storm坑之---同步问题

　　最近在做一个监控系统，用来监控网站上各个业务功能的调用量以及处理时间，以便及时发现问题，及时处理。做这种实时统计处理系统，自然首先想到了storm，于是现学现用，自然遇到了一些坑，而且不少是网上也难以找到的问题。在这里就做个记录，记录下这个最让我苦恼的错误。

　　首先我的业务逻辑是按分钟统计一分钟中的调用次数的数据，所以我在bolt里跑了一个定时器，定时将统计数据发到下一个bolt入库。所在我在定时器执行的代码里调用了OutputCollector发射到下一个bolt。本地调试没啥问题，就部署到外网环境测试。通常也没发现问题，但是偶尔会出现这种错误，作为开发人员最讨厌的就是这种可复现率很低的错误 。

　　这里是错误日志：

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

5675 [Thread-7-disruptor-executor[2 2]-send-queue] ERROR backtype.storm.daemon.executor - java.lang.RuntimeException: java.lang.NullPointerException     at backtype.storm.utils.DisruptorQueue.consumeBatchToCursor(DisruptorQueue.java:128) ~[storm-core-0.9.3.jar:0.9.3]     at backtype.storm.utils.DisruptorQueue.consumeBatchWhenAvailable(DisruptorQueue.java:99) ~[storm-core-0.9.3.jar:0.9.3]     at backtype.storm.disruptor$consume_batch_when_available.invoke(disruptor.clj:80) ~[storm-core-0.9.3.jar:0.9.3]     at backtype.storm.disruptor$consume_loop_STAR_$fn__1460.invoke(disruptor.clj:94) ~[storm-core-0.9.3.jar:0.9.3]     at backtype.storm.util$async_loop$fn__464.invoke(util.clj:463) ~[storm-core-0.9.3.jar:0.9.3]     at clojure.lang.AFn.run(AFn.java:24) [clojure-1.5.1.jar:na]     at java.lang.Thread.run(Thread.java:722) [na:1.7.0_15] Caused by: java.lang.NullPointerException: null     at clojure.lang.RT.intCast(RT.java:1087) ~[clojure-1.5.1.jar:na]     at backtype.storm.daemon.worker$mk_transfer_fn$fn__3549.invoke(worker.clj:129) ~[storm-core-0.9.3.jar:0.9.3]     at backtype.storm.daemon.executor$start_batch_transfer__GT_worker_handler_BANG_$fn__3283.invoke(executor.clj:258) ~[storm-core-0.9.3.jar:0.9.3]     at backtype.storm.disruptor$clojure_handler$reify__1447.onEvent(disruptor.clj:58) ~[storm-core-0.9.3.jar:0.9.3]     at backtype.storm.utils.DisruptorQueue.consumeBatchToCursor(DisruptorQueue.java:125) ~[storm-core-0.9.3.jar:0.9.3]     ... 6 common frames omitted             5697 [Thread-7-disruptor-executor[2 2]-send-queue] ERROR backtype.storm.util - Halting process: ("Worker died") java.lang.RuntimeException: ("Worker died")     at backtype.storm.util$exit_process_BANG_.doInvoke(util.clj:325) [storm-core-0.9.3.jar:0.9.3]     at clojure.lang.RestFn.invoke(RestFn.java:423) [clojure-1.5.1.jar:na]     at backtype.storm.daemon.worker$fn__3808$fn__3809.invoke(worker.clj:452) [storm-core-0.9.3.jar:0.9.3]     at backtype.storm.daemon.executor$mk_executor_data$fn__3274$fn__3275.invoke(executor.clj:240) [storm-core-0.9.3.jar:0.9.3]     at backtype.storm.util$async_loop$fn__464.invoke(util.clj:473) [storm-core-0.9.3.jar:0.9.3]     at clojure.lang.AFn.run(AFn.java:24) [clojure-1.5.1.jar:na]     at java.lang.Thread.run(Thread.java:722) [na:1.7.0_15]

　　如果你也遇到这个问题，相信你第一次看到这个错误一定很痛苦，因为错误日志中没有任何与自己的业务代码相关的记录。所以实在是无从定位问题的所在。痛苦至极的是复现还不那么容易。

