Hi,
In 0.9, I added an improvement to the shutdown API to better perform a
full cluster shutdown. This is the recommended way to shutdown the full
cluster (and not killing / stopping processes one by one). When killing /
stopping nodes one by one, the cluster will start to allocate the shards to
the remaining nodes, which is wasteful, the cluster wide shutdown API is
smart enough not to do it.
-shay.banon