Hi Venkata,

Thanks for starting the discussion.

On Spark 4.1: the support currently lives only on master and will be part of 
the 0.7.0 release (CELEBORN-2239, with the CELEBORN-2240 adaptation for 
SPARK-51756). I'd prefer to ship it there rather than backport it into a 0.6.x 
patch release — supporting a new Spark version line is a feature rather than a 
bug fix, so 0.7.0 is the right home for it. Users who need Spark 4.1 can pick 
it up once 0.7.0 is out.

On a 0.6.4 release: looking at branch-0.6 since 0.6.3, there's effectively only 
a single bug fix accumulated so far — CELEBORN-2253 (IndexOutOfBoundsException 
when reading shuffle data from HDFS). That alone doesn't feel like enough to 
justify cutting a new patch release right now. I'd suggest we let a few more 
fixes land on branch-0.6 first, and revisit 0.6.4 once there's a more 
meaningful set of changes to release.

Regards,
Nicholas Jiang

On 2026/06/27 04:11:20 Venkatakrishnan Sowrirajan wrote:
> Hi Celeborn community,
> 
> It has been close to 2 months since the 0.6.3 release. Curious, would there
> be a 0.6.4 with Spark-4.1 support or it will only go in 0.7? I don't see
> Spark-4.1 support in branch-0.6 though.
> 
> If it is something that can be released as part of celeborn-0.6.4 release,
> happy to backport it.
> 
> Regards
> Venkata krishnan
> 

Reply via email to