storm坑之---传递对象

继之前遇到的那个同步问题的坑之后(storm坑之---同步问题),最近对代码又做了调整和重构,并且又遇到了另一个storm开发中应该值得警惕的坑。接下来说说这个坑的大体情况。

  在我的storm程序中,Abolt需要将数据封装成一个对象同时发送给Bbolt和Cbolt各一份,Bbolt和Cbolt分别对对象做一定的处理后,更新到数据库。在查看日志时,意外的发现有些数据是不正确的诡异的,我先是怀疑算法问题,但又发现有部分数据又是正确的。算法应该没啥问题。纠结之下之后打印了更详细的日志,通过观察诡异数据的规律最后恍然大悟:肯定是Bbolt收到对象后对对象的修改影响到了Cbolt。在这里笔者几乎可以肯定的是:当Bbolt和Cbolt运行在同一个进程中时。发送给Bbolt和Cbolt的对象他们是公用的。Bbolt的修改会影响到Cbolt,反之亦然。如果Bbolt和Cbolt不是同一进程,则没有此影响。这就解释了为什么有的数据正常有的异常。

  下面举一个例子代码测试一下:

拓扑构建类:

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("test"new TestWordSpout());
         
        builder.setBolt("print1",new PrintBolt("PrintBolt1")).shuffleGrouping("test");
         
        builder.setBolt("print2",new PrintBolt("PrintBolt2")).shuffleGrouping("test");
         
        Config conf = new Config();
        conf.setDebug(false);
        conf.setNumWorkers(1);
        LocalCluster cluster = new LocalCluster();
        cluster.submitTopology("test-kafka-1", conf, builder.createTopology());
    }
 
}

spout类:

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
public class TestWordSpout extends BaseRichSpout {
    
    private static final long serialVersionUID = 1L;
    SpoutOutputCollector _collector;
     
    public void open(Map conf, TopologyContext context, SpoutOutputCollector collector) {
        _collector = collector;
    }
     
    public void close() {
         
    }
         
    public void nextTuple() {
        Utils.sleep(1000);
        Name name = new Name();
        name.setName("123");
        _collector.emit(new Values(name));
    }
     
    public void declareOutputFields(OutputFieldsDeclarer declarer) {
        declarer.declare(new Fields("word"));
    }
 
}

bolt类:

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
public class PrintBolt extends BaseRichBolt {
 
    private static final long serialVersionUID = 1L;
    private String name;
    int taskid;
 
    public PrintBolt(String name){
        this.name = name;
    }
     
    @Override
    public void prepare(Map stormConf, TopologyContext context,
            OutputCollector collector) {
        this.taskid = context.getThisTaskId();
         
    }
 
    @Override
    public void execute(Tuple input) {
        Name name = (Name) input.getValueByField("word");
        System.out.println(logPrefix()+name.getName());
        name.setName(this.name);
         
    }
 
    private String logPrefix(){
        return this.name+":";
    }
 
    @Override
    public void declareOutputFields(OutputFieldsDeclarer declarer) {
    }
}

可能发生的执行结果:

1
2
3
4
5
6
7
8
9
10
11
12
13
PrintBolt2:123
PrintBolt1:123
PrintBolt2:123
PrintBolt1:123
PrintBolt2:123
PrintBolt1:123
PrintBolt2:PrintBolt1
PrintBolt2:123
PrintBolt1:123
PrintBolt1:123
PrintBolt2:123
PrintBolt1:123
PrintBolt2:123

  从上边结果可以看到,PrintBolt2打印了PrintBolt1的修改。

  了解了这个情况,以后写代码就得要考虑到这种意外。如果一个对象会同时发送给两个bolt来处理,切bolt都要对此对象进行修改,在做修改之前一定要克隆一份,而不要直接修改!

原文地址:https://www.cnblogs.com/cxhfuujust/p/8950315.html