Kubernetes SIG CLI, 5 Oct 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes SIG CLI 20221005

Description

Kubernetes SIG CLI bi-weekly meeting on October 5th, 2022.

Agenda and Notes: https://docs.google.com/document/d/1r0YElcXt6G5mOWxwZiXgGu_X6he3F--wKwg-9UBc29I/edit#bookmark=kix.f00y7ug221g5

A

And so I have started, recording, okay, so uh welcome to the October 5th version of the kubernetes 6cli bi-weekly meeting. My name is Sean Sullivan I'll, be your host today uh looks like the agenda, isn't uh very big, but first we're going to go into announcements and the only announcements, or at least from what I could tell. Yes, the only announcements uh have to do with the 126 release tomorrow by the end of the day, is when keps, our enhancement freeze happens.

A

So you have about one more day. The six CLI has at least one cap. That looks like it's going to be implementable for 126., uh thanks a lot to Katrina in a very short putting in a lot of effort in that review, for the coupe cuddle, explain using open, API V3 um and then once this enhancement freeze happens tomorrow or the next major uh signpost is the code freeze which happens in November on November, 9th Wednesday, so so also.

A

In addition, uh mache would like us all to know that if you you've got some emails about calendars, that everything is okay. Is that a.

B

A

Summary of that matcha yeah.

B

You can uh safely ignore. uh We had some issues with the previous cons or with regards to uh uh ownership, and we just had to uh switch entirely to a new calendar, so nothing has changed with regards to meetings times, um except for uh with Katrina Road below the krm functions. A project meeting is being canceled in favor of the regular bi-weekly 60i calls. So if you have any uh okay, around related topics feel free to drop them and in our bi-weekly calls uh by the way. Sean.

B

Can you make me a co-host I'll check the agenda.

A

I I had a hard time understanding that last one. What was that much can.

B

You make me a co-host I'll be able to share it. Yes,.

B

A

A

Thank you very much um for that matcha. So uh now we're going to do introductions before we get into topics, and uh this is the time for uh your six Eli colleagues to uh to meet you uh if you'd like so, if you'd like to introduce yourself if you haven't been here before or if it's been a while. uh This is the time we take for that.

A

Would anyone like to introduce themselves.

C

uh I'm at Overton I'm, just checking in uh we uh I work for SAS, we use Kate's for deployments uh and so I'm, mainly looking to get a feel for. What's going on and so I figured attending a few meetings would help get me a little more oriented.

C

It's a good idea. Welcome.

D

And my name is David Goffman I'm, a senior software engineer and currently I'm providing the seminars for a CTU university and we are struggling with some missing uh still commands that we know today. I would like to present my free full request. I hope I ended the meeting who will have some.

D

We'll go forward with mine pull requests. Thank you.

C

B

A

Is there anyone else.

A

Okay, why don't we move on to the topics uh that uh they so we've we've added uh a couple of David's PRS here, uh We've also added. We may want to uh look at the create network policy link which has a general right uh approach to how we want to deal with create commands, and so so David.

A

It was important to to come to the meeting uh in order to understand our general approach to how we intend to uh develop the coup, cuddle, create command and, um and so uh just to be honest, just to be transparent.

A

It may not be possible to uh to merge your PRS, um but it was important I thought to at least have a discussion on where, if we're not going to keep adding particular resources to the coupe control create than what is our general approach to cube, cuddle create and we've written down a few things, and it's important to share that with the the community um so uh We've.

A

So back in what was this in 2020 with uh Phil when he was uh working uh when he was the Sig lead, he had some ideas on where the coup cuddle create command, should go and, and so I I think that that sounds like the correct path which may not align with the PRS David um and so uh I'd like to get other people's feedback on on that. If we still think that uh we want to uh do the the what's it, the generator approach.

B

That the overall generator approach is definitely one that we want to pursue, but most likely will require a little bit of our time uh to get done uh as a command as a double Gap.

B

um If I remember correctly, I covered this outfit last time or two weeks ago, where I mentioned that it would be interesting, because we've had that issue. A couple of times mentions to allow certain commands um to be extended with sub commands. Currently, if you try to create a plugin which which will be shadowing one of the built-in commands, that will not be possible, so, for example, you cannot create your color, create uh whatever resource.

B

We initially did that, because we uh we said that we don't wanna, allow that what we, uh but we said that we might reconsider that option in the future.

B

um This is the topic of the create sub commands is returning over and over again by various uh party uh I raised the topic of opening the create sub command for plugins.

B

So, instead of creating a built into kind of built-in command, people would be allowed to create their own suck meds uh uh their own login, which would hook into the create sub command. So, in your particular case, say this that will be create network policy.

