Elasticsearch fault tolerance mechanism: master election, replica fault tolerance, data recovery (study notes)

(1) The Elasticsearch cluster has 9 shards and 3 nodes;
(2) If the master node is down, the cluster will automatically elect the master, and the cluster status is red at this time;
(3) Replica fault tolerance: the new master will upgrade the replica to the primary shard . At this time, the cluster status becomes yellow;
(4) Restart the down node, master copy replica to this node, use the original shard and synchronize the changes after the downtime. At this time, the cluster status changes to green.