利用RabbitMQ实现RPC（python）

Posted 2020-10-30

    RPC——远程过程调用，通过网络调用运行在另一台计算机上的程序的函数\方法，是构建分布式程序的一种方式。RabbitMQ是一个消息队列系统，可以在程序之间收发消息。利用RabbitMQ可以实现RPC。本文所有操作都是在CentOS7.3上进行的，示例代码语言为Python。

RabbiMQ以及pika模块安装

yum install rabbitmq-server python-pika -y systemctl    start rabbitmq-server

 

RPC的基本实现

RPC的服务端代码如下：

#!/usr/bin/env   python import pika   connection = pika.BlockingConnection(pika.ConnectionParameters(host='localhost')) channel = connection.channel() channel.queue_declare(queue='rpc_queue')   def fun(n):     return 2*n   def on_request(ch, method, props, body):     n = int(body)     response = fun(n)     ch.basic_publish(exchange='',         routing_key=props.reply_to,         properties=pika.BasicProperties(correlation_id = props.correlation_id),         body=str(response))     ch.basic_ack(delivery_tag = method.delivery_tag)   channel.basic_qos(prefetch_count=1) channel.basic_consume(on_request, queue='rpc_queue') print(" [x] Awaiting RPC requests") channel.start_consuming()

以上代码中，首先与RabbitMQ服务建立连接，然后定义了一个函数fun()，fun()功能很简单，输入一个数然后返回该数的两倍，这个函数就是我们要远程调用的函数。on_request()是一个回调函数，它作为参数传递给了basic_consume()，当basic_consume()在队列中消费1条消息时，on_request()就会被调用，on_request()从消息内容body中获取数字，并传给fun()进行计算，并将返回值作为消息内容发给调用方指定的接收队列，队列名称保存在变量props.reply_to中。

RPC的客户端代码如下：

#!/usr/bin/env   python import pika import uuid   class RpcClient(object):     def __init__(self):         self.connection = pika.BlockingConnection(pika.ConnectionParameters(host='localhost'))           self.channel = self.connection.channel()           result = self.channel.queue_declare(exclusive=True)         self.callback_queue = result.method.queue           self.channel.basic_consume(self.on_response, no_ack=True,                                    queue=self.callback_queue)       def on_response(self, ch, method, props, body):         if self.corr_id == props.correlation_id:             self.response = body       def call(self,n):         self.response = None         self.corr_id = str(uuid.uuid4())         self.channel.basic_publish(exchange='',                                      routing_key='rpc_queue',                                    properties=pika.BasicProperties(                                            reply_to = self.callback_queue,                                            correlation_id = self.corr_id,                                          ),                                    body=str(n))         while self.response is None:             self.connection.process_data_events()         return str(self.response)   rpc = RpcClient()   print(" [x] Requesting") response = rpc.call(2) print(" [.] Got %r" % response)

代码开始也是连接RabbitMQ，然后开始消费消息队列callback_queue中的消息，该队列的名字通过Request的属性reply_to传递给服务端，就是在上面介绍服务端代码时提到过的props.reply_to，作用是告诉服务端把结果发到这个队列。 basic_consume()的回调函数变成了on_response()，这个函数从callback_queue的消息内容中获取返回结果。

函数call实际发起请求，把数字n发给服务端程序，当response不为空时，返回response值。

下面看运行效果，先启动服务端：

在另一个窗口中运行客户端：

成功调用了服务端的fun()并得到了正确结果(fun(2)结果为4)。

 

总结：RPC的实现过程可以用下图来表示（图片来自RabbitMQ官网）：

当客户端启动时，它将创建一个callback queue用于接收服务端的返回消息Reply，名称由RabbitMQ自动生成，如上图中的amq.gen-Xa2..。同一个客户端可能会发出多个Request，这些Request的Reply都由callback queue接收，为了互相区分，就引入了correlation_id属性，每个请求的correlation_id值唯一。这样，客户端发起的Request就带由2个关键属性:reply_to告诉服务端向哪个队列返回结果；correlation_id用来区分是哪个Request的返回。

