlumeEvent>> receivers = new ArrayList<>();

        for (String host : hosts) {
            for (String port : ports) {
                receivers.add(FlumeUtils.createStream(jssc, host, Integer.parseInt(port)));
            }
        }

        JavaDStream<SparkFlumeEvent> unionStreams = receivers.get(0)
                .union(receivers.get(1))
                .union(receivers.get(2))
                .union(receivers.get(3))
                .union(receivers.get(4))
                .union(receivers.get(5));

        return unionStreams;
    }

但我实际上并不知道我的群集在运行时会有多少个接收器。当我尝试在循环中执行此操作时，我得到了一个NPE。

 private JavaDStream<SparkFlumeEvent> getEventsNotWorking(JavaStreamingContext jssc, List<String> hosts, List<String> ports) {

        List<JavaReceiverInputDStream<SparkFlumeEvent>> receivers = new ArrayList<>();

        for (String host : hosts) {
            for (String port : ports) {
                receivers.add(FlumeUtils.createStream(jssc, host, Integer.parseInt(port)));
            }
        }

        JavaDStream<SparkFlumeEvent> unionStreams = null;
        for (JavaReceiverInputDStream<SparkFlumeEvent> receiver : receivers) {
            if (unionStreams == null) {
                unionStreams = receiver;
            } else {
                unionStreams.union(receiver);
            }
        }

        return unionStreams;
    }

ERROR：

16/09/15 17:05:25错误JobScheduler：作业生成器出错显示java.lang.NullPointerException 在org.apache.spark.streaming.DStreamGraph $$ anonfun $ getMaxInputStreamRememberDuration $ 2.apply（DStreamGraph.scala：172）在org.apache.spark.streaming.DStreamGraph $$ anonfun $ getMaxInputStreamRememberDuration $ 2.apply（DStreamGraph.scala：172）在scala.collection.TraversableOnce $$ anonfun $ maxBy $ 1.apply（TraversableOnce.scala：225）在scala.collection.IndexedSeqOptimized $ class.foldl（IndexedSeqOptimized.scala：51）在scala.collection.IndexedSeqOptimized $ class.reduceLeft（IndexedSeqOptimized.scala：68）在scala.collection.mutable.ArrayBuffer.reduceLeft（ArrayBuffer.scala：47）在scala.collection.TraversableOnce $ class.maxBy（TraversableOnce.scala：225）在scala.collection.AbstractTraversable.maxBy（Traversable.scala：105）在org.apache.spark.streaming.DStreamGraph.getMaxInputStreamRememberDuration（DStreamGraph.scala：172）在org.apache.spark.streaming.scheduler.JobGenerator.clearMetadata（JobGenerator.scala：270）在org.apache.spark.streaming.scheduler.JobGenerator.org $ apache $ spark $ streaming $ scheduler $ JobGenerator $$ processEvent（JobGenerator.scala：182）在org.apache.spark.streaming.scheduler.JobGenerator $$ anon $ 1.onReceive（JobGenerator.scala：87）在org.apache.spark.streaming.scheduler.JobGenerator $$ anon $ 1.onReceive（JobGenerator.scala：86）在org.apache.spark.util.EventLoop $$ anon $ 1.run（EventLoop.scala：48） 16/09/15 17:05:25 INFO MemoryStore：ensureFreeSpace（15128）调用 curMem = 520144，maxMem = 555755765 16/09/15 17:05:25 INFO MemoryStore：阻止broadcast_24存储为内存中的值（估计大小为14.8 KB，免费529.5 MB）线程中的异常＆＃34; main＆＃34; 显示java.lang.NullPointerException 在org.apache.spark.streaming.DStreamGraph $$ anonfun $ getMaxInputStreamRememberDuration $ 2.apply（DStreamGraph.scala：172）在org.apache.spark.streaming.DStreamGraph $$ anonfun $ getMaxInputStreamRememberDuration $ 2.apply（DStreamGraph.scala：172）在scala.collection.TraversableOnce $$ anonfun $ maxBy $ 1.apply（TraversableOnce.scala：225）在scala.collection.IndexedSeqOptimized $ class.foldl（IndexedSeqOptimized.scala：51）在scala.collection.IndexedSeqOptimized $ class.reduceLeft（IndexedSeqOptimized.scala：68）在scala.collection.mutable.ArrayBuffer.reduceLeft（ArrayBuffer.scala：47）在scala.collection.TraversableOnce $ class.maxBy（TraversableOnce.scala：225）在scala.collection.AbstractTraversable.maxBy（Traversable.scala：105）在org.apache.spark.streaming.DStreamGraph.getMaxInputStreamRememberDuration（DStreamGraph.scala：172）在org.apache.spark.streaming.scheduler.JobGenerator.clearMetadata（JobGenerator.scala：270）在org.apache.spark.streaming.scheduler.JobGenerator.org $ apache $ spark $ streaming $ scheduler $ JobGenerator $$ processEvent（JobGenerator.scala：182）在org.apache.spark.streaming.scheduler.JobGenerator $$ anon $ 1.onReceive（JobGenerator.scala：87）在org.apache.spark.streaming.scheduler.JobGenerator $$ anon $ 1.onReceive（JobGenerator.scala：86）在org.apache.spark.util.EventLoop $$ anon $ 1.run（EventLoop.scala：48）

这样做的正确方法是什么？

Answer 1

请你试试下面的代码，它会解决你的问题：

private JavaDStream<SparkFlumeEvent> getEventsNotWorking(JavaStreamingContext jssc, List<String> hosts, List<String> ports) {

    List<JavaDStream<SparkFlumeEvent>> receivers = new ArrayList<JavaDStream<SparkFlumeEvent>>();

    for (String host : hosts) {
        for (String port : ports) {
            receivers.add(FlumeUtils.createStream(jssc, host, Integer.parseInt(port)));
        }
    }

    return jssc.union(receivers.get(0), receivers.subList(1, receivers.size()));;
}

联合Spark Streaming中的Flume接收器列表

1 个答案: