Hi CHia-Ping,

Thank you for the review and joining the discussion. I agree that Kafka
native metrics are the right long-term home for these and that the existing
Yammer metrics for linux-disk-read-bytes and linux-disk-write-bytes should
eventually be migrated. However, I'd like to propose separating these
concerns:

KIP-1291 <https://cwiki.apache.org/confluence/x/co48G> (this KIP): Remain
purely additive - expose the 5 missing metrics under the existing
kafka.server:type=KafkaServer,name=linux-disk-* convention. The KIP's
stated guarantee is zero operational impact and no migration burden for
existing deployments.

Follow-on KIP: Once the 5 metrics are established, a separate KIP can
migrate all 7 linux I/O metrics to a dedicated native metrics group (e.g.,
type=linux-io-metrics) and formally deprecate the Yammer equivalents with a
clear timeline.

My concern with combining them in KIP-1291
<https://cwiki.apache.org/confluence/x/co48G> is that deprecation , even of
metrics that remain functional , is not a zero-impact change for operators
running dashboards and alerts against the existing MBean. It also opens
discussion on naming conventions, deprecation
timelines, and coordination with broader Yammer migration work, all of
which could delay shipping the 5 metrics that are the core of this KIP and
that operators need today.

I'm happy to file the follow-on KIP alongside this one so the community can
see the full intended path. Would that be acceptable?

Best,
Sahil Devgon

On Sat, Apr 4, 2026 at 1:22 PM Chia-Ping Tsai <[email protected]> wrote:

> hi
>
> I am concerned about adding more Yammer metrics. Instead, we should create
> a new metrics group (e.g., type=linux-io-metrics,cluster-id=aaa) for all
> seven metrics using Kafka native metrics. This way, the seven parameters
> can be grouped together as a single MBean. We can then deprecate the
> original Yammer metrics for linux-disk-read-bytes and linux-disk-write-bytes
>
> Best,
> CHia-Ping
>
> On 2026/02/25 14:50:44 Sahil Devgon wrote:
> > Hi,
> >
> > I would like to start a discussion thread on KIP-1291. In this KIP, we
> aim
> > to expose all 7 Linux I/O metrics from /proc/self/io instead of just the
> > current 2 (read_bytes and write_bytes).
> > The 5 additional metrics (rchar, wchar, syscr, syscw,
> > cancelled_write_bytes) enable operators to diagnose cache effectiveness,
> > write amplification, and I/O pattern inefficiencies.
> >
> > https://cwiki.apache.org/confluence/x/co48G
> >
> > Please review the KIP and feel free to share your thoughts.
> >
> > Thanks,
> > Sahil Devgon
> >
>

Reply via email to