The mit method synchronizes all the segments fsync in the memory to the disk, and finally clears the data in the translog (translog was not deleted in the 6.x version to realize sequenceIDs).

1.3 merge operation

The refresh operation will produce a large number of small segments, so each generated file will consume various resources, such as file handle, memory, CPU usage and so on. More importantly, each query request should check each segment in sequence; The more segments, the slower the retrieval.

ES will run a detection task, merge approximately-sized segments into a new large segment in the background, and delete the old segment.

1.4, multiple replication mechanism

ES has a multi-copy mechanism (default is 1 copy), and the primary and secondary segments of a segment cannot be segmented on the same node, which further ensures the reliability of data.

2. The process of expert system indexing.