Kubernetes API Expression Working Group, 8 Dec 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: WG API Expression Bi-Weekly Meeting for 20201208

Description

WG API Expression Bi-Weekly Meeting for 20201208

A

As usual, I always record to cloud and and send the recording to somebody who helps me. Okay, I'm recording.

B

Now, hello, welcome to server side, apply, no working group uh api expression.

B

Let's go, do you want to go? uh Do you want to drive kevin.

A

Yeah sure um um yeah I read through the document and uh thanks again jeff for for building it up, um I think, can we skip the table of contents, which is good tasks? I.

A

Think I I saw I saw the cap on the api clients. um I wanted to make comments for some typos, so it's all, but aside from that it sounds pretty good.

A

B

Yeah jenny's, not here is she.

A

uh I no no, it's only only the four of us.

A

uh Yeah, so from that, I I guess we we have to wait on on feedback with that.

A

um What's that with this one, with ssa, ctrl.

B

um It's the same girl yeah. We went to the ucla meeting uh with julian, and I mean we: don't they don't have specific feedback. I think everyone's happy with the current situation.

B

um The we mostly spend time talking about the manage fields uh being too verbose and so created two additional tasks at the top of the of this one is uh hiding it from the api server to reduce the size of the requests, and one is about hiding it in cube, curl and um so julian's. Working on the cube, coral part, which I don't expect to be very difficult and uh knight 42, is working on the api server part um that doesn't.

A

Look pretty good.

B

A

B

Much simpler than expected, yeah.

B

So yeah, that's it! For this part, these are the yes okay.

A

Yeah is the tracking sub resources still in review.

B

B

Why is this not merged.

B

Because jordan hasn't looked at it yet.

A

B

I'm thinking him right.

A

A

Okay, that that's that's the features. um Do we have anything else feature-wise. I think so.

B

uh I'm not am I looking at the right dog? Yes, okay! uh Yes, that's what I'm looking for! Okay! Yes! So um if we're trying to keep track of time, I think we have a few months before could freeze. um It was mid-february, I think, maybe even march waited somewhere in this somewhere in confirmed unconfirmed okay, but.

A

B

Excited um ideally, I'd like these to be merged uh within a month and same for these and same for these, and this is obviously going to be the most uh complicated con, complicated part and controversial, probably yeah. So I wouldn't be surprised if we end up having to maybe help on that, but these three we should be merged by the end of the month. Ideally I'd say by january.

A

I mean well, we still have an open discussion on how the how they accept header should be named, and from previous experience, the naming discussion can take half a year. What.

B

Accept hello, oh yeah, yeah! You mean for this one.

A

B

Yeah, okay, sure um but uh yeah. That's just.

A

I mean it's like.

B

It's an accept header and uh we already know it's probably going to be yamal plus something.

A

B

And so I'm not super super concerned. No.

A

Anyways was more uh punching on the, uh but.

B

I mean I kind of agree with you, but I I still believe it can be done by the end of the months. If.

A

I mean the code is mostly there, probably some testing and then.

B

Yeah yeah the code looks good uh yeah, I'm not super wide. We need to figure out the full back mechanism. I don't know if the api server is going to be able to properly fold back as uh defined by the spec.

A

Yeah, I I I don't know what the what the current mechanism for that is. If, if those apis have rt, is to the standard or.

B

A

To I'm trying that.

B

I'll ask someone and see if they, if they know about that.

C

B

Just to be proactive and see if we can uh falsely palam stuff worst case scenario, if we don't have a fallback mechanism built in the api server, we just manually fold back to yemo yep.

A

B

Let's go to box.

B

I'll, let you talk uh kevin.

A

um Yeah, um I don't know we we can go. I I don't see. The titles are those the two bugs we talked about jet earlier, like I can't see the I'll open the agenda myself.

A

What uh they're, looking into new bucks, those two links? What are they.

B

Yeah, so the first one is: um is the bug about they're both about uh spurious updates when, uh when there is no change yeah, that's the one we.

A

Just talked about.

B

Yes, uh so sometimes people send an object that doesn't change anything, but they end up with an update nonetheless, uh so.

B

I'm looking at it because I I think I I I know about most of these, and most of these are probably already fixed. To be honest,.

A

Yeah, the the long very exampley one looks to be resolved when once a person tests it with the current build, probably.

B

Yeah and so we'll have to be very careful to uh cherry pick, whatever has to be cherry picked because I know we have a lot of cherry picks that are on down and yeah. We'll have to do these they're, not serious. Suicide applied ga requirements, but uh I think it'd be better if we fix them.

