Hi Jiunn-Yang,

Thank you for the great questions.

JY1: As described in KIP-631, node.id is mainly for configuration
convenience — broker and controller remain two distinct roles internally.
The scope of this KIP is limited to the user-facing configuration. Whether
to standardize all usage of brokerId() vs nodeId(), introduce a
controllerId(), or fully migrate to nodeId() is a broader discussion that
deserves its own KIP.

JY2: I think broker-role metrics using BrokerId as a tag can remain as-is
since the broker role still exists. However, I do notice that the
controller also exposes kafka.server:type=registration-metrics, tagged with
BrokerId because it shares the same NodeToControllerChannelManagerImpl code
path. I've opened KAFKA-20199 to track this. As with JY1, deciding whether
to rename the tag to NodeId or ControllerId deserves its own KIP.

Best,
Ming-Yen

黃竣陽 <[email protected]> 於 2026年2月17日週二 下午5:51寫道:

> For JY2, the metric tag name is BrokerId, not broker.id. Should we update
> the tag name to NodeId to align with this change?
>
> Best Regards,
> Jiunn-Yang
>
> > 黃竣陽 <[email protected]> 於 2026年2月17日 下午5:46 寫道:
> >
> > Hello Ming-Yen,
> >
> > Thank you for the KIP.
> >
> > JY1: It appears that some broker.id accessor will remain even after the
> configuration is removed.
> > In which version do we plan to remove the remaining broker.id accessor?
> >
> > JY2: In the selector, we still use broker.id as a metrics tag. Should
> we also deprecate these metrics to
> > align with this configuration change and migrate them to use node.id
> instead?
> >
> > Best Regards,
> > Jiunn-Yang
> >
> >> Ming-Yen Chung <[email protected]> 於 2026年2月17日 上午10:54 寫道:
> >>
> >> Hi all,
> >>
> >> I would like to start a discussion on
> >> KIP-1232: Deprecate broker.id config
> >> <https://cwiki.apache.org/confluence/x/Hgp3Fw>
> >>
> >> With ZooKeeper removed in Kafka 4.0, node.id is the canonical node
> >> identifier for all process roles. This KIP proposes to deprecate
> broker.id
> >> from the server configuration in 4.3 and remove it in 5.0.
> >>
> >> Looking forward to your feedback.
> >>
> >> Best,
> >> Ming-Yen
> >
>
>

Reply via email to