本篇会给出如何使用python来读写kafka, 包含生产者和消费者.
以下使用kafka-python客户端
生产者
爬虫大多时候作为消息的发送端, 在消息发出去后最好能记录消息被发送到了哪个分区, offset是多少, 这些记录在很多情况下可以帮助快速定位问题, 所以需要在send方法后加入callback函数, 包括成功和失败的处理
# -*- coding: utf-8 -*-
'''
callback也是保证分区有序的, 比如2条消息, a先发送, b后发送, 对于同一个分区, 那么会先回调a的callback, 再回调b的callback
'''
import json
from kafka import KafkaProducer
topic = 'demo'
def on_send_success(record_metadata):
print(record_metadata.topic)
print(record_metadata.partition)
print(record_metadata.offset)
def on_send_error(excp):
print('I am an errback: {}'.format(excp))
def main():
producer = KafkaProducer(
bootstrap_servers='localhost:9092'
)
producer.send(topic, value=b'{"test_msg":"hello world"}').add_callback(on_send_success).add_callback(
on_send_error)
# close() 方法会阻塞等待之前所有的发送请求完成后再关闭 KafkaProducer
producer.close()
def main2():
'''
发送json格式消息
:return:
'''
producer = KafkaProducer(
bootstrap_servers='localhost:9092',
value_serializer=lambda m: json.dumps(m).encode('utf-8')
)
producer.send(topic, value={"test_msg": "hello world"}).add_callback(on_send_success).add_callback(
on_send_error)
# close() 方法会阻塞等待之前所有的发送请求完成后再关闭 KafkaProducer
producer.close()
if __name__ == '__main__':
# main()
main2()
消费者
kafka的消费模型比较复杂, 我会分以下几种情况来进行说明
1.不使用消费组(group_id=None)
不使用消费组的情况下可以启动很多个消费者, 不再受限于分区数, 即使消费者数量 > 分区数, 每个消费者也都可以收到消息
# -*- coding: utf-8 -*-
'''
消费者: group_id=None
'''
from kafka import KafkaConsumer
topic = 'demo'
def main():
consumer = KafkaConsumer(
topic,
bootstrap_servers='localhost:9092',
auto_offset_reset='latest',
# auto_offset_reset='earliest',
)
for msg in consumer:
print(msg)
print(msg.value)
consumer.close()
if __name__ == '__main__':
main()
2.指定消费组
以下使用pool方法来拉取消息
- pool 每次拉取只能拉取一个分区的消息, 比如有2个分区1个consumer, 那么会拉取2次
- pool 是如果有消息马上进行拉取, 如果timeout_ms内没有新消息则返回空dict, 所以可能出现某次拉取了1条消息, 某次拉取了max_records条
# -*- coding: utf-8 -*-
'''
消费者: 指定group_id
'''
from kafka import KafkaConsumer
topic = 'demo'
group_id = 'test_id'
def main():
consumer = KafkaConsumer(
topic,
bootstrap_servers='localhost:9092',
auto_offset_reset='latest',
group_id=group_id,
)
while True:
try:
# return a dict
batch_msgs = consumer.poll(timeout_ms=1000, max_records=2)
if not batch_msgs:
continue
'''
{TopicPartition(topic='demo', partition=0): [ConsumerRecord(topic='demo', partition=0, offset=42, timestamp=1576425111411, timestamp_type=0, key=None, value=b'74', headers=[], checksum=None, serialized_key_size=-1, serialized_value_size=2, serialized_header_size=-1)]}
'''
for tp, msgs in batch_msgs.items():
print('topic: {}, partition: {} receive length: '.format(tp.topic, tp.partition, len(msgs)))
for msg in msgs:
print(msg.value)
except KeyboardInterrupt:
break
consumer.close()
if __name__ == '__main__':
main()
关于消费组
我们根据配置参数分为以下几种情况
- group_id=None
- auto_offset_reset='latest': 每次启动都会从最新出开始消费, 重启后会丢失重启过程中的数据
- auto_offset_reset='latest': 每次从最新的开始消费, 不会管哪些任务还没有消费
- 指定group_id
- 全新group_id
- auto_offset_reset='latest': 只消费启动后的收到的数据, 重启后会从上次提交offset的地方开始消费
- auto_offset_reset='earliest': 从最开始消费全量数据