A

But the therapy deadlines are different ones. I think, and can also be off the code like yeah.

D

Different releases.

A

We can yeah yeah all right, yep.

A

And the second one was reese's version bumps: okay, um good the status wiping.

A

I don't know when we when we decide when we say we have to decide on something.

A

I I still don't know what to do with it.

B

Yes, I uh I'm gonna, move it down.

B

Yes, yeah yeah, we are still talking about it. I'm gonna try and grab some uh some attention from jordan, so we can have better guidance on that yeah.

A

I mean daniel had a lot of opinions about it in the start, I don't know if he still has.

A

I feel like the compromise we came up with that we talked about like prepare, would be a not so big step but doable and with some ex some some problems still remaining. But I don't know.

B

Yeah, I don't see a good reason to change right now.

A

um Once we want the dangling fields.

B

I'm writing I'm trying to make sure I understand the plan properly. I've written tests, extensive tests specifically to see what the behavior is and I have discussed with jenny about a possible fix which I before I started working on the fix. I wanted to make sure I understood this properly.

B

The bug is uh everything works, fine beside one specific test, so that's uh that's good news, uh but at least with all the tests that I've written. I know that um the change that I make if it breaks anything I know uh so that's good and I'm I just have to implement the fix now.

A

Okay sounds good.

A

um Quality feels type checking, I think you said stefan was on it.

B

Or no, I don't think so.

A

C

B

An opinion, but I don't think anyone's not in I I suspect, so, okay, and so interestingly, I saw someone else mentioned that we need quantity for our crds and we don't know how to do quantities per series like no one knows how to do quantity for crgs, because there is some form of like normalization with the quantity and.

B

And we can do that for crds today. So I'm we're curious if we should have a specific type with some specific normalization on the crds too. To do that and um if we did, then we could use the same mechanism and so as I'd apply to know when it is a sell.

A

Side um should we try to find somebody who who has experience with the area and cannon wants to do it? Is that something.

B

I mean we could, um I can try and ask someone.

B

I think that sounds some sort of difficult and not so difficult right, see.

A

What I'm saying like it sounds like special knowledge and if you know about it, it's easy to fix. Maybe.

B

Yeah for someone who knows nothing, it's gonna be somewhat difficult for uh someone that knows a little. It's gonna probably be easy, um so I don't know I'll I'll see I'll see if I can find someone yeah, it's also not probably the most important, but um to be discussed.

B

Yeah, so, okay, yeah, that's what I wanted to say if we can't find a good way- and we absolutely need that for ga- I think in smd we can probably do a quick walk around and see if this is named quantity or whatnot, or if it has the name of the thing that is quantity, then we can specifically say uh that this can be a string on it which would solve the main part. So we have a. We have a last minute walk around if we can find a propel solution.

A

uh In it can we pass to smd that something can have multiple types yeah.

B

A

B

I mean that's already an intel string. This is not an intel string because it's a float, intel string but smg doesn't make a difference between ins and floats. So if we say it's intel string in smg, everything's gonna work fine. So we could literally say if you see quantity, just assume it's interesting.

A

B

A

B

A

Stefan's problem was that the open api schema format is not precise enough. If we say it's into a string.

B

Yes, but for smg, it's good enough.

A

B

So we could say if it's quantity just say it's intel, string and server side of play is going to be solved.

A

A

A

um The next one directly converts city structural to smg, schema um jenny's. Not here you know, oh.

C

Yeah so I recently started looking into this, um jenny does have a potential fix. I think.

A

C

Still looking into whether that's something I want to adopt, but right now, um this is really hard to test, so I'm trying to see if we can just like improve our testing for this um before tackling it.

C

B

Do you have a do? You have a timeline for that jeff.

C

um Well, at least I want to get the testing done pretty quickly, but um I'm not really sure right now,.

B

Yeah, no, that's fine! I uh do you expect the test to be done by uh the end of the week, maybe or more time, yeah.

C

B

Yeah, that should be a good time like the tests, I'm tempted to believe that we can be done with most of these things by uh the end of the month, but I I know this is not reasonable, so I'm trying to understand where we're going to fail not being done by the end of the month.

B

A

uh John, I I don't know like I'm kind of looking forward to the end of the month, because I remember I remember the last end of the year where I had a lot of time and we got a lot of bugs done.

B

A

We we had this one crazy thing and it was quite fun figuring it out and spending a lot of time on it. I don't remember what it was so yeah.

B

A

I'm just seeing the the the admission thing I'm working on right now is not on the list yet so I'll have to.

B

Yes, and that's probably something we want to solve- is that a burglar future, let's mark, let's, let's mark it as a future.

