I believe the second one is correct as that is the actual description of the feature.
Elasticsearch went directly from version 2.x to 5.x to align version numbers across the stack, so there is no version 3 or 4. From earlier versions you will need to reindex the data.
In this case reindexing needs to be done acessing a running cluster with the data on the specified version, v1 or v2.
How you will reindex is entirely up to you, but you will need a cluster on v1/v2 with the data and a cluster on a newer version, basically you will read the data from the old cluster and write it in the new cluster.
The paid feature that can read data from v5 and v6 cluster will read the data from snapshots created on these versions, this would allow you to restore a v5/v6 snapshot on a v8 cluster for example.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.