超时元组的重要性是什么？

时间：2014-10-23 20:08:24

标签： apache-storm

我希望有人可以向我解释这种行为，因为它看起来很意外。

我看到除了在Spout上调用“fail”之外，元组的超时没有任何作用。元组本身仍将通过拓扑处理，除了acking / failing将不起作用。另一个问题是待处理元组的数量会增加 - 即使它们将流经拓扑，元组也不会被视为待处理。除非我遗漏了一些东西，否则这两个组合问题最多会使元组超时。完全没有意义，最糟糕的是非常有问题（因为它会将更多的元组扔进已经最大化的拓扑中）。

这是我的拓扑结构：

  • 我有一个鲸鱼嘴。在nextTuple上，它重新发出一个失败的元组，如果没有，则创建一个新的元组。
  • 我有一个螺栓需要4秒才能确定元组。
  • topology.max.spout.pending = 5
  • topology.message.timeout.secs = 5

我希望得到1或2个元组，并且4或3个元组超时 - 然后Bolt将接下来处理重新发送的元组。随着时间的推移，会有越来越多的元组被攻击（尽管它们会经常超时）。

我所看到的是，即使元组超时，它们仍然被博尔特处理。我假设Bolt有缓冲区/队列，并且没有从中清除超时元组。无论如何，这会导致所有元组超时，因为Bolt最终只会处理已经超时的元组。

我假设，并希望，我在这里遗漏了一些明显的东西......

两个问题：

  1. 我可以阻止Bolts处理已经超时的元组吗？
  2. 超时元组的重点是什么？它只会在Spout上调用失败，即使元组仍将由拓扑的其余部分处理！

    3. 感谢。

    脱粒机：

    public class SampleSpout extends BaseRichSpout {
    private static Logger logger = LoggerFactory.getLogger(SampleSpout.class);

    SpoutOutputCollector collector;
    Map<Integer, List<Object>> pending_map = new HashMap<Integer, List<Object>>();
    Queue<List<Object>> replay_queue = new LinkedBlockingQueue<List<Object>>();

    int contentCounter;
    int curMsgId;

    @Override
    public void declareOutputFields(OutputFieldsDeclarer declarer) {
        // unique-id always increments each time we emit.
        // msg-id gets incremented only when new tuples are created.
       declarer.declare(new Fields("msg-id", "content"));
    }

    @Override
    public void open(Map conf, TopologyContext context, SpoutOutputCollector spoutOutputCollector) {
        collector = spoutOutputCollector;
    }

    @Override
    public void nextTuple() {
        // either replay a failed tuple, or create a new one
        List<Object> tuple = null;
        if (replay_queue.size() > 0){
            tuple = replay_queue.poll();
        }else{
            tuple = new ArrayList<Object>();
            tuple.add(null);
            tuple.add("Content #" + contentCounter++);
        }

        // increment msgId and set it as the first item in the tuple
        int msgId = this.curMsgId++;
        tuple.set(0, msgId);
        logger.info("Emitting: " + tuple);
        // add this tuple to the 'pending' map, and emit it.
        pending_map.put(msgId, tuple);
        collector.emit(tuple, msgId);
        Utils.sleep(100);
    }

    @Override
    public void ack(Object msgId){
        // remove tuple from pending_map since it's no longer pending
        List<Object> acked_tuple = pending_map.remove(msgId);
        logger.info("Acked: " + acked_tuple);
    }

    @Override
    public void fail(Object msgId){
        // remove tuple from pending_map since it's no longer pending
        List<Object> failed_tuple = pending_map.remove(msgId);
        logger.info("Failed: " + failed_tuple);

        // put a copy into the replay queue
        ArrayList<Object> copy = new ArrayList<Object>(failed_tuple);
        replay_queue.add(copy);
    }
}

    螺栓：

    public class SamplePrintBolt extends BaseRichBolt {

    private static Logger logger = LoggerFactory.getLogger(SamplePrintBolt.class);

    OutputCollector collector;

    @Override
    public void prepare(Map stormConf, TopologyContext context, OutputCollector outputCollector) {
        collector = outputCollector;
    }

    @Override
    public void execute(Tuple input) {
        logger.info("I see: " + input.getValues());
        Utils.sleep(4000);
        logger.info("Done sleeping. Acking: "  + input.getValues());
        collector.ack(input);
    }

    @Override
    public void declareOutputFields(OutputFieldsDeclarer declarer) {
        // doesn't emit
    }
}

    主：

    public static void main(String[] args) throws Exception {
        Config conf = new Config();
        conf.setMaxSpoutPending(5);
        conf.setMessageTimeoutSecs(5);

        TopologyBuilder builder = new TopologyBuilder();
        builder.setSpout("spout", new SampleSpout());
        builder.setBolt("bolt1", new SamplePrintBolt()).shuffleGrouping("spout");

        LocalCluster cluster = new LocalCluster();
        cluster.submitTopology("local", conf, builder.createTopology());
}

