RabbitMQ Stream类型队列

RabbitMQ提供了三种类型的队列：

• Classic

• Quorum

• Stream

官方文档 对于流队列的描述是：高性能、可持久化、可复制、非破坏性消费、只追加写入的日志

使用场景：

• 一个队列将同一条消息分发给不同消费者

• 可重复消费消息

• 更高的性能

  • 存储大量消息而不影响性能

  • 更高的吞吐

基本使用

生产消息：

import pika
from pika import BasicProperties
from pika.adapters.blocking_connection import BlockingChannel
from pika.spec import Basic
​
​
STREAM_QUEUE = "stream_queue"
​
connection = pika.BlockingConnection(pika.ConnectionParameters("localhost", 5672, "/"))
channel = connection.channel()
channel.queue_declare(queue=STREAM_QUEUE, durable=True, arguments={"x-queue-type": "stream"})
​
for i in range(500, 600):
    msg = f"{i}".encode()
    channel.basic_publish("", STREAM_QUEUE, msg)
​
channel.close()
connection.close()

消费消息：

import pika
from pika import BasicProperties
from pika.adapters.blocking_connection import BlockingChannel
from pika.spec import Basic
​
​
def msg_handler(channel: BlockingChannel, method: Basic.Deliver, properties: BasicProperties, body: bytes):
    msg = f"获取消息：{body.decode()}"
    print(msg)
    channel.basic_ack(method.delivery_tag)
​
​
STREAM_QUEUE = "stream_queue"
​
connection = pika.BlockingConnection(pika.ConnectionParameters("localhost", 5672, "/"))
channel = connection.channel()
channel.queue_declare(queue=STREAM_QUEUE, durable=True, arguments={"x-queue-type": "stream"})
​
channel.basic_qos(prefetch_count=50)
channel.basic_consume(STREAM_QUEUE, on_message_callback=msg_handler, arguments={"x-stream-offset": 290})
channel.start_consuming()
​
channel.close()
connection.close()

Offset参数

可以通过x-stream-offset来控制读取消息的位置，对于改参数值的释义见下图，详情可参考：Offset Tracking with RabbitMQ Streams。

chunk

上图中有个chunk的概念，chunk就是stream队列中用于存储和传输消息的单元，一个chunk包含几条到几千条不等的消息。

Stream 插件

以上只是对Stream类型队列的简单使用，API和普通队列没有差异。若要体验完整的Stream队列特性，如：服务端消息偏移量追踪，需要启用stream插件。

不启用和启用流插件功能特性对比，可参考： Stream Core vs Stream Plugin。

服务端消息偏移量追踪

Stream提供了服务端消息偏移量追踪，客户端断开重连后可以从上次消费的下一个位置开始消费消息。

⚠️ 有些客户端不支持dedicated binary 协议，无法提供完整的流队列特性支持

使用docker启动一个rabbitmq服务并启用stream插件:

docker run \
 -d --name rabbitmq \
 --hostname=node1 \
 --env=RABBITMQ_NODENAME=r1 \
 --env=RABBITMQ_SERVER_ADDITIONAL_ERL_ARGS='-rabbitmq_stream advertised_host localhost' \
 --volume=rabbit_erl:/var/lib/rabbitmq \
 -p 15672:15672 -p 5672:5672 -p 5552:5552 \
 rabbitmq:3-management
 
docker exec rabbitmq rabbitmq-plugins enable rabbitmq_stream

这里使用rstream客户端来收发消息：

import asyncio
​
from rstream import (
    Producer
)
​
STREAM_QUEUE = "stream_queue"
CONSUMER_NAME = "py"
​
​
async def pub():
    async with Producer("localhost", 5552, username="guest", password="guest") as producer:
        await producer.create_stream(STREAM_QUEUE)
        for i in range(100, 300):
            await producer.send(STREAM_QUEUE, f"{i}".encode())
​
​
if __name__ == "__main__":
    asyncio.run(pub())

消费消息：

import asyncio
​
from rstream import (
    AMQPMessage,
    Consumer,
    ConsumerOffsetSpecification,
    MessageContext,
    OffsetType, OffsetNotFound
)
​
STREAM_QUEUE = "stream_queue"
CONSUMER_NAME = "py"
​
​
async def msg_handler(msg: AMQPMessage, context: MessageContext):
    print(msg)
    await context.consumer.store_offset(STREAM_QUEUE, CONSUMER_NAME, context.offset)
​
​
async def sub():
    consumer = Consumer("localhost", 5552, username="guest", password="guest")
    await consumer.start()
    try:
        offset = await consumer.query_offset(STREAM_QUEUE, CONSUMER_NAME)
    except OffsetNotFound:
        offset = 1
    await consumer.subscribe(STREAM_QUEUE, msg_handler,
                             offset_specification=ConsumerOffsetSpecification(OffsetType.OFFSET, offset),
                             subscriber_name=CONSUMER_NAME)
    await consumer.run()
​
​
if __name__ == "__main__":
    asyncio.run(sub())

Kafka简单对比

推荐阅读

Streams

Offset Tracking with RabbitMQ Streams

RabbitMQ 端口