B

The reason that we want to, we don't want to add them in to the default and to keep cuddle is because the um um that uh the cost of maintaining those create sub commands is growing over time and uh and the value that you can get out of the create is usually limited because it's helpful to have them simplistic and just for uh for starters.

B

But if you want to grow them eventually into sophisticated commands, that's not something that we want to invest our our time into maintaining, to allowing the mechanisms for plugins for create supplements. Would be the ideal middle ground and in the long run we would like to see something like generate where a command would be looking at the open, API definition of a of a resource and based on that it would be capable of creating either some kind of a stop or Ortho flag will allow you to modify uh the resource but um yeah.

B

We just need someone interested in pursuing this topic. Specifically I can't remember if we have already uh I kept written down, I don't recall, but there's uh definitely that that issue that Sean pointed out um should have some pointers.

E

Yeah the issue I think um also mentions that we would want it to be exposed through a separate gen command. Ideally because, as a teaching tool, we want to teach people the the like preferred flow of the declarative workflow so that you do gen and then apply instead of um an imperative uh teaching. Workflow and um another thing that that occurred to me. That's sort of related to this when I was um reviewing. The explain, opened.

E

Ipiv3 cup is that if we get that uh additional information into the open, API V3, then other commands could leverage that, to like explain, is also a teaching tool and it'd be really cool. If we showed that uh gen example in the explain, output right, um so I think there's a lot of value to this approach that we've discussed on on this issue.

D

So uh can you can I ask you so it will be new command, it can be gen or it may be book called design and I agree. Maybe it's actually a good idea to just go to a new command. The reason why, because create and again I came to this, took um three missing commands because they're actually required for getting this ticket and cka certificates.

D

Those actually commands actually uh part of the latest, uh a little foundational image and actually required to create using those three commands and definitely go to um documentation, copy paste yml file into your existing one. It may be a mess because actually you have to work on the indent and sometimes actually uh VI shifting all lines by in the bypatings so that that actually my reason why I create those three commands, because actually it's required for for passing the certification.

D

But if you will have a different command like design or Jam I would highly recommend uh remove two additional attributes. It's dry run, client and dash o, so people just can create a command and see manifest on output, and then the object can save this output manifest into file and then execute apply.

D

I agree with this approach as well, but we need some manifest generator for those three uh resources.

A

So so, just as a quick aside, uh one of the most common uses for coup cuddle create for the resources that are supported is a minus, minus dry run and then a minus o yaml, which will actually generate the the yaml for subsequent apply. It actually has a couple extra Fields, like the creation timestamp that you may have to to clear out before you apply it, but I think a very common uh workflow for people who are trying to generate yamls is to use it. Do it with cuddle, create and the dry run.

E

Right so that's like a sort of a workaround to what the Gen approach would do as the first class feature yeah.

C

E

Yeah this would be driven through open API, ideally instead of hard coding in in the specific command that would work for any resources. Even third party.

A

And- and these version SKU issue would would also be ameliorated, since the open API is coming from the the server and instead of you know, compiling in a particular version of a resource.

D

But if you create a new command, let's say design instead of create design, we don't need those two attributes right, because design is not for create design just design manifest file. In this case, you can include additional attributes.

D

What is our missing can create.

A

Yeah I think I think that was what uh Katrina was sort of saying was that, instead of the kind of the hack of the minus, oh yeah, well with the dry run, this will actually generate it. uh Natively is that is that correct, petunia.

E

um It would generate natively coming off of the open API and then it would. It would be a new command. Yes, so um we could have a different option set there, and that would be probably things like well by default, it's just gonna, it's not gonna, create anything. It would be client-side only it would.

E

It would output it and then, if you do want to then create you pipe it to create um using yeah yeah to create or apply um the the other thing that we could do with Jen uh would be to um have an option that makes them that throws the user into an Editor to so that they're, just using whatever being code enough, maybe as a starting point and then they they mess with it before um whatever they were going to do next and committed or create it or whatever that that may be so yeah, it would be a new command and we would we would have to think about what options we wanted to expose and whether or not we wanted to deprecate some options on crate once those are available, because they're really not and just basically pointing people to the new flow um in.

E

In short, there's a lot to think about here. I think we all agree about this plan. um That's sort of sketched out on this issue, um but it is a significant amount of amount of effort, and you know it requires uh the full kept process uh to to move forward, which is called an issue itself. I believe so we really need someone who's, passionate about this topic, to Champion it and and bring forward a cap for for this. For the sake to consider- and then actually you know put in the work to move this forward.