稍微复杂点的RPC

如果服务端定义了多个函数供远程调用怎么办？有两种思路，一种是利用Request的属性app_id传递函数名，另一种是把函数名通过消息内容发送给服务端。

1.我们先实现第一种，服务端代码如下：

#!/usr/bin/env   python import pika   connection = pika.BlockingConnection(pika.ConnectionParameters(host='localhost')) channel = connection.channel() channel.queue_declare(queue='rpc_queue')   def a():     return "a"   def b():     return "b"   def on_request(ch, method, props, body):     funname = props.app_id     if funname == "a":         response = a()     elif funname == "b":         response = b()       ch.basic_publish(exchange='',                      routing_key=props.reply_to,                      properties=pika.BasicProperties(correlation_id = \                                                            props.correlation_id),                      body=str(response))     ch.basic_ack(delivery_tag = method.delivery_tag)   channel.basic_qos(prefetch_count=1) channel.basic_consume(on_request, queue='rpc_queue')   print(" [x] Awaiting RPC requests") channel.start_consuming()

这次我们定义了2个不同函数a()和b()，分别打印不同字符串，根据接收到的app_id来决定调用哪一个。

客户端代码：

#!/usr/bin/env   python import pika import uuid   class RpcClient(object):     def __init__(self):         self.connection = pika.BlockingConnection(pika.ConnectionParameters(host='localhost'))           self.channel = self.connection.channel()           result = self.channel.queue_declare(exclusive=True)         self.callback_queue = result.method.queue           self.channel.basic_consume(self.on_response, no_ack=True,                                    queue=self.callback_queue)       def on_response(self, ch, method, props, body):         if self.corr_id == props.correlation_id:             self.response = body       def call(self,name):         self.response = None         self.corr_id = str(uuid.uuid4())         self.channel.basic_publish(exchange='',                                      routing_key='rpc_queue',                                    properties=pika.BasicProperties(                                            reply_to = self.callback_queue,                                            correlation_id = self.corr_id,                                            app_id = str(name),                                          ),                                    body="request")         while self.response is None:             self.connection.process_data_events()         return str(self.response)   rpc = RpcClient()   print(" [x] Requesting") response = rpc.call("b") print(" [.] Got %r" % response)

函数call()接收参数name作为被调用的远程函数的名字，通过app_id传给服务端程序，这段代码里我们选择调用服务端的函数b()，rpc.call(“b”)。

执行结果：

结果显示成功调用了函数b，如果改成rpc.call(“a”)，执行结果就会变成：

2.第二种实现方法，服务端代码：

#!/usr/bin/env   python import pika   connection = pika.BlockingConnection(pika.ConnectionParameters(host='localhost')) channel = connection.channel() channel.queue_declare(queue='rpc_queue')   def a():     return "a"   def b():     return "b"   def on_request(ch, method, props, body):     funname = str(body)     if funname == "a":         response = a()     elif funname == "b":         response = b()       ch.basic_publish(exchange='',                      routing_key=props.reply_to,                      properties=pika.BasicProperties(correlation_id = \                                                            props.correlation_id),                      body=str(response))     ch.basic_ack(delivery_tag = method.delivery_tag)   channel.basic_qos(prefetch_count=1) channel.basic_consume(on_request, queue='rpc_queue')   print(" [x] Awaiting RPC requests") channel.start_consuming()

客户端代码：

#!/usr/bin/env   python import pika import uuid   class RpcClient(object):     def __init__(self):         self.connection = pika.BlockingConnection(pika.ConnectionParameters(host='localhost'))           self.channel = self.connection.channel()           result = self.channel.queue_declare(exclusive=True)         self.callback_queue = result.method.queue           self.channel.basic_consume(self.on_response, no_ack=True,                                    queue=self.callback_queue)       def on_response(self, ch, method, props, body):         if self.corr_id == props.correlation_id:             self.response = body       def call(self,name):         self.response = None         self.corr_id = str(uuid.uuid4())         self.channel.basic_publish(exchange='',                                      routing_key='rpc_queue',                                    properties=pika.BasicProperties(                                            reply_to = self.callback_queue,                                            correlation_id = self.corr_id,                                          ),                                    body=str(name))         while self.response is None:             self.connection.process_data_events()         return str(self.response)   rpc = RpcClient()   print(" [x] Requesting") response = rpc.call("b") print(" [.] Got %r" % response)

