Hi all,
During the review of https://github.com/apache/kafka/pull/21555, we have
opted to rename the executor configs and metrics.

*Configs*

   - group.coordinator.executor.threads ->
   group.coordinator.background.threads

*Metrics*

   - executor-thread-idle-ratio-avg -> background-thread-idle-ratio-avg
   - executor-queue-time-ms-(p50|p95|p99|p999|max)
   -> background-queue-time-ms-(p50|p95|p99|p999|max)
   - executor-processing-time-ms-(p50|p95|p99|p999|max)
   -> background-processing-time-ms-(p50|p95|p99|p999|max)

Thanks,
Sean

On Fri, Feb 13, 2026 at 8:00 AM David Jacot via dev <[email protected]>
wrote:

> Thanks for the update, Sean. The KIP looks good to me.
>
> Best,
> David
>
> On Fri, Feb 13, 2026 at 8:27 AM Sean Quah via dev <[email protected]>
> wrote:
>
> > Thanks for the feedback.
> >
> > dj08: I wonder whether we should use 1s for the interval combined with
> > > offloading enabled by default. The offloading preserves the coordinator
> > > threads while providing back-pressure because a group can have only one
> > > assignment inflight. Going from 5s to 1s would ensure that groups
> remain
> > > more dynamic.
> >
> > I've updated the default to 1s.
> >
> > dj09: Should we have min/max configs on the server for the new interval,
> > > similar to other intervals?
> >
> > I've added those too. I've renamed min.assignment.interval.ms to
> > assignment.interval.ms, otherwise the names look weird.
> >
> > dj10: What's the reason for using -1 by default for the Timestamp field?
> It
> >
> > seems we can just use zero.
> >
> > I wanted a value to indicate "no timestamp". 0 works just as well as -1.
> > I've updated the KIP.
> >
> >
> > > dj11: I wonder whether we should have a plan to sunset
> > > the *.assignor.offload.enable=false too. I believe that this will hurt
> > > users more than anything else. We could perhaps mention this in a
> future
> > > work section.
> >
> > Updated the Compatibility, Deprecation, and Migration Plan to mention
> this.
> >
> >  dj12: In the migration section, we should also mention the offloading
> >
> > enabled by default.
> >
> > Updated the Compatibility, Deprecation, and Migration Plan section
> >
> > Thanks,
> > Sean
> >
> > On Thu, Feb 12, 2026 at 1:08 PM David Jacot via dev <
> [email protected]>
> > wrote:
> >
> > > Hi Sean,
> > >
> > > Thanks for the update. I have a few more comments to consider:
> > >
> > > dj08: I wonder whether we should use 1s for the interval combined with
> > > offloading enabled by default. The offloading preserves the coordinator
> > > threads while providing back-pressure because a group can have only one
> > > assignment inflight. Going from 5s to 1s would ensure that groups
> remain
> > > more dynamic.
> > > dj09: Should we have min/max configs on the server for the new
> interval,
> > > similar to other intervals?
> > > dj10: What's the reason for using -1 by default for the Timestamp
> field?
> > It
> > > seems we can just use zero.
> > > dj11: I wonder whether we should have a plan to sunset
> > > the *.assignor.offload.enable=false too. I believe that this will hurt
> > > users more than anything else. We could perhaps mention this in a
> future
> > > work section.
> > > dj12: In the migration section, we should also mention the offloading
> > > enabled by default.
> > >
> > > Best,
> > > David
> > >
> > > On Tue, Feb 10, 2026 at 10:14 PM Sean Quah via dev <
> [email protected]
> > >
> > > wrote:
> > >
> > > > Thanks all for the feedback so far. I will start a vote within the
> next
> > > few
> > > > days unless anyone raises further feedback.
> > > >
> > > > Thanks,
> > > > Sean
> > > >
> > > > On Mon, Jan 26, 2026 at 10:42 PM Dongnuo Lyu via dev <
> > > [email protected]
> > > > >
> > > > wrote:
> > > >
> > > > > Thanks Sean for the update!
> > > > >
> > > >
> > >
> >
>

Reply via email to