A

So so, I'd like to thank David for kind of instigating this very interesting discussion, I I, think that um uh we, we probably were probably not going to go forward with the particular PRS that you put time into and and uh we're sorry for that.

D

um No problem, it's not sorry I'm, just wondering when we'll get this commanded do we have some timeline.

A

That's yeah, that's a great question. I.

E

Feel like I was just saying: like things don't just just happen, we need folks to to Champion them who are passionate about about um the need for them and about making them happen right. So um someone needs to step up. Perhaps yourself, if you or it sounds like you're, pretty passionate um we'd love to have you uh Champion that that kept processing and drive it forward for the sake um there's nobody in particular, you can see the issues from 2020 there's, no one in particular. Who has claims to work on that so far, I.

D

Can take this ownership? uh Just uh I need someone who I will able to communicate during um that process, because I need someone who will be getting the approval, because I cannot just uh create a pull requests and keep uh tracking on the on my local machine. I would let some guy, who will be actually helping me I, can carry the command and I can create a new pull. Requests belongs to this command. Currently, for example, I work in new command. It's called format, so my new command.

D

What I'm working right now format actually will be formatting yml. So usually, when you know, when you pay and by the way just give me some ideas should I continue or not so ideas here you can place yml file, but maybe your identical indents are not correct. Maybe it's actually some uh parameters like let's say Mount volume or volume actually shifting, maybe two spaces left or maybe extra space to the right.

D

So my idea of the format will analyze your manifest your manifest file and also will be validate the schema and I believe right now you are using the Swagger when I look on the not documentation. Look inside the code of the chip, steel click apply, you're in the Swagger, so usually getting Swagger uh schema and based on the Swagger I can actually correct my yml output.

D

So, instead of just give the exception what we everybody getting every time you uh yml cannot be accepted, it will be helped to correct the environment because again, why I like the way ml, it's better than Json, but yml has a problem in dance and sometimes it's not easy to validate or can visually see something wrong in the maybe extra space. So this format function. What currently I'm working will be at least try to format your yml. If it's some problem is intense.

E

I'm hearing two questions there, um both related to the process for getting uh fairly major new features accepted into qpedal.

E

So um if it's a net new command that you've come up with yourself as as in this, this last case with the format sounds very cool. um What we usually recommend folks to do is start with a plugin. That is a proof of concept and they can circulate, and folks can try it out and see how it works. And then, if it proves successful and popular, then we can start the enhancement proposal process to potentially add it to cube cuddle proper.

E

But we usually want to see that that first step of like sort of proof of concept that folks can try out, as as a plug-in for commands like that, um the case with uh Cube Kettle generate is a little different, because this is something that the Sig has already agreed to in that old discussion on the issue and today and probably some other meetings in the past, where we already have that initial consensus, so um I think we're ready to move forward with the enhancement proposal process at this point and that you can do proof of Concepts um in in support uh of of your enhancement proposal.

E

But really what needs to happen is writing up in the enhancements repo there's. This template that you fill out with your proposal and the plan and what the um details of the command would look like, and that is where the Sig will all come together to collaborate on the design and and support you in making the plan uh something that we're able to move forward with. And it's only after.

E

um We have that merged as implementable, that the work on the feature itself begins and it gets added to the Milestone. um For uh inclusion in a future release to actually come back to the announce some of the announcements we made today about what's going into the next release, that's has to do with this whole process of how the kubernetes project makes enhancements and and plans uh work together.

E

um So yeah, the really the next step for, if you wanted to Champion, generate, isn't to make more PR's it's to uh well I guess it is a PR2 which I think the the pr uh with the enhancements proposal um for for that feature. Does that make sense, makes.

D

Sense, maybe I will keep contact with you if you'll send your personal information.

D

um I would glad to have personal contact with you and you'll. Maybe help me to navigate with all this process because again, I'm new.

E

um I think we uh prefer to work in the open as a stick. So if you join the six CLI Channel, then you can get support from any Sig member who has more experience with the process instead of relying on one particular lead to help you, okay,.

D

E

D

You but I can start working. This generate because generate already a proof right. I can start working on the generate and make the request, or my put my pull request, move to under generate uh Repository.

E

um No, we we need to go through the enhancements process to see what exactly that would look like and design and get the design approved in the future tracked um through the enhancement proposal process. For that one as well, I should so.

A

So just a quick note: if you look in the chat Eddie's pointing out that there there was a proof of concept uh under and Eddie has the link uh where he's he's worked on the coupe cuddle generate as a proof of concept.

D

A

That'd be a good place to to start appreciate uh all the uh the enthusiasm thanks David you're welcome.

