Closing IndexWriter
might wait a long time for merges to complete
#107513
Labels
>bug
:Distributed/Engine
Anything around managing Lucene and the Translog in an open shard.
Team:Distributed
Meta label for distributed team
Another instance of #89821: I saw a case where a cluster took 3+ minutes to apply a cluster state, and on closer inspection it was waiting on aborted merges to complete here:
The associated merge thread was busy here:
AFAICT today there's no mechanism to abort the big read that
checksumEntireFile
requires when the corresponding merge is aborted. In this context it is plausible that this read might take a few minutes to complete (the data was coming in chunks from S3). I think we should have such a mechanism.The text was updated successfully, but these errors were encountered: