Kubernetes SIG Instrumentation, 2 Sep 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SIG Instrumentation 20210902

Description

SIG Instrumentation Bi-Weekly Meeting Sept 2nd 2021

A

All right welcome everyone. This is sick. Instrumentation today is september 2nd, and we have uh two items on the agenda. The first one is by han. If I understood correctly, I believe han can't make it today. I think he he said he yeah. He can't make it today. So I guess we'll go right ahead to our second point, um which is kept review for 1 23.

B

A

Want something.

B

Yeah I can take this one. I think I put this on the agenda yeah. um I went ahead and if I can just pull up the notes, uh if you could give me co-chair, I will share my screen.

B

So we have this dock for cap planning and I noticed that I don't think we've done anything and freeze is next week, so um I copied over everything from last release and.

A

You should be good.

B

Thank you. I copied over everything from last release for 123 and I think we should probably discuss um so like here's. The list of everything, uh and I think we need to make an in-out call uh today. Also is uh the soft deadline for production readiness.

B

So uh if we want to include things in 123 like the questionnaire should be done, uh and so given that I mean it's still possible for things to make it even if it's not yet done, that's just a suggested sort of soft deadline, uh but it will be harder, especially for net new things. If it's not done um so, I know that merrick wanted to get this one into 123..

B

uh I have not heard anything else from him. Does anybody have any concerns or, like plus minus, to inclusion on this one.

C

Plus, it seems like a thing to do sooner, rather than later.

B

Yeah, especially given deportation cycles, so I did. I did an initial quick review of it. I was missing prr, so I pointed that out uh and I think that's fine to alpha in 123.

B

uh next we have metric cardinality enforcement uh that one's been sitting in alpha for a bit. um Is there any plans to bring that one to beta? I know that we we had this to do from last time to update the cap and I don't think it's been updated yet.

A

No, it hasn't been.

B

I uh suspect, on a lot of these, we're going to be running into bandwidth issues, um but uh we'll we'll see what there is to see.

B

uh And, and is there anybody on the call who would be interested in picking that one up or uh otherwise we'll just let it lie.

B

Okay sounds like no um defending against logging sneakers by static analysis that one's not a code change, and I saw that they indicated they wanted ga at this release. So I think uh that one's been taken over by security and possibly testing as well. So I don't think there's much input from us. um This one had a note that it would ga this release, uh so I haven't heard from clayton uh as to whether or not it should be gaining, but we maybe want to put ga question mark beside this. One.

A

So, as far as I'm aware, none of the like uh major dashboard and alerting rule collections have picked up these these metrics. So I would question their um like how much they've been used that we are able to validate it, able to ga it as much as I would like them to be stable?

A

I'm not sure I I would have that confidence today.

B

Yeah- and I mean to to be clear- I guess gaying this- I'm not sure what would be involved in that because it wouldn't involve bumping the metrics to stable. Yet uh because, uh if we're talking about the framework that we discussed with sig architecture, ga would trail the feature going. Ga anyways so.

A

I mean in this case the feature is literally the the metrics right.

B

A

The scheduler features do exist already, it's just about creating the metrics that reflect the scheduler state.

B

Yeah, so I guess uh someone, let's we'll see, if maybe we can.

B

I'll put a note on here.

B

uh And then uh this tracing thing david? Do you know if this is ready for 123? I have? I saw there was some back and forth on the cap, but it didn't look good to go yet to me.

C

For 123, as in for beta.

B

uh No, this is still net new. This is the unique logging metadata. Oh.

C

Sorry, sorry, um uh that's blocked on me right now. um Okay, I would put it as net new in 123..

C

um We still haven't settled on a design that I think meets the criteria given to us by the sig api machinery. Folks that previously reviewed so um I'll have to take another look. They have a new iteration, but I haven't had time to review yet.

B

Okay, so that one's waiting on you so I'll put that as alpha question mark um structured blogging, I have not heard anything about like I saw that they talk. They had an issue about migrating another component, but I haven't seen any updates to the cap.

B

Does anybody know what's going on with this.

C

No, maybe just a sign, I don't think I saw merrick on the call, so maybe just a sign.

B

Yeah, I think him earlier, and I know that this morning uh or my morning there was a working group. Logging call so I'll, maybe go and try to check their notes and see what the status is.

B

uh If you guys are retracing, is that going to beta.

C

No I'd like to keep that in alpha for 123 and plan on beta and 124.

B

B

uh Redesigned event api, um so this one has been kicked down the road, a bunch and continues to get kicked down. The road and uh oh.

D

There was a to-do.

B

uh Do we have you on the call.

D

Yes, yes, I'm here so I reached out to to chelsea and it looks like she does not have the capacity to work on this.

B

So that's been three cycles in a row that she has not had capacity to work on this. We don't really want to leave this in perma-beta. We have like something like half of the core components have been migrated to the new framework. We can't ga it until everything's migrated.

