Load the zookeeper node children as well as leaving a children watch, then invoke the applyZKChildren implementation to process the children list.
Load the zookeeper node data as well as leaving a data watch, then invoke the applyZKData implementation to process the data string.
Reconnect to the zookeeper, this maybe invoked when zookeeper connection expired and the node data watcher previously registered got dropped, hence re-attache the data wather here.
Optional, the destination to report the stats to
TODO: pick up new rev of Cluster once it's ready Soon enough the Cluster will be defined in a way that we can directly managing the managers in a more flexible way, by then we should be able to do batch update we want here.
zookeeper client talking to the zookeeper, it will only be used to read zookeeper
the zookeeper path representing the cache pool