On my elastic.co hosted cluster, i had 3 hot nodes with around 80% disk usage. I decided to add 2 warm nodes with no modification to the hot nodes configuration(i still have 3). Before creating the new warm nodes, elastic cloud carries out a task "performing a rolling change" which took about 1 hour (see screen shot below) which appeared to consist of trying to move all of the shards off each of the hot nodes in turn onto the other 2 hot nodes until they exceed the 93% disk watermark, turning my cluster red and loosing data in the process. Finally the warm nodes were created, but i dont understand why it needs to try to move all shards around when all i want to do is add a new node?