A

A

What would the feature because, right now, I would see it as a bug that I'm mutating at mission control. I can break managed fields: okay, okay, okay, I uh I get it here's. My original feature would be the admission controls can track managed fields and we're still unsure if that's required for ga or not, I'm gonna remove features.

B

And uh drugs because I mean yeah, I thought it would make sense, but I don't think it does it's. I mean we don't have actual features, maybe maybe api clients, but I mean that's.

A

B

Add it wherever you want now.

A

Above status, wiping uh oh yeah, I think I'll I'll put it here uh and it is.

B

It'd be really really nice if the last months before could freeze, we were just focusing on conformance tests and we didn't have to get any approval from anyone. Saying oh, have you thought about this and that and um have you fixed these and uh I'm not going to put this code blah blah blah at the last minute, like we did for uh my future last cycle, so I'd really like if we could have just a conformance test for the leicester and maybe documentation for the last months.

A

Yeah we should.

B

Set ourselves a deadline of end of january and I think it's reasonable based on these, but on the other hand we have not moved forward as much as expected. Maybe on these.

A

Yeah, but it feels like we're picking a pace, but we have more people helping now than yeah a few weeks ago, yeah. Okay, I have.

D

The feeling I'm.

A

Able to do more despite the new job, because I'm I'm working in a fixed schedule. Yeah my life is more in order. I don't know.

B

B

Cool, um that's it! Oh uh andrea! Do you want to talk about these.

B

About the api service.

D

Yeah, I I picked it up, I haven't started looking did yet so I'm gonna my plan is to learn a bit a little bit more about uh like learn, get some context and then think jenny see. Where can we start and um yeah? Maybe sentence like the updates.

B

Yeah, I hope she's not feeling super well, so I don't know if it's gonna be easy for you to reach out to her, but um uh just so you know.

D

Okay, yeah yeah, um I'm gonna send this luck. um If I have doubts so that uh in case.

B

You remember, as you can see, feel free to uh maybe find the cherry pick that uh I'll figure out the testing framework maybe or work on uh start conformance testing. I mean if, if you set some good uh example for how to write a conformance test, that's going to be super useful for everyone. When we have to focus on that.

D

Oh yeah, I I have sometimes in the next week, so definitely pick something up awesome. Thank you.

A

All right, um I added my task.

B

Yeah good, do you want to talk about it.

A

um Yeah I've been I've. I've spent the evening uh looking on how I could test the replicated problem in a test. First, I saw an integration test with that that adds web hooks and executes some some stuff on it or or quality for every resource, and I'm I'm looking on how they do it or if I, if I can add our test to that or if I just replicate it for a fake, broken web hook, yeah, and then I have a test that I can try to fix on.

A

I I don't I I don't expect it to be very hard to fix, checking I I don't know if we have validation or proper validation code for the managed fields. Yet to be honest,.

B

I mean I'm sure we know we can pass them. So we know.

A

B

I mean, I don't think we're going to val okay, a good question. What is validation, I don't think we're gonna validate if it actually makes sense, as opposed to a validate that it's uh is properly written.

A

I mean we should check for the fields type and uh what was it? The two fields that were removed, that that made everything go boom, um yeah.

B

Yeah yeah yeah, but we should basically pass the uh past the structure and make sure that it's valid as in it's not garbage, but we're not going to look if you're trying to own a field that doesn't exist or something like this are we.

A

No, no, I I wouldn't expect that what we can do is um if, like what the, what the special does specific issue had, if there's many fields managed field entry without any managed fields, it's probably broken, because that shouldn't exist. I think yeah yeah, so yeah. I would not check every like the fields themselves. I don't know if we can validate them even.

B

Yeah yeah um yeah. I don't know how we'll do that, let's keep it simple.

A

B

A

Are we good yeah documentation is for the future, probably.

B

I mean if someone's bald and look at it, it's not difficult.

A

B

It's like, if you have five minutes, it's basically about adding one sentence in a couple of places. So it's not it's not the end of the world. It's it's probably very simple. It's like adding sentences in a few places, um so, ideally yeah I'll, try and add some to you.

A

uh One one thought on the user: documentation like.

A

Okay, I can't look at your screen while typing I have to look at mine. It's confusing.

A

Something is some idea we can, we may can explain the new apply to users easier, giving them best practices.

A

Because, right now, documentation is pretty technical and and.

B

Okay, good, I think we're done for today. What do you think yeah.

A

I think so too.

B

Cool. Thank you jeff. Thank you, andrea.

A

Thank you, kevin yeah. Thank.

B

A

A