与第一种实现方法的区别就是没有使用属性app_id，而是把要调用的函数名放在消息内容body中，执行结果跟第一种方法一样。

一个简单的实际应用案例

下面我们将编写一个小程序，用于收集多台KVM宿主机上的虚拟机数量和剩余可使用的资源。程序由两部分组成，运行在每台宿主机上的脚本agent.py和管理机上收集信息的脚本collect.py。从RPC的角度，agent.py是服务端，collect.py是客户端。

agent.py代码如下：

#!/usr/bin/python import pika import libvirt import psutil import json import socket import os import sys from xml.dom import minidom   #配置RabbitMQ地址 RabbitMQServer=x.x.x.x   #连接libvirt，libvirt是一个虚拟机、容器管理程序。 def get_conn():     conn = libvirt.open("qemu:///system")     if conn == None:         print '--Failed to open connection to   QEMU/KVM--'         sys.exit(2)     else:         return conn   #获取虚拟机数量 def getVMcount():     conn = get_conn()     domainIDs = conn.listDomainsID()     return len(domainIDs)   #获取分配给所有虚拟机的内存之和 def getMemoryused():     conn = get_conn()     domainIDs = conn.listDomainsID()     used_mem = 0     for id in domainIDs:         dom = conn.lookupByID(id)         used_mem += dom.maxMemory()/(1024*1024)     return used_mem   #获取分配给所有虚拟机的vcpu之和 def getCPUused():     conn = get_conn()     domainIDs = conn.listDomainsID()     used_cpu = 0     for id in domainIDs:         dom = conn.lookupByID(id)         used_cpu += dom.maxVcpus()     return used_cpu   #获取所有虚拟机磁盘文件大小之和 def getDiskused():     conn = get_conn()     domainIDs = conn.listDomainsID()     diskused = 0     for id in domainIDs:         dom = conn.lookupByID(id)         xml = dom.XMLDesc(0)         doc = minidom.parseString(xml)         disks = doc.getElementsByTagName('disk')         for disk in disks:             if disk.getAttribute('device') == 'disk':                 diskfile = disk.getElementsByTagName('source')[0].getAttribute('file')                 diskused += dom.blockInfo(diskfile,0)[0]/(1024**3)     return diskused   #使agent.py进入守护进程模式 def daemonize(stdin='/dev/null',stdout='/dev/null',stderr='/dev/null'):     try:         pid = os.fork()         if pid > 0:             sys.exit(0)     except OSError,e:         sys.stderr.write("fork #1 failed: (%d) %s\n" % (e.errno,e.strerror))         sys.exit(1)     os.chdir("/")     os.umask(0)     os.setsid()     try:         pid = os.fork()         if pid > 0:             sys.exit(0)     except OSError,e:         sys.stderr.write("fork #2 failed: (%d) %s\n" % (e.errno,e.strerror))         sys.exit(1)     for f in sys.stdout,sys.stderr,: f.flush()     si = file(stdin,'r')     so = file(stdout,'a+',0)     se = file(stderr,'a+',0)     os.dup2(si.fileno(),sys.stdin.fileno())     os.dup2(so.fileno(),sys.stdout.fileno())     os.dup2(se.fileno(),sys.stderr.fileno())   daemonize('/dev/null','/root/kvm/agent.log','/root/kvm/agent.log')   #连接RabbitMQ connection = pika.BlockingConnection(pika.ConnectionParameters(host= RabbitMQServer)) channel = connection.channel() channel.exchange_declare(exchange='kvm',type='fanout') result = channel.queue_declare(exclusive=True) queue_name = result.method.queue channel.queue_bind(exchange='kvm',queue=queue_name)   def on_request(ch,method,props,body):     sys.stdout.write(body+'\n')     sys.stdout.flush()     mem_total = psutil.virtual_memory()[0]/(1024*1024*1024)     cpu_total = psutil.cpu_count()     statvfs = os.statvfs('/datapool')     disk_total = (statvfs.f_frsize * statvfs.f_blocks)/(1024**3)     mem_unused = mem_total - getMemoryused()     cpu_unused = cpu_total - getCPUused()     disk_unused = disk_total - getDiskused() data = {             'hostname':socket.gethostname(),#宿主机名             'vm':getVMcount(),#虚拟机数量             'available memory':mem_unused,#可用内存             'available cpu':cpu_unused,#可用cpu核数             'available disk':disk_unused#可用磁盘空间             }     json_str = json.dumps(data)     ch.basic_publish(exchange='',                      routing_key=props.reply_to,                      properties=pika.BasicProperties(correlation_id=props.correlation_id),                      body=json_str                      )     ch.basic_ack(delivery_tag=method.delivery_tag) channel.basic_qos(prefetch_count=1) channel.basic_consume(on_request,queue=queue_name) sys.stdout.write(" [x] Awaiting RPC requests\n") sys.stdout.flush() channel.start_consuming()

