We are using elasticsearch 1.3.7 version with 5 data only nodes and 3 master only nodes, now we are working hard on elasticsearch upgrade
Now Facing some miscount issue ,one time same query gets 177 hits and another time getting 0 hits for same query.
getting random results (0 hits or 177 hits ) for every curl execution.
Checked after cache clear also, facing same issue.
If you use the cat shards API, do all shards of the index you are querying have the same number of documents? Does the count stay consistent if you use a preference parameter to ensure you always hit the same shards?
I would recommend that you prioritise upgrading, as a lot of improvements related to resiliency and correctness have gone intro more recent versions. Just because shard A have more documents than shard B does not necessarily mean that shard B does not have documents that are missing from shard A. Setting replica to 0 and then increasing it to 1 again would make the shards consistent, but may lose data. In order to make sure you capture all records, you may need to reindex data while making sure you extract data from all shards using the preference parameter. If you have an external data source that is the master, you could also reindex from that instead.
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.