warrenzhu25 opened a new pull request, #41083: URL: https://github.com/apache/spark/pull/41083
### What changes were proposed in this pull request? Add the config `spark.files.fetchFailure.unRegisterOutputThreshold` to control the number of fetch failed failures needed for one specific executor to unregister map output on this executor and allow to disable unregister. ### Why are the changes needed? Spark will unregister map output on the executor when fetch failed from this executor. This might be too aggressive when fetch failed is temporary and recoverable, especially when re-computation is more expensive than retry failed fetch. ### Does this PR introduce _any_ user-facing change? Yes. Added the config `spark.files.fetchFailure.unRegisterOutputThreshold` ### How was this patch tested? Added test in `DAGSchedulerSuite` -- 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]
