rdblue opened a new pull request #1139: URL: https://github.com/apache/iceberg/pull/1139
Spark executors use a new thread for each task, and Iceberg reuses a thread-local evaluator when running row group filters. The combination of new threads and thread-local prevents those evaluators from being garbage collected. The evaluator holds state for the current row group, which can be large for the dictionary row group filter. This PR avoids a large memory leak by cleaning up references to the row group's dictionaries and other state after each evaluation. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