　　经过我多次猜测尝试，终于测出了问题的所在。下面我先贴出一个会报这个错误的例子代码：

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

public class Main {       public static void main(String[] args) {         TopologyBuilder builder = new TopologyBuilder();         builder.setSpout("spout",new TestWordSpout());                   builder.setBolt("dispatch", new WordDispatchBolt()).shuffleGrouping("spout");         builder.setBolt("print",new PrintBolt()).fieldsGrouping("dispatch", new Fields("word"));                   Config conf = new Config();                   conf.setDebug(false);         conf.setNumWorkers(1);         //conf.put(Config.TOPOLOGY_MAX_SPOUT_PENDING, 1);         LocalCluster cluster = new LocalCluster();         cluster.submitTopology("test-kafka-1", conf, builder.createTopology());     }   }

　　

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

public class TestWordSpout extends BaseRichSpout {      　　private static final long serialVersionUID = 1L;     boolean _isDistributed;     SpoutOutputCollector _collector;     String[] words = new String[] {"nathan", "mike", "jackson", "golda", "bertels"};     public TestWordSpout() {         this(true);     }       public TestWordSpout(boolean isDistributed) {         _isDistributed = isDistributed;     }               public void open(Map conf, TopologyContext context, SpoutOutputCollector collector) {         _collector = collector;     }           public void close() {               }               public void nextTuple() {         Utils.sleep(1000);         final Random rand = new Random();         final String word = words[rand.nextInt(words.length)];         _collector.emit(new Values(word), word+new Random().nextDouble());     }           public void ack(Object msgId) {         System.out.println("### ack:"+msgId);     }       public void fail(Object msgId) {         System.out.println("### fail:"+msgId);     }           public void declareOutputFields(OutputFieldsDeclarer declarer) {         declarer.declare(new Fields("word"));     }

　　

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

public class WordDispatchBolt extends BaseRichBolt{       private OutputCollector collector;           @Override     public void prepare(Map stormConf, TopologyContext context,             OutputCollector collector) {         this.collector = collector;                   new Thread(new Runnable() {                           @Override             public void run() {                 while(true){                     send();//不做sleep休眠，否则抛出此异常的几率太小，不容易观察到                 }             }         }).start();     }       public void send(){         this.collector.emit(new Values(new Random().nextDouble()));     }     @Override     public void execute(Tuple input) {         String word = input.getStringByField("word");         this.collector.emit(new Values(word));         this.collector.ack(input);     }       @Override     public void declareOutputFields(OutputFieldsDeclarer declarer) {         declarer.declare(new Fields("word"));     }       }

　　

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

public class PrintBolt extends BaseRichBolt {       private static final long serialVersionUID = 1L;       @Override     public void prepare(Map stormConf, TopologyContext context,             OutputCollector collector) {     }       @Override     public void execute(Tuple input) {         System.out.println(input.getValue(0));     }       @Override     public void declareOutputFields(OutputFieldsDeclarer declarer) {               }     }

　　这个代码很简单，就不做详细介绍了。在WordDispatchBolt类里我启动了另一个线程来发射数据到下一个bolt。我的业务代码中与此类似，是通过Timer定时发送数据的（Timer底层其实也是线程，就不多说了）。但是Timer是按分钟调用的，所以出现问题的几率小的可怜，这里我故意零停顿的调用，让此异常发生的几率更大一些。 

　　如果运行以上例子代码，你也肯定遇到前边贴出的错误异常。如果不知道是OutputCollector的同步问题，相信解决起来绝对让人痛不欲生。既然知道了是同步问题，要么避免在别的线程里调用collector,要么改成同步的。以下是我简单想到的解决方案。（如果有大神还有更好的，希望留言指教）

　　对WordDispatchBolt类做如下修改：

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

public class WordDispatchBolt extends BaseRichBolt{       private OutputCollector collector;           @Override     public void prepare(Map stormConf, TopologyContext context,             OutputCollector collector) {         this.collector = collector;                   new Thread(new Runnable() {                           @Override             public void run() {                 while(true){                     send(new Values(new Random().nextDouble()));//不做sleep休眠，否则抛出此异常的几率太小，不容易观察到                 }             }         }).start();     }       public synchronized void send(List<Object> tuple){         this.collector.emit(tuple);     }     @Override     public void execute(Tuple input) {         String word = input.getStringByField("word");         send(new Values(word));         this.collector.ack(input);     }       @Override     public void declareOutputFields(OutputFieldsDeclarer declarer) {         declarer.declare(new Fields("word"));     }       }

 

　　到这里，这个坑算是基本得到解决了。之后可能还要大量使用到storm，遇到坑是再做记录。

 

　　”把遇到的坑记录下来，让后遇到者可以有更多的网络资源查询，以减少排查问题的时间和纠结“