http://www.zlovezl.cn/articles/40/

简介：

　　Zookeeper 分布式服务框架是 Apache Hadoop 的一个子项目，它主要是用来解决分布式应用中经常遇到的一些数据管理问题，如：统一命名服务、状态同步服务、集群管理、分布式应用配置项的管理等。

具体简介可以参照这篇文章。

zkpython的安装：

python中有一个zkpython的包，是基于zookeeper的c-client开发的，所以安装的时候需要先安装zookeeper的c客户端。安装步骤如下：

# 首先下载zookeeper
wget http://labs.renren.com/apache-mirror//zookeeper/zookeeper-3.3.3/zookeeper-3.3.3.tar.gz
tar xzvf zookeeper-3.3.3.tar.gz
cd zookeeper-3.3.3/src/c/
./configure
make
make install# 然后下载zkpython
wget http://pypi.python.org/packages/source/z/zkpython/zkpython-0.4.tar.gz#md5=3de220615aaddf57f1462b78d32477f9
tar xzvf zkpython-0.4.tar.gz
cd zkpython-0.4
python setup.py install

这样就完成了zkpython的安装。

一个简单的demo：

之后让我们来写一个简单的demo吧。（demo中用到的zkclient.py:https://github.com/piglei/zkpython_example/blob/master/zkclient.py）

# coding: utf-8
import logging
from os.path import basename, joinfrom zkclient import ZKClient, zookeeper, watchmethodlogging.basicConfig(level = logging.DEBUG,format = "[%(asctime)s] %(levelname)-8s %(message)s"
)log = loggingclass GJZookeeper(object):ZK_HOST = "localhost:2181"ROOT = "/app"WORKERS_PATH = join(ROOT, "workers")MASTERS_NUM = 1TIMEOUT = 10000def __init__(self, verbose = True):self.VERBOSE = verboseself.masters = []self.is_master = Falseself.path = Noneself.zk = ZKClient(self.ZK_HOST, timeout = self.TIMEOUT)self.say("login ok!")# initself.__init_zk()# registerself.register()def __init_zk(self):"""create the zookeeper node if not exist"""nodes = (self.ROOT, self.WORKERS_PATH)for node in nodes: if not self.zk.exists(node):try:self.zk.create(node, "")except:pass@propertydef is_slave(self):return not self.is_masterdef register(self):"""register a node for this 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 masterself.get_master()def get_master(self):"""get children, and check who is the smallest child"""@watchmethoddef watcher(event):self.say("child changed, try to get master again.")self.get_master()children = self.zk.get_children(self.WORKERS_PATH, watcher)children.sort()self.say("%s's children: %s" % (self.WORKERS_PATH, children)) # check if I'm masterself.masters = children[:self.MASTERS_NUM]if self.path in self.masters:self.is_master = Trueself.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 timetime.sleep(1000)

　　这个简单的demo所做的事情，就是通过在zookeeper的/app/workers节点下建立临时的子节点( flags=zookeeper.EPHEMERAL | zookeeper.SEQUENCE )，每次create完成之后检查自己是不是在最小的MASTERS_NUM(例子中为1，即单master)里。如果是的话，作为master运行，否则的话，作为slave运行。

　　这样的话，当我们的master挂掉以后，与zookeeper之间的连接也会中断，过了指定的TIMEOUT以后，master之前在worker下的子节点就会被删除，于是slave节点之前设置的watcher会被触发，再次检查自己是否为master，如果是的话则完成切换。

demo运行结果：

# 第一个实例
Connected in 20 ms, handle is 0
[2011-09-09 12:40:43,702] INFO     login ok!
Node /app/workers/worker created in 4 ms
[2011-09-09 12:40:43,708] INFO     [ worker0000000022(slave) ] register ok! I'm worker0000000022
[2011-09-09 12:40:43,709] INFO     [ worker0000000022(slave) ] /app/workers's children: ['worker0000000022']
[2011-09-09 12:40:43,709] INFO     [ worker0000000022(master) ] I've become master!# 这时再起第二个实例
Connected in 64 ms, handle is 0
[2011-09-09 12:43:08,334] INFO     login ok!
Node /app/workers/worker created in 11 ms
[2011-09-09 12:43:08,346] INFO     [ worker0000000023(slave) ] register ok! I'm worker0000000023
[2011-09-09 12:43:08,347] INFO     [ worker0000000023(slave) ] /app/workers's children: ['worker0000000022', 'worker0000000023']
[2011-09-09 12:43:08,347] INFO     [ worker0000000023(slave) ] ['worker0000000022'] is masters, I'm slave# 杀掉master，第二个实例发生的变化
[2011-09-09 12:44:06,016] INFO     [ worker0000000023(slave) ] child changed, try to get master again.
[2011-09-09 12:44:06,017] INFO     [ worker0000000023(slave) ] /app/workers's children: ['worker0000000023']
[2011-09-09 12:44:06,017] INFO     [ worker0000000023(master) ] I've become master!