Hi,all
Now elasticsearch can't detect which part of documents are inconsistent between primary and replica. I see cassandra use hashtree to imply it.
is it feasible on elk?
Regards
ES currently uses a file-based recovery approach, syncing Lucene segments that are present on the primary but missing on the replica. As primary and replica creates/merges segments independently from each other, this can often mean that all the segments need to be resynced (if they haven't been synced shortly before). We are currently working on Sequence numbers which introduce an ordering on indexing operations. This ordering can then be used to speed up shard recoveries, by identifying which specific operations need to be replayed to the recovering replica instead of falling back to a file based sync.
Ok, thanks for your reply. I am looking forward the exciting feature.