Kubernetes SIG CLI, 12 Jul 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes SIG CLI 20230712

Description

Kubernetes SIG CLI Bi-Weekly Meeting on July 12th, 2023.

Agenda and Notes: https://docs.google.com/document/d/1r0YElcXt6G5mOWxwZiXgGu_X6he3F--wKwg-9UBc29I/edit#bookmark=kix.1u9inena3d6c

A

So this is the July 12th uh kubernetes 6 CLI bi-weekly meeting I'm Sean Sullivan I'm your host. uh Why don't we go directly into the announcements and the first is that the release code freeze is coming up very soon. It's going to be next Tuesday um 6 p.m. Pacific time, I think it's about midnight or 1 a.m. uh European Time, Western, European time. So we've got a couple days to finish um and then it'll be about a month or four weeks afterwards that the release will be cut in on good schedule for August 15th.

A

So does anybody else have any other announcements? That's the only announcement we have on the agenda.

A

Okay, if, um if you need any of the reviewers or approvers for this code, freeze please reach out to us, probably through slack would be best um again. That's happening this coming up Tuesday in about a week. So this is the part of the meeting where we get to meet new members.

A

um So if you're new to 6cli or you haven't been here for a while, this is an opportunity to uh to network with your colleagues. So uh if you'd like to introduce yourself, please go ahead, and if you don't want to it's completely voluntary, don't uh and sure no problem is there. Anyone who'd like to introduce themselves.

A

Okay, why don't we move on to the next section, which is the cap updates and the sub project updates?

A

um I I put a link in for the websocket transition cap that actually is scheduled to have to have Alpha submitted by this coming up code, freeze I've been in contact. The person that I've been in contact, mostly with, is Antonio o'jaya of uh who works with Sig API machinery and Sig networking, and uh it there it's Touch and Go. Whether or not this will make alpha this particular release, um but I will show this particular tab.

A

This is the the cap uh transitioning from Speedy to websockets and for those of you who aren't familiar with it since we've got plenty of time, I think for this meeting, I'll just give a quick overview, which is that the there's some Coupe control commands, which upgrade the connection from HTTP to a bi-directional streaming protocol, Coupe control, exec, Coupe control copy, attach and port forward use speedy, bi-directional, speedy, uh bi-directional uh streaming protocol, but that's been deprecated for eight years and so for quite a long time now, kubernetes has been trying to upgrade that to the more modern websockets, and so this is part of that effort.

A

um And does anybody have any questions before we go on to sub-project updates.

A

Okay, let me go back to this.

A

Okay, so I think that we're um we're we're at the point where we're going to go into the oh, the agenda items unless anybody would like to do a sub-project update.

A

Okay, so the item- that's on the agenda, um let me open that up. uh It sounded like it was.

A

uh This is to deprecate the usage of no opt def value, which means that uh the equal between the key and the value in a flag or an argument or a flag, sorry uh can be missing, which causes quite a bit of confusion in some cases, um I think that this particular issue Falls clearly in the category of things which makes sense, but which are probably really really difficult to change, because people are depending on them, but uh there's a couple people who would like to work on this and they want feedback from the Sig uh Ricardo it doesn't sound.

A

Like was able. Oh, oh, cool, Ricardo's here, Ricardo do you? Do you want to I know that varsh is here as well? Would you guys like to take it over now.

B

uh Yeah sure I can, if Russia doesn't want I can I can just go so yeah I. uh So this is. uh This is an issue.

C

That I figured out.

B

So someone was a learning, kubernetes actually asked.

C

Me like hey, why can't.

B

I use the following some places and I cannot use people in other places or but.

A

Why this happens and uh I I.

B

Have to say this indicator and figured out this kind of problem uh honestly uh it. It does make sense to me uh the approach that was taken, I I, think that's. Why actually the jumping into the issue as.

A

Well and say: hey, idds and I know.

