I'm having the same problem. Thought I would throw in my configs etc here too.
Similar to Unassigned shards, v2 (Unanswered).
Each morning at 9am (JST, which is UTC0) a new index for logstash is created in elasticsearch. At this time, these new shards are UNASSIGNED. They remain in this state until manually rerouted using the
_cluster/reroute api. This can take a few minutes, or anywhere up to an hour to settle down, during which time, kibana has no access to information, and logstash begins erroring while elasticsearch is basically unavailable.
My elasticsearch.yml on all 3 nodes is the same, and basically unchanged:
discovery.zen.ping.unicast.hosts: ["10.73.11.105:9300", "10.73.12.105:9300", "10.73.13.105:9300"]
Any help is appreciated. Its frustrating to have to do manual operations on this cluster every morning.