What would be the correct way to handle backups of a cluster that is used for time-based indexing, as you update the cluster to new versions of Elasticsearch ?
Basically, what we do now is we backup older indices on a different storage before deleting them when they're too old, in case we need to go back to them later. But since we regularly upgrade the cluster, at some point we won't be able to restore those indices on it, because their backup has been made on the N-2 version of Elasticsearch. The only way I can think of achieving restoration in this case is to create a node using the N-1 or N-2 Elasticsearch version, restore the backup on it then reindex to the production cluster, which seems cumbersome.
Are there good practices or tips we could use to achieve this kind of lifecycle management ?