B

The feeling, but my proposal is that we we have some sort of well-defined behavior from now on and like if I should be a Boolean. It is a bullion. If a flag should be a string, it should be a string, and we don't have this- this middle class, where you you may have a bullet which is a screen, um because I think that overall, in the long term, for uh people starting to use that and even I I keep doing these kind of mistakes.

B

Myself sometimes uh we'll have some some sort of a well-defined Behavior Uh I'm, not sure if everyone understood the problem here, but uh no up. That Vol is like if you pass a flag, but you don't pass a family on a stream flag. It's it's gonna, assume that there is a default value, so.

D

B

As an example, if you, if you try to do the best.

E

B

Street it's gonna use the validate, which is not strict, and it's going to create unemployment to the main street right. So that's that's the proposal. I I have no idea to be honest on how to start that. I think that uh we wanted at least to bring this to the sea to know if it does make sense to rain that we had before starting them and when people say yeah, it doesn't make sense and I know that this is gonna, probably.

C

Take some fine.

B

Any care right, but uh at least I think it would be good that everyone was aware.

D

B

This kind of behavior.

C

um So in general, I'm, very supportive of the idea, uh I I was actually meant to comment on your initial proposal that you posted on Twitter even before the issue was open, but I just somehow for the island. Sorry about that.

C

um The issue that I see currently at least is the places where we rely on that kind of behavior, especially the dry run, because it has a double meaning and if I remember correctly on one uh at some point in time was inclined to draw a double meaning because it, uh if you're not aware dry, run, allows you to be both a strength.

C

And there was a PR to drop and only switch to the string valve because you can, between Drive run, you can take between the server client and and on, um but for backwards compatible. We have to support the true false if we can initially support the previous behavior and figure out a path towards with those few places where this is a problem.

C

um That would be probably the initial step for the cap, and on top of that, what we can do already is probably write, a very simple verification: script that will ensure that no new command with this particular change are being added or that nobody will try to modify and set that black anywhere any anywhere else in the keep Cloud code, with just those few exceptions that we uh that we are exclusively aware of, and we will try to figure out a way out of that situation. That would be.

C

That was my proposed approach uh and I'll I'll. Let others think.

A

So there's a couple of hands: uh I think varsho is first. Would you like to go.

D

Yeah, so this is just a question of the kind of behavior year, apart from dry, run and validate. It also occurs in other places like Cube CTL config, especially the Boolean, the supposedly Boolean Flags like insecure, skip, TLS verify and looking at the code. uh It looks like it's considered a string and we literally set the default value to be false, and then we set the no of default value to be true in case the flag is not present, so I just wanted to know if there was a specific reason in doing that.

D

Why not straight away declare that as a Boolean and have a default value instead of having a hidden value where the flag is not set.

C

I'm guessing that some of those instances will be or historical reasons either nobody checked that there is a Boolean or there were some odd instances.

C

um I I, don't think we can have a a rule to set for everything will most likely have to get through uh through a git history of any particular command and ensure that we don't write anyone.

C

But with those so yeah I I I'm, fully aware that that's not the kind of uh response you were looking for, but it will be Case by case uh in most uh in in those few uh instances. But if it if there is a simple replacement that we can do, then just go ahead again.

A

Did that answer your question in versa,.

D

Yes, at that time answered my question next Monday.

A

D

Eddie has his hand.

A

E

Yeah I I, it's the first time I've seen this issue, sorry I'm, looking at where we use it and after reading the P flag docs, it makes sense why we did it that way, and it's I see that yeah. This is messy and tricky. Where did people actually get bit by this.

C

We kind of had an example: um Sean Kenny's quote a little bit up. There was a an explicit example with strict and validated oh yeah. So that's one example.

C

um If you try something like that, you're actually uh not gonna validate with strip, but the stroke will be picked by the name of the deployment and the value they will pick the default value.

C

Yeah, that's that's the working of that slack. It's a little bit messy.

E

The default value is strict, the no option default value is also strict.

E

So am I what am I missing.

