@Wenchen Fan <[email protected]> Gentle ping! This is a very common issue in many platforms running with external shuffle services like Celeborn etc. This will be a very good improvement to get it merged in OSS to benefit the overall community.
Regards Venkata krishnan On Sat, Oct 31, 2026 at 7:12 PM Venkatakrishnan Sowrirajan <[email protected]> wrote: > Hi Wenchen, > > Thanks for all your time and reviews so far on > https://github.com/apache/spark/pull/58437. It would be great if you can > take a look at it again when you get a chance. I would really like to see > this get merged. > > I really appreciate all the reviews so far and hope to get this merged > soon. Thanks! > > PRs: > > Primary PR: [SPARK-59138][CORE] Preserve reliably-stored shuffle map > outputs on executor loss <https://github.com/apache/spark/pull/58437> > > Follow on PRs: > 1. [SPARK-59807][CORE] Reject executor hold while an unreliably-stored > shuffle is active <https://github.com/apache/spark/pull/59083> > 2. [SPARK-59806][CORE] Make shuffle tracking storage-aware > <https://github.com/apache/spark/pull/58687> > > Regards > Venkata krishnan >