B

uh I'm not really sure what we should do with this one and I've been seeing prs come in to migrate components, but then, like there's no cap, so.

D

um You know what I know, someone in my team who may want to take over this one. So let me like steal a new to-do, but it's not too uh um it's it's another person instead of a chelsea I'll I'll talk to her, and I I think she might be like interested in taking over this one. So.

B

Great, I'm gonna also put you in here as our an owner.

D

Yep, that's fine.

B

And then last one here, uh so this is this kubernetes system components, log sanitization, the runtime sanitization. So in theory this was punted over the fence to seek security, but there haven't been any updates from them and it's been an alpha for over three releases.

B

Now uh I don't think anyone's using it, and I still like I asked them last release like please update this cap or I think we should deprecate it because I don't think anybody's using this and we had concerns initially when we reviewed this proposal as to whether or not we even wanted this in the first place, um I haven't heard anything uh it I mean somebody said that they were a new owner, but like there's, no kep updates, uh so I would say maybe we should just deprecate. This is anyone using this.

C

I believe gk uses this.

B

I see I think if gke is using this and gke does not want this to be deprecated. Then gke really needs to graduate this to beta, because it's just been sitting there as an alpha feature.

C

I agree, um I think fantu who's on that is still the correct person um who should have an interest in keeping this around so.

B

C

B

The issue like had gotten so neglected that had rotted and closed, which is why I think it had gotten missed. um So I really think we need to graduate this, because I mean it's just been sitting.

B

I think it went uh alpha in 120 and then it basically hasn't been worked on since so.

C

Okay, I'll all bug thunder.

C

Yep then uh we'll go from there.

C

But I agree: we should make progress on this or deprecate it soon.

B

Duke I will assign this to you.

B

Okay, I'll put a question mark there as well, since it's not confirmed so, I guess.

D

B

Only thing we've confirmed that we wanted in 123. Is this one, so I'll add the milestone to that.

B

And we have a lot of follow-ups to do.

B

Yeah we've been getting some things from the um okay, this one's already in the milestone good.

A

Has has han indicated whether he wants to get the revisiting of the metric stability classes in, for.

B

Yeah, uh I know that's a great point and I will throw this on here.

B

uh It would be net new and.

C

B

I haven't heard anything from him I think he's slammed. So I.

A

Mean like all all my time that I can uh find I'm working with han on this, but like it's, it's been difficult enough to figure out the details, and really most of this is rolling out the process right. So it's also really hard to it's. Just just like the initial metric stability classes, it's hard to say when this is alpha beta, unstable.

A

Right because technically, our metric stability classes in the first place are only in beta right now, because we haven't actually g8 any metrics.

B

No, no metric stability, all right.

A

B

G8 last release.

A

B

Yeah, uh I think, um and by last release I mean 121..

B

We added a bunch of steering stable things. Yeah time keeps passing yeah. We do have some stable things so that that is ga, that's done, but yeah. I agree that uh so if this is process based, we maybe don't necessarily have to follow the like entry cycle as much. I think it's basically like anything that's in tree in terms of changes. Those are what we really have to follow: the ta, the normal cap, time frame for so.

A

So there was some entry stuff that tan wanted to do um or that we wanted to do. um I mean, first of all, because we're introducing new class stability classes that would be intriguing and then there's some uh like static analysis that we wanted to add. um That is partly done.

B

Well, the static.

A

Analysis would.

B

Because that's static analysis, uh it wouldn't necessarily need to be like the fact that we're following the sort of standard entry cycle for the static analysis for security stuff, like that's, really a formality. That's not, I think, a hard requirement, but yeah.

D

Definitely for.

B

Like new metric stability classes, that probably will have to follow um the standard kept cycles, so um I suspect, excuse me this won't be ready just because I think that it'll take some time for like prr to review this, because I get architecture, sign off, etc and we're only a week away, and I know that han is still really busy um and probably won't have time to work on this until next week. So uh I guess.

A

B

This will probably be 124.

A

I I think so as well. We have a really rough, like draft of a of a cap that we started to write, but um it's it's far from like, like we haven't, even opened the pr for it right.

B

Yeah, so we've got the two things we know are going to 123 and I think everything else is kind of up in the air. This one. I will apply the milestone.

B

And this one says it's approved, uh it was.

B

Approved but not lg tm, uh so I can go back and lgtm this one and then uh we should be good. uh I don't have to seek security through this in the collection spreadsheet, but if not, um then we can do that.

B

Okay, uh I think that's everything I have for caps. So uh thank you. Everybody for your patience going through this every cycle always got lots of features to work on uh that was, I guess all I had for that agenda item. Let me make sure that we have everybody on the attendance.

A

I think that was it in terms of agenda items for today, since han couldn't make it. So unless there is anything that anybody wants to discuss now, we could call it a day early.

A

Anything else that anybody wants to discuss.

B

I think we sounds like we're good. I will stop the recording.