mridulm commented on pull request #28769: URL: https://github.com/apache/spark/pull/28769#issuecomment-906722390
The soft reference to iterator introduced in this pr unfortunately ended up causing iterators to not be closed when they go out of scope (which would have happened earlier in the finalize) This is because java is more conservative in cleaning up `SoftReference`'s. The net result was we ended up having 50k files for SHS while typically they get compacted away to 200 odd files. Changing from `SoftReference` to `WeakReference` should make it much more aggresive in cleanup and prevent the issue - which we observed in a 3.1 SHS +CC @thejdeep - please file a bug and include the contributors/reviewers of this PR for reviewing your change as well. Thx ! -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
