The default value of spark.dynamicAllocation.shuffleTracking.enabled was changed from false to true in Spark 3.4.0, disabling it might help.
[1] https://spark.apache.org/docs/latest/core-migration-guide.html#upgrading-from-core-33-to-34 Thanks, Cheng Pan > On Sep 6, 2024, at 00:36, Jayabindu Singh <jayabi...@gmail.com> wrote: > > Dear Spark Users, > > We have run into an issue where with spark 3.3.2 using auto scaling with STS > is working fine, but with 3.4.2 or 3.5.2 executors are being left behind and > not scaling down. > Driver makes a call to remove the executor but some (not all) executors never > get removed. > > Has anyone else noticed this or aware of any reported issues? > > Any help will be greatly appreciated. > > Regards > Jay > --------------------------------------------------------------------- To unsubscribe e-mail: user-unsubscr...@spark.apache.org