1.12 Elasticsearch refresh

1. What Is refresh?

• An Elasticsearch operation that writes data from the memory buffer into the filesystem cache, after which it becomes searchable.

2. Why Is refresh Needed?

• Make the index searchable.
  • Note that after refresh, the data is in the filesystem cache and has not been persisted to disk, so data may be lost. Therefore Elasticsearch provides Elasticsearch translog.
  • Only after Elasticsearch flush is performed will segment files be persisted to disk.

3. Why Not fsync Directly to Disk?

• The refresh operation is lighter than fsync, although it sacrifices data reliability.

4. When refresh Is Triggered

4.1. Manual

• POST /my_index/_refresh can be used to force a refresh.

4.2. Automatic

• By default, refresh happens once per second.
• The refresh interval can be configured.

  PUT /my_index/_settings
      {
        "index" : {
          "refresh_interval" : -1
        }
      }

5. refresh Process

Write data from the memory buffer into the filesystem cache. A Lucene segment file is generated at this point.

6. References

• Refresh API | Elasticsearch Guide [7.13] | Elastic
• Understanding the Elasticsearch Write Process - Elastic Chinese Community
• Understanding ES refresh, flush, and merge - Sansi’s Blog - CSDN