【python】kafka在与celery和gevent连用时遇到的问题

前提:kafka有同步,多线程,gevent异步和rdkafka异步四种模式。但是在与celery和gevent连用的时候,有的模式会出错。

下面是我代码运行的结果。

 

结论:使用多线程方式!

 

使用同步方式可以成功发送数据

def send_data_kafka(data):
    try:
        client = KafkaClient(hosts=broker_list)
        topic = client.topics[topic_name]
        with topic.get_sync_producer() as producer:  
            for d in data:
                print "send data"
                msg = json.dumps(d)
                producer.produce(msg)
            producer.stop()
    except Exception, e:
        LOGGER.exception("error in send_data_kafka")
        print e

 

使用rdkafka异步,只打印了一条send data之后卡住

def send_data_kafka(data):
    try:
        client = KafkaClient(hosts=broker_list)
        topic = client.topics[topic_name]
        with topic.get_producer(use_rdkafka=True) as producer:  
            for d in data:
                print "send data"
                msg = json.dumps(d)
                producer.produce(msg)
            producer.stop()
    except Exception, e:
        LOGGER.exception("error in send_data_kafka")
        print e

 

 

使用多线程,可以正常生产所有数据

def send_data_kafka(data):
    try:
        client = KafkaClient(hosts=broker_list)
        topic = client.topics[topic_name]
        with topic.get_producer() as producer:  
            for d in data:
                print "send data"
                msg = json.dumps(d)
                producer.produce(msg)
            producer.stop()
    except Exception, e:
        LOGGER.exception("error in send_data_kafka")
        print e

 

没有用with,rdkafka异步,打印了所有的send data,后卡住

client = KafkaClient(hosts=broker_list)
topic = client.topics[topic_name]
producer = topic.get_producer(use_rdkafka=True)  # 异步,使用rdkafka库,速度最快的方案

def send_data_kafka(data):
    try:
        for d in data:
            print "send data"
            msg = json.dumps(d)
            producer.produce(msg)
        producer.stop()
    except Exception, e:
        LOGGER.exception("error in send_data_kafka")
        print e

 

posted @ 2017-10-26 19:30  匡子语  阅读(1766)  评论(0编辑  收藏  举报