Elasticsearch nodes are leaving the cluster continuously

Note that's just an overview - you should delay shard reassignment as we restart things (default is 60s) so stuff is not moving around on you during this process; set back to 60s when done (or I like 300s to give nodes time to reboot for upgrades, etc.):

PUT /_all/_settings
{"settings": {"index.unassigned.node_left.delayed_timeout": "1800s"}}}