Mridul M
Mridul M
+CC @waitinfuture, @jiang13021
Gentle ping on this @waitinfuture
> However, you might want to link to this PR in your comment: https://github.com/apache/celeborn/pull/2609. The PR you are referring to in the comments is a test I wrote for another...
The test failure is not related to this change. +CC @waitinfuture, @ErikFang
+CC @RexXiong, @waitinfuture Please see the [fix here](https://github.com/apache/celeborn/pull/2639/commits/dcd8465dd98ee468de8579acf28a0234ce391fdd) for the determinate stage failure case we are discussing [here](https://github.com/apache/celeborn/pull/2609#discussion_r1697821776). I would prefer decoupling the two changes though - one specifically for...
Actually, let me think a bit more about the determinate stage case though ... there might be corner cases here that need handling (partial recomputation will have an issue on...
I have reverted the changes for determinate stage handling here @RexXiong, @waitinfuture - and limiting it to barrier stage alone. Let us handle task reattempt for determinate stages in #2609....
Gentle ping @waitinfuture, @RexXiong
Thanks for the review @waitinfuture, have addressed comments.