zookeeper应用场景之Master竞选

zookeeper应用场景中提出了对于Master节点管理的问题,如何保证集群中Master可用性和唯一性,下面就利用zookeeper来实现。

在确保zookeeper集群节点安装配置的前提下,假设zk已经对外提供了正常的服务,通过下面的步骤来实现Master竞选

  1. Client连接到zk上,判断znode /Roles/workers是否存在,不存在则建立,znode的类型是PERSISTENT类型,保证不会随着C1的session断开而消失。
  2. Client在/Roles/workers下面建立一个SEQUENCE|EPHEMERAL类型的znode,前缀可以是worker,由zk保证znode编号是递增而且是暂时的,EPHEMERAL在前文说了,一旦session断开创建的znode也会消失。
  3. Client通过getChildren获取所有的/Roles/workers下znode列表,并且设置一个Watcher等待通知,返回值有多少个znode数量就对应Client来竞选。
  4. 对于步骤4返回的节点列表进行排序,找到最小的worker编号,如果是和自己创建的一致(步骤2返回值),那么久代表自己的编号是最小的,自己就是Master。如果发现自己的编号不是最小,那么就等待通知,一旦Watcher触发,就在Watcher回到步骤3。

上面的机制主要利用了zk的几个特性

  • 对于N个客户端同时请求create一个znode,zk能保证顺序的一致性,并且保证每个客户端创建的znode节点是递增并且唯一。
  • 因为创建的znode是临时的,一旦session断开,那么znode就会从zk上消失,从而给每个设置Watcher的客户端发送通知,让每个客户端重新竞选Master,编号小的肯定是Master,保证了唯一性。

下图是上面的逻辑图
image

下面是实现的代码,默认是连接本地的zk服务,端口是2181,zkclient模块位于zookeeper python接口只需要运行多个下面的脚本就会能实现Master的竞选。

  1. 先后在三个终端上面运行下面的脚本,模拟为c1,c2,c3三个client,创建的节点依次是/Roles/workers/worker0000000000,/Roles/workers/worker0000000001,依次是/Roles/workers/worker0000000002
  2. 发现c1成功竞选了Master,然后c2和c3都是slave
  3. 把c1关了从而导致依次是/Roles/workers/worker0000000000消失,一段时间后c2和c3会重新竞选,c2会成为master,c3是slave
  4. 重新启动c1,发现c1立马加入集群,消息里面变更表示创建了新的znode依次是/Roles/workers/worker0000000003,重新竞选,c2还是master

PS:上面步骤3里面一个客户端关闭后经历了一段时间znode才会删除,原因是这段时间内zk的session还没有被清除,因为关闭是通过ctrl+c关闭的。但是加了一个客户端,znode里面创建,就会通知其余注册了watcher的客户端

#!/usr/bin/env python2.7
# -*- coding: UTF-8 -*-

import logging
from os.path import basename, join

from zkclient import ZKClient, zookeeper, watchmethod

logging.basicConfig(
    level = logging.DEBUG,
    format = "[%(asctime)s] %(levelname)-8s %(message)s"
)

log = logging

class GJZookeeper(object):

    ZK_HOST = "localhost:2181"
    ROOT = "/Roles"
    WORKERS_PATH = join(ROOT, "workers")
    MASTERS_NUM = 1
    TIMEOUT = 10000

    def __init__(self, verbose = True):
        self.VERBOSE = verbose
        self.masters = []
        self.is_master = False
        self.path = None

        self.zk = ZKClient(self.ZK_HOST, timeout = self.TIMEOUT)
        self.say("login ok!")
        # init
        self.__init_zk()
        # register
        self.register()

    def __init_zk(self):
        """
        create the zookeeper node if not exist
        |-Roles
             |-workers
        """
        nodes = (self.ROOT, self.WORKERS_PATH)
        for node in nodes: 
            if not self.zk.exists(node):
                try:
                    self.zk.create(node, "")
                except:
                    pass

    @property
    def is_slave(self):
        return not self.is_master

    def register(self):
        """
        register a node for this worker,znode type : EPHEMERAL | SEQUENCE
        |-Roles
             |-workers
                     |-worker000000000x         ==>>master
                     |-worker000000000x+1       ==>>worker
                     ....
        """
        self.path = self.zk.create(self.WORKERS_PATH + "/worker", "1", flags=zookeeper.EPHEMERAL | zookeeper.SEQUENCE)
        self.path = basename(self.path)
        self.say("register ok! I'm %s" % self.path)
        # check who is the master
        self.get_master()

    def get_master(self):
        """
        get children, and check who is the smallest child
        """
        @watchmethod
        def watcher(event):
            self.say("child changed, try to get master again.")
            self.get_master()
        try :
            children = self.zk.get_children(self.WORKERS_PATH, watcher)
        except zookeeper.ConnectionLossException:
            print "losing connection with zookeeper..."
            return False
        except :
            return False
        children.sort()
        self.say("%s's children: %s" % (self.WORKERS_PATH, children)) 

        # check if I'm master
        self.masters = children[:self.MASTERS_NUM]
        if self.path in self.masters:
            self.is_master = True
            self.say("I've become master!")
        else:
            self.say("%s is masters, I'm slave" % self.masters)


    def say(self, msg):
        """
        print messages to screen
        """
        if self.VERBOSE:
            if self.path:
                log.info("[ %s(%s) ] %s" % (self.path, "master" if self.is_master else "slave", msg))
            else:
                log.info(msg)

def main():
    gj_zookeeper = GJZookeeper()

if __name__ == "__main__":
    main()
    import time
    time.sleep(1000)

标签:Zookeeper