Thanks for the KIP, Patrick! It looks like you addressed Guozhang's and Bruno's very good feeback, and I like the result. The example especially helps clarify how this property might be useful.
I'm in favor of this proposal. Thanks, -John On Fri, 2021-11-05 at 12:03 +0100, Bruno Cadonna wrote: > Hi Patrick, > > Thank you for the KIP! > > - Maybe some more details in the motivation would help to better > understand the background of the KIP. Currently, it is hard to judge > whether record metadata should be exposed or not. Can you maybe give an > example? > > - Could you please replace RYW abbreviation with read-your-writes (at > least that is my guess about the meaning of RYW)? > > Best, > Bruno > > > > On 03.11.21 22:43, Guozhang Wang wrote: > > Thanks Patrick, > > > > I looked at the KIP and it looks good to me overall. I think we need to > > double check whether the record metadata reflect the "last processed > > record" or the "currently processed record" where the latter may not have > > been completely processed. In `ProcessorContext#recordMetadata` it returns > > the latter, but that may not be the preferred case if you want to build the > > consistency reasoning on top of. > > > > Otherwise, LGTM. > > > > > > Guozhang > > > > On Wed, Nov 3, 2021 at 1:44 PM Patrick Stuedi <pstu...@confluent.io.invalid> > > wrote: > > > > > Hi everyone, > > > > > > I would like to start the discussion for KIP-791: Add Record Metadata to > > > State Store Context. > > > > > > The KIP can be found here: > > > > > > https://cwiki.apache.org/confluence/display/KAFKA/KIP-791:+Add+Record+Metadata+to+State+Store+Context > > > > > > Any feedback will be highly appreciated. > > > > > > Many thanks, > > > Patrick > > > > > > >