The current implementation of 'ejectNode' can be possibly destructive to data if it is run against a node that is in the cluster and responsible for vbuckets.
We need to a) update the REST API documentation to describe specifically what it is to be used for and b) not allow ejectNode to be run on a node that is serving data
If absolutely necessary, a 'force' option could be implemented to allow ejection of an active node, but I don't think it's needed at this time.
|For Gerrit Dashboard: &For+MB-3423=message:MB-3423|
|4458,3||don't allow ejection of active cluster members. MB-3423||ns_server||Status: MERGED||+2||+1|