Zookeeper client disconnects and reconnects with an error

[ChenJhua]

1. start python process，create zk client，hosts has two effective zk server
2. find zk client connect zk server and kill this zk server
3. don't reconnect other zk server and has exception
4. I use or not use kazooretry has simple exception
5. I use connection_listener has simple exception

   Traceback (most recent call last):
   File "dmoc_update_cfg.py", line 110, in <module>
   update_cfg.create_tem_node(hid)
   File "dmoc_update_cfg.py", line 35, in create_tem_node
   if not self.kr(self.zk_obj.exists, self.path):
   File "/usr/lib/python2.7/site-packages/kazoo/retry.py", line 129, in __call__
   return func(*args, **kwargs)
   File "/usr/lib/python2.7/site-packages/kazoo/client.py", line 1033, in exists
   return self.exists_async(path, watch=watch).get()
   File "/usr/lib/python2.7/site-packages/kazoo/handlers/utils.py", line 69, in get
   raise self._exception
   kazoo.exceptions.SessionExpiredError

   I don't know why has exception and reconnect effective zk server need do something?

[StephenSorriaux]

Hi,

Sounds related to https://github.com/python-zk/kazoo/issues/457

[ChenJhua]

I trying #457 ,but not use; If I have three hosts A,B,C, and kazoo client connecting to A, then if I kill A, kazoo client will connect to two other useful hosts B,C, such as B. But if I start A now and kill B, kazoo client will not select A, because A has been removed from the list of valid hosts?

[StephenSorriaux]

A is not removed from the list. If you set hosts to A, B and C, you will first connect to A, If you get disconnected from A, you will connect to B. Again, if you get disconnected from B, you will move to C. And so on.

[ChenJhua]

Yeah,I know it,but during this period, A is back to normal, can we connect to A? Do you use the list to mark invalid hosts or valid hosts and remove them from the list if they cannot be connected?

Traceback (most recent call last):
  File "/usr/lib/python2.7/site-packages/kazoo/client.py", line 1033, in exists
    return self.exists_async(path, watch=watch).get()
  File "/usr/lib/python2.7/site-packages/kazoo/handlers/utils.py", line 69, in get
    raise self._exception
kazoo.exceptions.SessionExpiredError

when I capture SessionExpiredError exception，I use kazoo client _reset()，B server down will reconnect to A。

Using _reset() has those effects？