Kubernetes SIG CLI, 4 May 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SIG CLI Intro - Phil Wittrock & Mengqi Yu, Google (Any Skill Level)

Description

Want to view more sessions and keep the conversations going? Join us for KubeCon + CloudNativeCon North America in Seattle, December 11 - 13, 2018 (http://bit.ly/KCCNCNA18) or in Shanghai, November 14-15 (http://bit.ly/kccncchina18)

SIG CLI Intro - Phil Wittrock & Mengqi Yu, Google (Any Skill Level)

Join the SIG CLI introduction session to learn our mission, recent accomplishments and discuss future work. We will also focus on how new contributors can get involved in the SIG.

About Phil
Phillip Wittrock is Senior Software Engineer at Google, Kubernetes SIG CLI Lead, and a member of the Kubernetes Steering Committee. Phillip’s hobbies include debating how kubectl is pronounced, talking about Kubernetes at social events, and appreciating code as art.

About Mengqi
Mengqi has been working in the SIG CLI since 2016. And he is one of maintainers of the SIG CLI.
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.

A

Okay, my name is Phil wit, rock and I'm. One of the sync leads for six. You lie. This is ma che, another sink lead and this is Munchie who's, one of the maintainer of qu control in sig CLI. So we're doing the cgi intro today and before we get started. I just wanted to pull the audience here. Who's contributed to coop control.

A

Alright, who's attended, six CLI meeting who's used to control.

B

A

Who's used coop control and decided that the error messages weren't good enough. Okay, about the same amount. That's what I expected! Okay cool, so Munchie is going to give the cgi intro and talk a little bit about what our goals are, what we're trying to accomplish in 60, Li and Q control, and how to become a contributor how to become involved and how we develop things.

C

Okay, let's get started hi everyone I'm monkey from Google I have been working in the 6a RI for over one and a half years. So I'm here to represent in the 60 I, took give the introduction.

C

So the topic I'm going to cover today missions our responsibility, rows in the six and our existing sub-project and future sub project and our ongoing efforts, and also how you can get involved and how to contribute.

C

Ok, let's start with our mission. First, our mission is to build Q control, CLI and the its front. For example, customized I will talk, talk a bit more about this in the future slides.

C

So we want to make Q controls about API extensions and we want to keep control support. Unlimited version skew and the second mission is: we want to provide SDKs to the two peoples who want to be able to say oh I, to make their life easier and the third mission is: we want to move Q control codebase to be out of the core kubernetes / kubernetes repo, so we can have our own development and release cadence.

C

So maybe like her, we can release. Bi-Weekly is much better than like three months and then our responsibilities, so the sick need to develop the roadmap and identify and the founded, valuable new sub projects and maintaining the sub projects, including like a bug, fixing and ensuring the test. Houseless.

C

So there are multiple rows in the cig. First, there are chairs chairs are responsible for governing the cig and currently in our sick. Our chairs also serves as the technical ease our current chairs has Margie. You want to wave your hand to others and also fill it and Tony from Alibaba with not here, and then we also have members with different privilege for maintained nurse okay approvers, so they can DTM and approve PR, and the four reviewers think I. Only LD TM button not approve a PR.

C

So let's talk a bit about our sub projects, so currently we only have one sub project, which is Q control and we are planning to have the customized as the another sub project for the sig it's still pending and under discussion, so customizes a template, free customized to of declarative configuration for its application. So the basically the idea is so you have a base for for your application and then you can have multiple variant and then you have a customized Toyama fail.

C

You can't declarative, specified, name, prefix labels and the annotations for all the objects defined in your configuration files. For example, if you have a application, you can have a deaf version of it and also have a production version of it. So that's the idea- and there is a link put industry in the slides where you can fund all these sub projects for all the six.

C

So our ongoing effort, so first we want to know where we are moving a client-side logic to the server-side, so we can make the cube control singer.

C

So there are two work-in-progress projects. The first one is a server set apply which will take several colors to finish because it complicated and then we have the server side printing. We are going to finish this work in this culture and it will be ten hour by default in the 1.11 release.

C

And the second effort is, we are developed libraries to handle unstructured objects, so we can avoid tightly coupled with the Trinity types so, which is a good thing for us and the third we want to. We are making Cube control works with extensions and version school in terms of API extensions. We are going with open, API schema based approach instead of using the static ghost drug and we are also using the sub resource and then we are also a building CI extensions, aka plugins. So it's currently in our first date.

C

We have proposals to continue improve this framework.

C

For the people who want to learn more details about all of this, you are welcome to go to the six ERI deep dev session, which is hosted by philip on friday, 1110.

C

So how you can get involved, so you can join the slack channel, which is a sick. Do I you can also join the Google group mailing list. You can post and discuss questions there, and you are also welcome to join the Sigma T bi-weekly sick meeting, which will be a 6 p.m. Denmark time and you can introduce yourself in the Sigma team. Let us know you.