B

Try to create the non-default value any like I, don't remember, which one was the other one day.

B

Whatever it's not free, the deployment is going to be created, it validates three and the name is going to be whatever you try to use as a flag.

E

I see yeah. This is definitely a bug. Then right. This is a P fly. Cobra bug.

B

So I I was looking into the D flag uh uh code and the problem is that uh there is no way to know uh when you are passing a flag or uh or an argument without using the equal name in this case, if the people in this case, when we use the note of that ball. So that's that's why this is the only flag. This is the only kind of flood that that you should be using The Eagle right.

B

The confusion is because we allow people using equal, War space to separate floods and arguments when you are using string Flats, it kind of confuses because Android Ram, you should use equal sign but like let's say image, because there is no default value. You shouldn't be using people. You can use space, so that's kind of a bad user experience.

E

Yeah this is this is definitely a foot gun um that sucks okay and did sorry I'm just catching up. Did you have a proposed fix.

B

uh No because, as I could tell people like there is no easy fix right. So that's why I I have rated this issue and in the decision to actually to stop using this note up.

B

So uh you should not. You should know which, which value you want to pass into a stream and play YouTube. That's it right so, but that's like a long-term uh piece that we kind of interest rate to see what other things on that and really start working on care. For that.

E

Yeah we should, we should definitely add a lint rule to to not allow future PR sad. This in.

E

Okay, thank you.

C

E

C

And the rule can be added rather quickly even already in 128, and the rest will probably take a little bit longer uh to figure out the passwords. But having the lending right away would be definitely very, very reasonable, so that we don't uh to the already existing buckets.

E

Yeah we should. We should also raise this on on kadev and make sure other people are aware of this like this, this is a spooky foot gun that I didn't know, existed.

E

I wonder if we can get Mark to to come to a call to discuss it too.

E

Like this violates the the Unix philosophy of of command line, args.

C

Mark was actually commenting on the issue that you considered that a blocking program, yeah.

E

C

E

Don't think Cobra could do much better, though yeah.

C

I'm, just I I didn't entirely on the Empire State, but yeah.

E

Yeah there should be no semantic difference between an equal and not an equal. This is yeah.

E

Have we seen this before? Is there any history around this.

E

E

There is there's one issue from 2018.

E

It's cat: it's talking about Cascade,.

E

And it looks like it got fed about closed, which shouldn't happen anymore on our repos.

E

There is a p flag issue: that's old.

E

Looks like it was closed and not merged.

E

It was closed with the intent to come up with a better strategy.

E

2017. all right.

E

Or this is something different: what is this?

E

B

It's different meaning yeah, because I have just jumped into it before it's. There is no 2021.

E

I know that McBride is maintaining Cobra I, don't know if he actually still is.

C

I think, based on his recent comments when he left AWS, uh the new place still allows him to work in Cobra to a certain degree, but I think he's currently on a uh on a holidays, so probably yeah it. It just happened that were not enough in the middle of many people holidays.

C

Yeah definitely worth reaching out to him I I that will be able to quickly resolve this and.

C

Checking all the possible values people are affecting is probably a worthy goal.

E

Yeah I guess it's it's a it's a breaking change, no matter what right for people expecting the behavior either way.

C

Yeah, unfortunately,.

E

That we should run by Jordan Jordan always has really good ideas in these situations.

C

I mean I would probably start with uh write down all the possibilities that we have and once we have the possibilities, all the pros and cons write it down as a cap, and then we can pull into order. Okay,.

E

C

Matter what uh having the initial um homework done before pulling in Jordan will be required.

C

uh I, don't want to throw out him a problem, then without possible approaches, even if he comes up with entirely different idea than ours uh at least.

C

E

C

E

Ricardo and varsha, how do you feel about writing all this down in the cap.

B

I was just hiking Russian and it's like from what we've taken care of this and I can just keep helping so perfect.

D

Yeah I can write down a gap for this uh and for the lender um 1.28 is the deadline is next week right.