B

uh If you're curious, what the enhancements for a new command should look like, uh we've been working on events command, so the enhancement describing a new event command and what the uh the enhancement should look like. You'll find it under kubernetes enhancements. There's six CLI cap six CLI uh 1448 is a description. What the document should look like what it should include, what are their requirements and everything we're currently working on pushing the events over to data in 126, so that should give you a rough view.

B

What do we expect from such document describing the command? Oh.

D

Thank you. Thank you. Can you share this link also to the chat please.

C

I will do it.

D

Yeah and that's actually enhancement, it will be uh just new uh readme file.

B

Another directly, you need to create an issue under a kubernetes enhancement.

B

um Every cap uh every document has its own issue and the issue are roughly describing the uh the idea behind the this proposal and then you'll you'll be also linking all the issues and all and the progress of the work, uh and then out of that you will be creating a new pull request, specifically adding um uh the documents required by the cap. There are three documents: uh two Youngs one is describing the cap.

B

One is uh following the pride Readiness and the other one is the readme MD, which is the entire description of the uh on the.

D

C

D

In the git I should I create a new Branch or how it's actually again. Can you give me just start Point? How can start you.

B

Create a pull request with all those uh directories have a look at uh Cube cuddle events and uh that's 40 and 40 issue in the kubernetes enhancements it holds. It has all the links to all the implementations, uh both and uh in the cap and the actual code that is adding those commands. Okay yeah.

D

I will try to review.

B

Is your starting point and then just follow all the links, and you should be able to figure out how this works excellent.

C

C

A

So great, let's uh so I think we've uh We've covered that pretty pretty well. If it's okay with the uh the Sig we can move on to the next topic for with Noah didn't, is Noah here.

A

Okay, so Noah's uh mentioning uh a particular PR that needs review.

C

B

This is just asking for reviews on this particular pair. uh There has been uh several back and forth uh from what I see so.

B

We just need to have a look at it and add and make sure that whatever has been written in this document makes sense. Doesn't look like super complicated to someone has to spend some time and double check if it makes sense.

C

A

Can we move on to the standups.

A

Okay, Brian's been doing some real good work with the translate really appreciate. All of that effort of Brian sure.

F

Yeah so uh since last meeting I did create a blog post that we talked about uh after I created. So it was a great exercise, but after I created, it um I think I kind of realized that it didn't really fit the uh the content. That's in the uh k8s.io blog, so uh I think it was uh Tim Bannister or maybe somebody else that mentioned there's another blog uh kds.dev. That might be a better place for it.

F

But uh in the meantime, Katrina pointed me to a community survey request that can be used to send a poll out to the entire community and they also offer some uh help and kind of crafting what that poll looks like in analyzing the data as well as publicizing it. So that might be a good approach.

F

So I did create a draft poll and uh that's uh kind of I guess what we might submit to them to publicize for us uh and then at that point, I think we kind of talked about maybe doing a blog post um at the same time as that poll is released to kind of add context or further publicize, it I guess and then I think part of it is like figuring out like what are we going to do with the data once we get it um are we, you know I, think what's on the table, is you know hey?

F

Should we continue doing translations or should we take them out all together or maybe there's some Middle Ground, where we put them on hold and and I did reach out to some people in the Sig docs localization on slack and one of the things they uh there were several uh good comments and things, but one of the things was like hey.

F

Maybe there's some threshold that you know you have to meet this threshold in order and able to in order to turn on uh translations for a particular language, and so maybe there's some Middle Ground there, where we turn them off and we say: look unless you get to some percent um complete, then then we're not going to turn it on for this language uh and then I think there's probably some coordination efforts and everything else on a release by release basis, to kind of look at what the percentages for each language and then reach out to whoever it is that that can be a resource for translating things.

F

So so yeah, that's that's where it stands right now, um I guess next steps are uh figuring out if the pole is ready and if so, when do we want to release that? And what do we hope to do with the results.

A

When you quick question Brian, when you talk to the localization channel, uh were there did it sound like there were experts that we would be able to rely on to get some more translations for particular languages or there's certain? Are they missing? Do they have holes for particular languages? Do they have really good coverage for other languages? Did? Did you get any a sense of for any of that.

F

um I mean I think there are resources there. It's there's some crossover, obviously in the people that that translate the docs and I think they are largely unaware that uh that maybe not even aware that that Coupe cuddle has translations or if they are don't know how they, where they are, um where they are set. So I think there's people there that we can tap as resources.

F

We probably need to make it easier and more more clear how they can contribute I, don't know that a lot of people even know how to contribute the translations.

