For sake of proper API design and to simplify unit testing a cluster node should be able to be started and stopped over and over again.
Currently, a node's API is asymmetrical in that resources are "started" in constructor and may be stopped explicitly using method stop(). However, shell exposes method start() for starting current node, too. Due to the lack of support for starting a node separately, this is working once, only.
For sake of proper API design and to simplify unit testing a cluster node should be able to be started and stopped over and over again.
Currently, a node's API is asymmetrical in that resources are "started" in constructor and may be stopped explicitly using method
stop(). However, shell exposes methodstart()for starting current node, too. Due to the lack of support for starting a node separately, this is working once, only.