collect.py代码如下：

#!/usr/bin/python import pika import uuid import json import datetime   #配置RabbitMQ地址 RabbitMQServer=x.x.x.x class RpcClient(object):     def __init__(self):         self.connection = pika.BlockingConnection(pika.ConnectionParameters(host=RabbitMQServer))         self.channel = self.connection.channel()         self.channel.exchange_declare(exchange='kvm',type='fanout')         result = self.channel.queue_declare(exclusive=True)         self.callback_queue = result.method.queue         self.channel.basic_consume(self.on_responses,no_ack=True,queue=self.callback_queue)         self.responses = []       def on_responses(self,ch,method,props,body):         if self.corr_id == props.correlation_id:             self.responses.append(body)       def call(self):         timestamp = datetime.datetime.strftime(datetime.datetime.now(),'%Y-%m-%dT%H:%M:%SZ')         self.corr_id = str(uuid.uuid4())         self.channel.basic_publish(exchange='kvm',                                      routing_key='',                                    properties=pika.BasicProperties(                                          reply_to = self.callback_queue,                                        correlation_id = self.corr_id,                                        ),                                    body='%s: receive a request' % timestamp                                    ) #定义超时回调函数        def outoftime():             self.channel.stop_consuming()         self.connection.add_timeout(30,outoftime)         self.channel.start_consuming()         return self.responses   rpc = RpcClient() responses = rpc.call() for i in responses:     response = json.loads(i)     print(" [.] Got %r" % response)

  本文在前面演示的RPC都是只有一个服务端的情况，客户端发起请求后是用一个while循环来阻塞程序以等待返回结果的，当self.response不为None，就退出循环。

  如果在多服务端的情况下照搬过来就会出问题，实际情况中我们可能有几十台宿主机，每台上面都运行了一个agent.py，当collect.py向几十个agent.py发起请求时，收到第一个宿主机的返回结果后就会退出上述while循环，导致后续其他宿主机的返回结果被丢弃。这里我选择定义了一个超时回调函数outoftime()来替代之前的while循环，超时时间设为30秒。collect.py发起请求后阻塞30秒来等待所有宿主机的回应。如果宿主机数量特别多，可以再调大超时时间。

  脚本运行需要使用的模块pika和psutil安装过程：

yum install -y python-pip python-devel pip install pika wget --no-check-certificate https://pypi.python.org/packages/source/p/psutil/psutil-2.1.3.tar.gz tar zxvf psutil-2.1.3.tar.gz cd psutil-2.1.3/ && python setup.py install

  脚本运行效果演示：