    输出：

    30084 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [0, Content #0]
30085 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [0, Content #0]. Will now sleep...
30097 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [1, Content #1]
30097 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [2, Content #2]
30097 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [3, Content #3]
30097 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [4, Content #4]
34086 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [0, Content #0]
34086 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [1, Content #1]. Will now sleep...
34087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Acked: [0, Content #0]
34087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [5, Content #5]
38087 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [1, Content #1]
38087 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [2, Content #2]. Will now sleep...
38089 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Acked: [1, Content #1]
38089 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [6, Content #6]
-- So far, so good… however, now it's time for things to timeout.
40082 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [5, Content #5]
40082 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [4, Content #4]
40082 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [3, Content #3]
40083 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [2, Content #2]
40083 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [7, Content #5]
40084 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [8, Content #4]
40084 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [9, Content #3]
40085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [10, Content #2]
42088 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [2, Content #2]
-- Acking a timed-out tuple… this does nothing.
42088 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [3, Content #3]. Will now sleep…
-- Why is it looking at tuple #3?  This has already failed.
45084 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [6, Content #6]
45085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [11, Content #6]
46089 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [3, Content #3]
46089 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [4, Content #4]. Will now sleep...
50084 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [10, Content #2]
50085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [7, Content #5]
50085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [8, Content #4]
50085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [9, Content #3]
50085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [12, Content #2]
50085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [13, Content #5]
50085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [14, Content #4]
50085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [15, Content #3]
-- More timeouts…
50090 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [4, Content #4]
50090 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [5, Content #5]. Will now sleep...
54091 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [5, Content #5]
-- Yet the Bolt looks at tuple #5 which timed out 15 seconds ago…
54091 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [6, Content #6]. Will now sleep...
55085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [11, Content #6]
55085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [16, Content #6]
58091 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [6, Content #6]
58092 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [7, Content #5]. Will now sleep...
60085 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [15, Content #3]
60086 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [12, Content #2]
60086 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [13, Content #5]
60086 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [14, Content #4]
60086 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [17, Content #3]
60086 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [18, Content #2]
60086 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [19, Content #5]
60086 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [20, Content #4]
62093 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [7, Content #5]
62093 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [8, Content #4]. Will now sleep…
-- It's clear that the Bolt looks at tuples even if they have timed-out.  It's queue will get longer and longer and tuples will always timeout.
65086 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [16, Content #6]
65087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [21, Content #6]
66094 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [8, Content #4]
66094 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [9, Content #3]. Will now sleep...
70087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [20, Content #4]
70087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [19, Content #5]
70087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [18, Content #2]
70088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [17, Content #3]
70088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [22, Content #4]
70088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [23, Content #5]
70088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [24, Content #2]
70088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [25, Content #3]
70095 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [9, Content #3]
70095 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [10, Content #2]. Will now sleep...
74096 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [10, Content #2]
74096 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [11, Content #6]. Will now sleep...
75088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [21, Content #6]
75088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [26, Content #6]
78097 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [11, Content #6]
78097 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [12, Content #2]. Will now sleep...
80087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [25, Content #3]
80087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [24, Content #2]
80087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [23, Content #5]
80087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [22, Content #4]
80087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [27, Content #3]
80087 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [28, Content #2]
80088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [29, Content #5]
80088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [30, Content #4]
82098 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [12, Content #2]
82098 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [13, Content #5]. Will now sleep...
85088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [26, Content #6]
85088 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [31, Content #6]
86098 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [13, Content #5]
86099 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [14, Content #4]. Will now sleep...
90100 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [14, Content #4]
90101 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [15, Content #3]. Will now sleep...
90216 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [29, Content #5]
90216 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [30, Content #4]
90216 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [28, Content #2]
90217 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Failed: [27, Content #3]
90217 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [32, Content #5]
90217 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [33, Content #4]
90217 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [34, Content #2]
90217 [Thread-10-spout] INFO  com.appnexus.bsg.billing.storminator.SampleSpout - Emitting: [35, Content #3]
94101 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - Done sleeping. Acking: [15, Content #3]
94101 [Thread-8-bolt1] INFO  com.appnexus.bsg.billing.storminator.SamplePrintBolt - I see: [16, Content #6]. Will now sleep…
-- Problem gets exacerbated…  Bolt is now looking at tuples that have failed 30 seconds ago.

1 个答案:

答案 0 :(得分：3)

“超时”是一个捕捉任何失踪元组的功能 - 要么是发送给已经死亡的工人，要么是Bolt根本没有发生或失败。

因此，“topology.message.timeout.secs”应该设置为您认为拓扑永远不会超过的某个值，但这也是您需要处理多长时间的要求。如果你的价值太低，你就有可能超时仍然“活着”的元组，并堵塞你的系统（就像在原帖中一样）。如果您的值太高，则可能需要等待太长时间才能重新处理失败的元组（同样，这取决于您的要求）。

在超时“实时”元组的情况下，一种解决方案是在所有系统之间同步时钟，在每个元组上发出时间戳，并让每个单独的Bolt简单地传递太旧的元组，从而假定为时间进行。

在我的拙见中，这种行为应该由Storm本身来处理。 Storm可以使用所有元组发送（spout，timestamp）数据，并且在每批超时时，只需告诉每个Bolt失败的任何旧版本（spout，timeout-timestamp） - 比较足够快，并且Bolts可以转储（spout） ，timeout-timestamp）数据一旦遇到来自spout的元组就足够新了。

相关问题
最新问题