E

Yeah we should we should try to get in before then, but we can definitely get a winter roll in Past release.

E

I think you muted Sean.

A

Sounds like we have a plan and some action items uh do we feel like we've? uh Is there anything else to cover on this issue.

A

If not we'll move to Eddie with the kubecon.

E

Yeah announcement.

A

E

Center card, thank you all for taking.

C

E

Drive in this one appreciate it, this is uh definitely going to be a tricky hairy situation that there's no clear answer for well. I appreciate y'all's work for kubecon 2023. We have our maintainer track submission due on Sunday, so we started a draft of a simple maintainer talk, um we're going buzzwordy and asking the title is something like what's new with Cube control and customize, uh which we had 50 people at our maintainer session last time. So hopefully we can get more, that's rad, uh I guess.

E

My question for the group here is: are there any any topics? Any caps, any work that you think we should highlight in this talk or in the submission? And so that's for the maintainer track talk and then we also have the option to do a one minute.

E

Video update from our Sig again for the kubecon Keynotes, so Marley and I did this, for what was the last one we did was that EU I think it was EU, yeah, EU yeah and it was rad like it was so cool to have like our our little video slot in front of the entire room of just like talking about the work and telling people to come, hang out and help so I think we should definitely double down and make it even better for next year, so start brainstorming things that we can highlight work that was done by y'all or other people, and we can get those in a cfp and a uh the the video highlight.

E

So does anyone have thoughts off the top of that? Also? Who is planning to attend kubecon, Chicago I? Think it's in November, that's I can't see all the hands put put a put a yes in chat or something.

E

E

Yeah I know we need the budgeting world for travel is, is still squeezed. huh Okay,.

E

Oh, are they damn? Okay, um so any any things that come to mind that you want to bring up or discuss around topics or Caps or work that was done.

E

uh Let's see what, if we, what what caps did we get into.

E

C

uh uh For 128, we pushed the events to our IGA and interactive delete uh that was released as Alpha I can't remember what we did for 127.. That was like ages ago. Yeah.

E

And you said events is events move to Beta. Is that what it was.

C

A lot will depend on how far Sean will go with Speedy, but that's to be decided and either way by Chicago or which will be in November will probably be further ahead, uh either way, because that's like in four months, so you know- that's like whatever we will say today will be history in November.

E

126 came out when.

C

uh End of last year.

E

C

That was already.

E

Covered for Europe right, yeah.

C

Okay, 127 was post Europe because that wall is end of May, okay or end of uh hold on. Are we released 127 before you.

C

Oh yeah Arda is just pointing out good examples. The uh apply sets that Katrina was working for apply prune resolution.

C

I'm, pretty sure that we can cover both of these topics easily nice ah there yeah, if you have any suggestions or any additional proposals, that we should cover during the presentation just pink either of the leads on the 6ci channel. That's probably the best option, uh no matter what we will write in the uh in the cfp submission the actual uh presentation will be prepared towards the the week before or the week when uh we will be doing the presentation. Yeah.

E

Yeah and the deadline for that video is August right.

C

Yeah, but that one is it: the deadline is a little bit shorter.

E

Let me let me get that exact date hold on.

E

Deadline for the video is.

E

August 10th Thursday, August 10th.

E

Here is the announcement for that.

E

Okay, we still have some time.

E

Cool any other thoughts around kubecon us Chicago.

E

A

Okay, thanks for all that feedback, it'll be nice to see folks in person for those of us uh who have were able to get the companies to pay for the get the travel budget. It seems like that's a big question for a lot of people, myself included.

A

um So that looks like we've covered all of the items on the agenda. uh Is there anything else last minute that we want to bring up and talk about before I? We close this.

A

Okay, we uh we canceled the the two previous meetings and uh seems like we've covered the topics that we wanted to cover this time so we're going to and that it's good to see everybody and we'll see you again in two weeks.

A

Good luck on the code. Freeze, foreign.