Before did not know why there is this problem, and now understand the
zookeeper cluster only provide services normally run the server in more than half the time!
I built a pseudo-cluster 3 servers
we first start one, and then query the state:
before because this problem is particularly distressed. .
When you open two servers, there is no problem, if not delete the data inside the version-2 and pid file restart
when all clusters are open, there is no problem
