NightOwl888 commented on PR #938: URL: https://github.com/apache/lucenenet/pull/938#issuecomment-2113877070
> Very nice! Do you think that maybe the thread-unsafety of FSDirectory (see my comment above about the race conditions) could be causing that? And maybe if .NET 5+ performs better than .NET FX, that is exacerbating it? I'll look at improving the thread safety here next and see if that can have better results. Because one very real possibility in this code (and AFAICT Java Lucene 4.8 too) under heavy load is that an fsync/FlushFileBuffers gets missed because of thread races. Now that you mention it, it is possible that is the cause, (or a contributor) since we could be Fsyncing a partially written buffer to disk. While building a production quality implementation may take some time, if we locked all access to write, flush, and sync on all files, it would make testing the scenario pretty simple. I will set that up and run the check. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: dev-unsubscr...@lucenenet.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
