zookeeper 心跳上报

1.问题描述
zookeeper服务端通过zookeeper客户端上报心跳来确定zookeeper客户端的存活状态。当网络状态不稳定的时候，经常会造成zookeeper客户端心跳上报丢失，导致zookeeper服务端错以为zookeeper客户端死掉，然后删除相应结点。测试代码如下：

from kazoo.client import KazooClient, KazooState
import logging
import time

def my_listener(state):
  if state == KazooState.LOST:
    # Register somewhere that the session was lost
    logging.warning('handle lost')
  elif state == KazooState.SUSPENDED:
    # Handle being disconnected from Zookeeper
    logging.debug('handle being disconnected')
  else:
    # Handle being connected/reconnected to Zookeeper
    logging.debug('handle being (re)connected')


if __name__ == '__main__':
  logging.basicConfig(format='(%(name)s): %(asctime)s [%(levelname)s] %(module)s %(funcName)s %(message)s',
level=logging.DEBUG)

  logging.debug('starting...')

  zk = KazooClient(hosts='192.168.0.142:2181', timeout=30)
  zk.start()
  if zk.exists("/test999") is None:
    zk.create("/test999", bytes("test999".encode('utf-8')), ephemeral=True,makepath=True)
  zk.add_listener(my_listener)
  i=0
 while True:
      print(i)
      try:
        zk.set("/test999", bytes(str(i), encoding='utf8'))
        time.sleep(i+1)
        i=i+1
      except:
        print("error sleep 5s")
        time.sleep(5)
        #解决方法2
        # if zk.exists("/test999") is None:
        #   zk.create("/test999", bytes("test999".encode('utf-8')),

  zk.stop()

当正常运行时，zookeeper下会创建一个test999的结点

image.png

人为制造网络不畅，通过防火墙阻断与服务器的连接：
在命令行输入 iptables -A OUTPUT -d 192.168.0.142 -p tcp --dport 2181 -j DROP

image.png
等待一段时间后 ，输入 iptables -F
zookeeper客户端与服务器端连接建立，但是test999结点消失了

解决办法：
1.创建永久结点代替临时结点
2.在重新建立连接后自己代码手动添加 相同的 临时结点