C

F

So I guess I guess I mean I, guess the next step I mean I'm thinking. The next step is to come to con consensus on what the poll should contain and uh try to get that out there and then I mean ultimately I guess you know I feel a little bit um conflicted because I'm, obviously not a a user of the translations. So I don't want to be the one to come in and say, like hey, I'm, really advocating for let's take out translations so I'm, that's kind of where I'm hesitant on on doing that.

F

So I don't know what the right way is to involve more more stakeholders into this decision. I guess that makes sense.

A

Was there any? Was there any one uh place where we were getting feedback on the incompleteness of translations before, uh or was this instigated by ourselves?.

F

No, it came from. There was a pull request. A few weeks back that somebody was literally taking out all the I-18n stuff and I was like well. Okay, that's probably a little bit drastic, but the the sentiment behind it was that hey! You know these translations are really incomplete, so they're, not even useful to me so I think we should just take them out and I was like okay, I can kind of see that um and I I don't know. Maybe maybe it's an overblown thing.

F

I mean it's not hurting anything that we have that in there or is it I guess that's what the poll is hoping to find out is when, when somebody is using it in Japanese and and they they run, Coupe cuddle help and like 90 of it is in English. Is that is that, worse than nuke and I? Think that's where the poll comes in? Is that worse than no translations to people who do use couple in another language.

E

Yeah the two concerns that I have that to address this, like finding out that basic information like exactly what you just said, but to actually find that out. Based on the poll, we need to make sure that we can get the pole to the right set of people um which is potentially folks who uh you know, aren't English first language. So we need to be able to reach the right communities with their pool and hopefully the the community survey.

E

um Collaborators can help us at that angle and the other one is that if we get an overwhelming response that um that we do want to keep these and make them better, how do we find the right set of people to help out with that, and should we be doing something in the poll to see if there's folks who want it.

B

E

Are interested in helping like how do we approach approach that.

F

Yeah and that's a good point and uh and there's also that I mean there's also the option that you know it's not causing any harm to have the we could. We could also just use issues. I mean we've had one person raise an issue about it, but you know, maybe maybe we just kind of monitor the issues and if more people start raising issues about it, then we deal with it. um I don't want to.

F

This is certainly not like a big thing that I care a lot about and I'm trying to push through or anything um so I I, don't know, I mean there's all I guess what I'm saying is: there's also the option of doing nothing and so I'm I'm open to that option as well. If people think that that is an okay approach,.

E

um Pr to try to delete everything is a good tripwire for us to think like. Oh, the translations have degraded to a state where somebody is bringing this opinion um so I think taking that opportunity to at least do the poll and get the information from communities to the extent we can um makes a lot of sense and I'm still in favor of that personally,.

F

Yeah I can get on board with that. So so the poll is a good place to like at least gather more information and, like you said, maybe, and do it in a way because we're gonna get one shot at the poll, so we can use that at the same time to collect uh maybe add. Another question is like that says: if you're willing to help out with translations, uh put your contact, you know, put your GitHub handle or email address or whatever here.

E

Yeah, that sounds good.

C

A

Thanks for that, uh Brian, um please let us uh know uh what well you could reach out to me if, if I can help uh to get the poll out, uh but it sounds like you guys are already uh have already made good progress. There.

A

Okay uh Katrina, would you like to talk about uh customize.

E

Yeah, this is just a quick follow-up on what I mentioned at our last uh bi-weekly. um Can you hear me um yeah? It's just a quick follow-up. um I mentioned that I was going to be putting together, especially to help coordinate the dog's migration efforts, and I did do that. So I have a link to the spreadsheet here and uh in case it's not self-explanatory, there's also more detailed instructions than the issue that is tracking the overall consolidation effort.

E

um So just to recap what this means: customized docs, are scattered across a bunch of different places, and it would be really helpful to customize this community to improve that situation and also helpful to the maintainers so that we are more easily able to keep those docs up to date um and respond to documentation requests. So this is a something that we could really use the community's help on. It has a very low barrier to entry.

E

It's very simple: straightforward granular tasks, not the most glamorous, perhaps but a big help to the community and, if you're interested in getting involved in customized, but aren't familiar with the tool. It's also a great way to you know as you review these docs you'll learn more about the customized tool, you'll test out the examples and then you'll be in in a better position to start making code contributions in the future.

E

So um if you, if you're looking to get involved an approachable way to help the Sig this, this docs issue is uh maybe a good opportunity for you.

E

And yeah, if you have any follow-up questions on that, you can comment on the issue or there's a customized Channel, a customer specific channel in the slack.

A

Cool, thank you very much so uh before we wrap up, is there anything else that somebody would like to bring up while we have the Sig together.

A

Okay looks like we're: gonna get some time back today.

A

um I will stop recording.