Problems with mixed version cluster during rolling upgrade?

I have been confounded trying to upgrade my cluster from 2.0.0 to 2.3.2 (see Node upgraded 2.0.0 to 2.3.2 can’t communicate with other nodes in cluster). I have given up trying to in-place upgrade. Instead, I am bringing new nodes up with 2.3.2, and then I will pull out the nodes with 2.0.0.

I brought a new 2.3.2 cluster on-line, and it seems to be working (as opposed to when I brought an upgraded node online and all started failing). I know that a new node will not allocate to a node with an older version. When I brought the 2.3.2 node on-line, shards started allocating to it. I am guessing this is a one-way allocation. Is this a risky situation to be in? I don't want to pull a 2.0.0 node out, or add another 2.3.2 node in, until I know that all is well.


Only if you lose the 2.0 nodes.