+1 (non-binding)

Thanks Aravind, glad to see this shaped into a CIP. The client-side
approach has real advantages - sendfile path stays intact, workers stay out
of key handling, clean Spark integration.

Worth flagging for the record that CIP-22 is narrower than CELEBORN-2311
<https://issues.apache.org/jira/browse/CELEBORN-2311> it's Spark-only
(Flink/MR clients uncovered), and spark.io.encryption derives keys locally
with no KEK/DEK split or external custody - which is the part that matters
for the HIPAA/PCI/SOC 2/FedRAMP deployments I had in mind. The KeyService
SPI with pluggable KMS providers from CELEBORN-2311 is still needed for
those.

Suggest landing this as the Spark path and keeping CELEBORN-2311 open as
the umbrella for the engine-agnostic worker-side track. Happy to drive that
once this lands.

- Karthik

On Wed, May 6, 2026 at 10:16 AM Mridul Muralidharan <[email protected]>
wrote:

> +1
>
> This is based on what we are currently in the process of rolling out
> internally in production.
>
> Regards,
> Mridul
>
> On Wed, May 6, 2026 at 1:53 AM Aravind Patnam <[email protected]>
> wrote:
>
> > Hi all,
> >
> > This is a call to vote to contribute CIP 22: Encryption at Rest for
> > Celeborn Shuffle Data
> > <
> >
> https://docs.google.com/document/d/1xBrLtpb8bk8CdJENiM3aLJCbFThRKLor8uAjSsfXoxE/edit?usp=sharing
> > >to
> > Apache Celeborn.
> >
> > Please do vote accordingly:
> > [ ] +1 approve
> > [ ] +0 no opinion
> > [ ] -1 disapprove (and the reason)
> >
> > Thanks once again!!
> >
> > --
> > Aravind K. Patnam
> >
>


-- 
Thanks,
Karthik

Reply via email to