Have you tried using the split index API (assuming you are on a recent version)? Not sure if it would work on a completely full shard but it could be worth exploring...
That was my initial answer to be honest but the error message makes me think that the limit is per index and not per shard? Unless this is an error message coming directly from Lucene?
As you can see from the discussion @subash it would be useful to know which version of Elasticsearch you are using as well as get some index and/or shard statistics.
Yeah it's from Lucene so "index" means "shard" in Elasticsearch terminology.
The fundamental issue is that deleting a document requires indexing another document to record the delete, and that happens before the deleted document goes away. I suspect that splitting the index would work, but if that doesn't work then reindexing would work too.
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.