Kubernetes SIG Instrumentation, 15 Sep 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SIG Instrumentation 20220915

Description

SIG Instrumentation Bi-Weekly Meeting Sept 15th 2022

A

This meeting is being recorded.

B

Hi everyone welcome to Sig instrumentation uh bi-weekly meeting. This is uh September 15th of 2022.

B

So here let me share my screen.

A

B

So uh we only have a couple things on the agenda, which is.

B

Caps, so cap freeze is coming up. Does anyone have anything that they want to propose? Because if so, uh they need to get prr review by into in the next two weeks and enhancement? Freeze is in three weeks.

B

So from the looks of it, we have a couple things in the pipeline. I have two things uh uh it sounds like Damien is just going to get rid of these.

A

B

And I am not sure what the status of contextual login is here.

C

Alpha yeah we we should reach out to Patrick and or Merrick to see what their plans are.

B

Oh they're going to try to go to Beta on 126..

B

A

A

B

And what about this one David: are you on top of this one here.

C

Which one is that.

A

A

C

B

uh The cubelet of inflammatory tracing.

C

uh I think cubelet. We should leave an alpha for another release.

A

C

Yes, I can I can ask Sally if she has plans, um but.

B

So we just want to bake it a little bit longer.

C

Yeah I think so I think I'd like to give people a chance to try it out in Alpha before we turn it on by default. Okay,.

B

I will reach out to Merrick about uh the structure, logged and stuff, because we have a ton of structured logging and stuff what about what about APS server tracing? Do we want to leave this in.

C

Oh that one we should definitely graduate to Beta I. Think.

B

um Okay, so do you have a cap for this.

C

The the cap already says it's going to Beta, but in like three releases ago, so I just need to update the version. um I'll make sense is.

B

There any work that needs to be done.

C

uh Yes, so we successfully updated the version of open Telemetry used in etcd.

A

C

We did get a doesn't need to be in three four three.

B

Five, it does need to be a three.

A

C

I'll double check, I.

A

B

C

Yeah I thought it was in 3.5, but.

B

Okay, let's, let's discuss it offline, but I I'm, pretty sure we need it in three four.

B

um So it does so. There aren't many things on the agenda here, but there seems to be a couple more things document log. So it's a sub project. Okay, yeah.

A

I'll just assign this to myself.

B

And then uh that and then KSM planning.

A

uh Can we put this on um hello, everyone, so yeah I I I did um hello. Everyone Hi am I, ordered yeah yeah. You.

B

D

Okay yeah, so you might remember me from the last meeting. I was I'm just basically working on KSM these days um and uh I I was tracking this board and I uh uh basically I had the spls. For uh you know, almost all these uh issues and some more you know some minor issues on as well, but uh what I wanted to talk about was uh there's this.

D

Fourth, uh this is issue uh number four on the on the board on on that same board that we were on yeah, so I was just proposing that you know, because this is something that is that this is something that you know is maybe a bit too much for the single person. So maybe we can, you know, put this out as a good first timer and you know attract more contributors.

D

Instead of you know a single person taking it up.

B

uh Yeah, who normally leads the case some stuff, it's Damian right, yeah,.

D

B

uh Have you reached out to Damian about this.

D

um Not not particularly about this, though,.

B

I would reach out to Damian directly. um I am I'm, not super involved with KSM stuff I mean I I, don't object to what you're saying I think it makes sense.

B

um Okay, there might be a protocol in like promoting these metrics right, like they might be, based off of a certain amount of time.

D

Yeah so I think Catherine sent uh multiple PR's for uh promoting uh some metrics from experimental to stable, okay and yeah I. Think I think it's just the usage. If, if you know it makes sense if it, if it's been there for a while um but yeah she would, she probably knows better yeah yeah.

E

I think another relation issue is that I put in this PR. We need to have a mechanism to detect it's similar to what we did in for the uh kubernetes repo, but for case and repo. We had no check to check a weather. This stable measures has been modified by some contributors or not it's just a document in the in the doc file. Okay, this one is stable. These are not stable, but no pre-submitted checks.

D

Yeah I saw that comment yeah that makes sense.

B

So I mean if someone wants to Port over the stability framework to KSM I mean like it's just a wrap around Prometheus, so it should theoretically just work. um It depends how many weird metrics types KSM supports.

B

uh If they're doing a lot of like gauge funks custom collectors, then it's basically going to be a lot less useful.

E

Yeah, do we have some plan to promote these two or two Prometheus, because I think not only this rifle as a rival, for example, for example, LCD also, using maybe permission library to export metrics. So if we can have this Library information, so I.

B

Don't think Prometheus really wants this. This is like very specific to um yeah, versioning, metrics and stuff. uh We do it in kids because, like people depend on kids but for normal metrics, it's a little.

B

Yeah I I mean I I, brought it up to Prometheus guys before and uh I, don't think they were interested in upstreaming it so.

B

So it would make more sense just to pull in the the component-based library in 2ks I've been doing that there. If you wanted to do something like that.

A

Yeah makes sense, yeah.

B

But it it can be a little bit involved. You would need new static analysis, scripts and stuff.

A

D

Just one more thing that I wanted to link up just to get Focus opinion on it.

D

um Basically, this comment right here, so it's uh regarding the placing the CLI toolkit, so we would be um duplicating kind of just you know, um being independent of the uh P Flags library and uh instead incorporating cobra um for the flags wiper, for you know, watching config files and um basically the loading um hard thing loading whenever there's a config there's a change in the config files or whatever and then uh Damien and I.

D

uh We actually talked about this uh regarding the future plans, where we want to reload on the fly, so we won't. We won't restart the binary, but we would just you know, um invalidate the cache or just you know, just uh the word, those specific areas as soon as we detect that there's any uh you know, config uh changes uh in the in the supplied. You know config, Maps or files or whatever so.

D

um Yeah, it's the uh I, think it's the third Point uh that Damien commented um at the bottom. I can actually.

A

D

Yeah the future plans Beyond version three yeah, the first link there yeah.

D

So that's basically.

A

This this one, um no, the one, uh okay, let me just uh this one.

B

Ability to configure this uh okay.

A

Yeah yeah yeah yeah.

B

I'm cool with this.

A

D

The first two paragraphs that he talked about- that's already been done: the context cancellation logic, it's already there in the wiper PR, but yeah. The third point is a bit more complex, but it would definitely you know it's worth the effort and um we kind of decided to do that. uh You know after version three, if that's fine, I, actually I wanted to ask this thing: uh where uh is that board accurate? So if, let's say all those PR's emerged all those issues at this, then we go ahead with the version 3 release for KSM.

B

uh I am not the best person to ask about that, but I would assume yes.

B

This right, you're talking about this right, yeah uh I, would assume yes for version three yeah I, think that makes sense.

A

B

uh But keep in mind I'm not like a cube State metrics domain expert, so I deferred to Damien, but he's not here.

A

Yeah yeah I'll thank him. Yeah.

B

All right so I think that's it for our agenda today.

B

um Thank you, everyone for attending. If does anyone, have anything else that they wanted to to pop up as a topic.

B

Okay, great then I will give everyone 15 minutes of their time back. um If you guys have time, please review our caps. um We would appreciate feedback on them. um Since cap deadline is approaching, um cool well have a nice week. Everyone see you guys next time, bye.