- 旧group_id(即kafka集群中还保留着该group_id的提交记录)
- auto_offset_reset='latest': 从上次提交offset的地方开始消费
- auto_offset_reset='earliest': 从上次提交offset的地方开始消费
- 全新group_id
性能测试
以下是在本地进行的测试, 如果要在线上使用kakfa, 建议提前进行性能测试
- producer
# -*- coding: utf-8 -*-
'''
producer performance
environment:
mac
python3.7
broker 1
partition 2
'''
import json
import time
from kafka import KafkaProducer
topic = 'demo'
nums = 1000000
def main():
producer = KafkaProducer(
bootstrap_servers='localhost:9092',
value_serializer=lambda m: json.dumps(m).encode('utf-8')
)
st = time.time()
cnt = 0
for _ in range(nums):
producer.send(topic, value=_)
cnt += 1
if cnt % 10000 == 0:
print(cnt)
producer.flush()
et = time.time()
cost_time = et - st
print('send nums: {}, cost time: {}, rate: {}/s'.format(nums, cost_time, nums // cost_time))
if __name__ == '__main__':
main()
'''
send nums: 1000000, cost time: 61.89236712455749, rate: 16157.0/s
send nums: 1000000, cost time: 61.29534196853638, rate: 16314.0/s
'''
- consumer
# -*- coding: utf-8 -*-
'''
consumer performance
'''
import time
from kafka import KafkaConsumer
topic = 'demo'
group_id = 'test_id'
def main1():
nums = 0
st = time.time()
consumer = KafkaConsumer(
topic,
bootstrap_servers='localhost:9092',
auto_offset_reset='latest',
group_id=group_id
)
for msg in consumer:
nums += 1
if nums >= 500000:
break
consumer.close()
et = time.time()
cost_time = et - st
print('one_by_one: consume nums: {}, cost time: {}, rate: {}/s'.format(nums, cost_time, nums // cost_time))
def main2():
nums = 0
st = time.time()
consumer = KafkaConsumer(
topic,
bootstrap_servers='localhost:9092',
auto_offset_reset='latest',
group_id=group_id
)
running = True
batch_pool_nums = 1
while running:
batch_msgs = consumer.poll(timeout_ms=1000, max_records=batch_pool_nums)
if not batch_msgs:
continue
for tp, msgs in batch_msgs.items():
nums += len(msgs)
if nums >= 500000:
running = False
break
consumer.close()
et = time.time()
cost_time = et - st
print('batch_pool: max_records: {} consume nums: {}, cost time: {}, rate: {}/s'.format(batch_pool_nums, nums,
cost_time,
nums // cost_time))
if __name__ == '__main__':
# main1()
main2()
'''
one_by_one: consume nums: 500000, cost time: 8.018627166748047, rate: 62354.0/s
one_by_one: consume nums: 500000, cost time: 7.698841094970703, rate: 64944.0/s
batch_pool: max_records: 1 consume nums: 500000, cost time: 17.975456953048706, rate: 27815.0/s
batch_pool: max_records: 1 consume nums: 500000, cost time: 16.711708784103394, rate: 29919.0/s
batch_pool: max_records: 500 consume nums: 500369, cost time: 6.654940843582153, rate: 75187.0/s
batch_pool: max_records: 500 consume nums: 500183, cost time: 6.854053258895874, rate: 72976.0/s
batch_pool: max_records: 1000 consume nums: 500485, cost time: 6.504687070846558, rate: 76942.0/s
batch_pool: max_records: 1000 consume nums: 500775, cost time: 7.047331809997559, rate: 71058.0/s
'''
该公众号是我个人维护的, 初衷是可以适当备份自己的笔记, 之后想可能会帮到更多的人. 发文时间不固定, 尽可能每篇文章直奔主题, 不去占用大家宝贵的时间. 如觉得对你有用, 欢迎关注交流.
