Our add-on service called "PALLAS", deployed in several clusters. It defined a set of role types called "Pallas Datanode xxxx", which the numeric suffix is the port of the role instance listened, so at most 14 datanode instances can run in the same host with different port. Now we want to expand the number of datanode instances in a host, e.g., add new role type named "Pallas Datanode 8351-8364". For all the exists datanode instances formed a big storage cluster(not the cm cluster), which support our online services, so we can't stop the instances simultaneously. The role type expansion for the exist clusters is inessential, we prepared a new cluster to hold hosts with more datanode instances.
An alternate choice is to define a new service with more datanode role types, but this will generate duplicate role types(different name, with the same port in different cluster), also increase our management difficulties.
