Operator tooling: cluster repair

[sanmiguel]

There have been scheduler bugs in the past that have led to orphaned executors: the only way currently to deal with these is to forcibly kill them off (e.g. using riak-mesos framework teardown) and start over.

We should investigate how we can provide tooling for an operator to manually bring a node back under control of a scheduler.

/cc @seanjensengrey

[seanjensengrey]

Not only scheduler bugs, but ZK corruption, etc. If we are to support running Riak clusters on Mesos with the same kind of uptime and longevity we see on bare metal we need to have ways to transition a node back to a normal operating state w/o killing it.