Archive OpenMetrics?

[Richard Hartmann]

Dear all,

I have written and re-written this email I don't know how often, and I
have put off writing it off even more often.

Various parts of CNCF have become more pushy so it's time to ask:
Should we ask CNCF to archive OpenMetrics?

It's become a drag on Prometheus, OpenTelemetry is certainly more
successful, and we can still re-use the spec as the basis for
Prometheus Exposition Format.


Ideally reply with YES or NO.


Best,
Richard

[Alexis Richardson]

Yes 

--
You received this message because you are subscribed to the Google Groups "OpenMetrics" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openmetrics...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openmetrics/CA%2B_k06MV%2BnJvzpoAeYNJ8SmHGSwGSRp9yrtLUp4SyWrAwnHU%3DQ%40mail.gmail.com.

[Richard Hartmann]

Dear all,

I selected the wrong mailing list by mistake, I had intended to write
to the maintainer-only one.

Done's done, and maybe this is the better choice in hindsight anyway.


Best,
Richard

PS: Given that Alexis is one of the maybe first one or two dozen
people on Earth I talked with about this project idea many many moons
ago, it feels as if we've come (almost) full circle.

[Kevin Wiesmueller]

I still think there is a need and a place for a standard format that brings projects together and improves on the Prometheus exposition format to allow for systems other than Prometheus to be part of the ecosystem efficiently.

Maybe OM was too close to the Prometheus format, maybe it's just a lack of backing, investment, or need. If OM dies I'd like to understand how we can bring some of the improvements made with it to a reusable format.

Not having an open standard like OM and relying on Prometheus instead means that for every idea there is a valid argument that "Prometheus doesn't need this feature" which I worry will keep the ecosystem inflexible to change. Maybe the answer to that is OTel.

To view this discussion on the web visit https://groups.google.com/d/msgid/openmetrics/CAOSi4U4MXLxZmWMHPO535WJXsHcP-OOAS16ohr1DMNvW%2BMoh3g%40mail.gmail.com.

[Ilja Shmelkin]

Hello you all,

I have written my dissertation thesis closely related to monitoring. 

Over the years when I was doing research, I really had noticed the need for standardized monitoring formats, e.g. metrics, logs, checking, etc.

I even cited OM for being a promising project towards finally bringing some standardization into monitoring.

I don’t really know why the idea of archiving is even in the table but standards are definitely needed. 

Viele Grüße 

Dr. Ilja Shmelkin

Am 23.06.2023 um 16:39 schrieb 'Kevin Wiesmueller' via OpenMetrics <openm...@googlegroups.com>:



To view this discussion on the web visit https://groups.google.com/d/msgid/openmetrics/CAOx_-b68O2z%3DjBC0V_%3DM2niW8ArGDHO9TScokON5k0S%2B7psWHQ%40mail.gmail.com.

[Richard Hartmann]

It has been pointed out to me that some context is missing. In an
internal email to a few people, more context can (and was) assumed
than in an email to a public mailing list. Let me expand.

My thinking is along the lines of "Fold the spec in part/in full into
Prometheus, release a Prometheus exposition format 1.0, base PEF 1.1
or 2.0 on top of that with native histograms and a few other cleanups,
and then go from there".

My main intention is to regain speed which we lost; with my Prometheus
hat on, I believe that we owe this to the wider ecosystem.

[Richard Hartmann]

Having a distinct standard is part of why CNCF asked me to spin it out
into its own separate project all those years ago.

Prometheus has become a lot more willing to support
non-direct-Prometheus use cases, and I'd expect us to do the same
thing there.

OTLP metrics exists and is compatible; though for anyone within the
Prometheus ecosystem it will likely be cheaper & quicker to use an
opinionated & optimized wire format over a lingua franca one.

[Alexis Richardson]

+1

[Lee Calcote]

“...ago, it feels as if we've come (almost) full” moon. :)

> --
> You received this message because you are subscribed to the Google Groups "OpenMetrics" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to openmetrics...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/openmetrics/CA%2B_k06MjqePJhc_XRGb7iMLxmAODBed93ewFRpXpO%3DAhWgEnsg%40mail.gmail.com.

[Richard Hartmann]

On Fri, Jun 23, 2023 at 6:17 PM Ilja Shmelkin <ilja.s...@gmail.com> wrote:

> I even cited OM for being a promising project towards finally bringing some standardization into monitoring.
>
> I don’t really know why the idea of archiving is even in the table but standards are definitely needed.

I fully agree. It's not about stopping standardization, it's about
speeding up the standardization process.

At the same time, part of being honest is that https://xkcd.com/927/
applies to all standards. That being said, Prometheus is something
~everything in CNCF aligns itself to, so no matter the name, there's a
standard to speak.

[Alexis Richardson]

CNCF is meant to be an incubator and accelerator for code that -- if
people want to -- they could standardise around. But CNCF is not a
standards body. It is ok to adjust the strategy

[Rob Skillington]

I would love to see it become a standard and an IETF RFC since it would likely help it become a long lived standard. However if OpenMetrics being an active CNCF project however does not necessarily help support that goal then I agree that perhaps retirement as a CNCF project makes sense.

Or do we think it being standardized should come later after evolving some more in Prometheus again?

Curious on long term thoughts of how this looks longer term rather than purely just the here and now too.

--
You received this message because you are subscribed to the Google Groups "OpenMetrics" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openmetrics...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/openmetrics/CAOSi4U5s7pLgSNpto0J8Kt8KD52-v4CJE7MMkmk-ZAR097OrAA%40mail.gmail.com.

[lei...@gmail.com]

I have been doing monitoring consulting part-time for a long time and
participated in IETF before. IETF is doing standardization of syslog.
But I think it's unlikely it would take on open-metrics and monitoring
unless it is a de-facto standard and effort is supported by all
vendors. It is not at that point but things could lead there. The need
for monitoring standardization has existed for at least 20 years but I
have not seen sufficient traction until Open Metrics and Open
Telemetry. However quickly things began moving to vendor specific
protocols and clients in the fast growing SAS cloud monitoring field -
with 3 -4 vendors competing head-head there I don't see them wanting
to cooperate right now even though everyone supports Open Telemetry
and their solutions are all based around this concept.

> To view this discussion on the web visit https://groups.google.com/d/msgid/openmetrics/CABakzZbqtFSQU_Qo2%3D9vfaS%3D8g%2BabWD%2BA-BneuUum2GGZzqK5A%40mail.gmail.com.

[Rob Skillington]

How would we proceed Richie to begin archiving the project?

I’m supportive.

To view this discussion on the web visit https://groups.google.com/d/msgid/openmetrics/CAFCy1BiG6ZVpuYUejipDu6K9MizP8mp1EzLN00x3_1udBSCYYA%40mail.gmail.com.

[Ben Kochie]

Sorry for the long delay, I've been thinking about this quite a bit over the last couple months.

I think it's time to retire OpenMetrics as a stand-alone project. There just isn't enough support behind it as it is.

I would like to see OpenMetrics work picked up by the Prometheus community going forward. There's still a need to have a good quality, used in production, open standard.

To view this discussion on the web visit https://groups.google.com/d/msgid/openmetrics/CABakzZbqtFSQU_Qo2%3D9vfaS%3D8g%2BabWD%2BA-BneuUum2GGZzqK5A%40mail.gmail.com.