? For a distributed storage system, the data is stored separately in a number of nodes, how to make the data distribution on different nodes of equilibrium, to ensure that its high availability? The so-called equilibrium, it is to point to each node in the system with uniform load, and the node was found in the uneven situation or add/remove, timely adjustment, maintain uniform state, this article explores Elasticsearch data distribution method, this paper described Elasticsearch 5.5, is the background of the
?? In the h, Shard as the smallest unit of data distribution/migration, the data to the node mapping separated into two layers: one is the number according to the Shard mapping (Route), another is Shard to node mapping (the Allocate),