A

Next week, yeah excellent.

C

So, let's talk about how you can contribute so before you contributing to learn a bit, how to read Co code and sign the C sign the since a FCoE, and you should also learn understand some basic community concepts.

C

For example, you need to know what is a part of what is life cycle and how does the service work under it? It's also strong recommended to understand a different object management model, so there are two different models. One is a declarative. The other is imperative, so Inc Renaud is. We recommend to use the declarative model and the in Q control q control apply as using the declarative models. So all the others are using the imperative models.

C

So our code base, so the CMD /q control, is the entry point of cube control. But there is, there are not enough of a lot of code there.

C

Instead, most code lives in the pkg cube control code so be sure to be familiar with it to contribute, and then let's talk a bit about github workflow, so to contribute, you need to familiar with our github workflow. So first, you need to fork the repo and then make changes to your local repo and then run the test, make sure all passing locally and then send the PR and gets revealed by the cig, River and cattle Aditya and then get approved, get approved by the segment.

C

A nurse and Kathy are approved label and the insurer, the priests admit its passing. So then the merge part can merge the PR into the upstream Oh.

C

Even though the queue control codebase is still in the terminal is large community repos, we are tracking our issue, mostly in kubernetes large cube control, repo the issues there are often labeled with size, small, medium large extra-large. That means you need to commit a different amount of time to work on different size issue. Your.

A

Hours may vary, but the relative numbers should be somewhat correct. Yep.

C

The issues is also often labeled with different type.

C

For example, there are cleaner types. That means you really need to refactor some package library and there are also documentation types, which means you. You are going to write dr. go further for some existing package and also there are features in future requests.

C

Usually you need to write a new library and there are back types and also test coverage, which means you will need to audit the existing test and read more test case for that will increase the test coverage so for the new contributors, we recommend you to start with a documentation issue, our test coverage issue because that's easier to get your PR merged and you can also be familiar with the contributing process, and you have also welcome to like a funder code that doesn't have a test and a test for them.

C

To get start, you need to adopt a issue so for the new contributor there are also labels labeled, with help wanted and good first issue to start with.

C

So we recommend you to do some research before you start and then please make sure you comment on the issue to let others and segment a nurse knows. You are working on that. This can help prevent multiple people working on the same issues.

C

Then, for the more experienced contributor you can adopt harder issues, for example, feature a new feature or a big refactoring task. So sometimes you will need to write a design dog and present in the same meetings and discuss with the segment inners and other members, and you also need to provide at least weekly update on the issue. So, like you can light the segment.

C

Who knows what is the status and if you are stucked and if you are stuck feel free to reach out to the maintainer to let them know and help you you can use either slack channel or on the motion at the monkey hub.

C

So there are multiple teams, github teams you can escalate. If you need, for example, there are feature requests. If you want to submit a new feature request, you can mention that so people related people will know and also, if you want to get more attention on your peers, you can mention 60 or I PR review.

C

So for people who are interesting to our hosts summer interns, there are multiple program versus since F is working with google Summer of Code. The second is the authority program, so you can find the links in the slice.

C

If you are interest.

C

So there are some general tips can make your peers moving faster, so first try to make your PR smaller, so a reviewer don't need to reserve a big chunk of time to review a giant, PR and second read tests. So there are three kind of tests for Q control unit, test, III test and also Sam D tester, which is a test event in a shell script and also it's recommended to write document and comments in your code. So it will be easier to maintain in the future and then there are some resources.

C

There are some links in the slice, contributor, guide and contributor Jaitley's, and also our 6ar I'm meeting node, and also you can find the recordings in the document in the dock.

C

Okay, anyone have questions so.

A

Yeah, that would be great.

A

Yeah excellent question so I'm actually on the steering committee and did some of the work on the charters, forsakes, so I'm, probably a good person to ask about this.

A

The direction we're planning on heading there is six, come up with their own charters and find their own rolls and there's some basic ones, such as reviewer, an approver which are well established and kind of well known, and we plan on most SIG's keeping those roles. Some SIG's like say, release, have additional roles such as related to making sure upgrade tests remain healthy or burning down issues or that sort of stuff, and so we're across the project.

A

We anticipate that different SIG's will figure out, adapt new roles and they won't be static in sig CLI, specifically one wall role, we're looking at adding is a role for health of the project like built copping, for instance, is one thing where we have these test dashboards that have our tests, but we don't have anyone responsible for reviewing those and making sure they're healthy and past. We had build crop across the project, but it was not I didn't scale and so we're thinking about.

A

Okay, well maybe we need a special role to allow people to contribute in the form of keeping those tests healthy, but maybe they're not reviewing code or something like that. They don't have time to. Does that answer your question or what were you looking for in the roles I.

A

See so specifically between those two roles, the difference there is the approvers are meant to be sort of pseudo tech leads right and who are familiar enough with the codebase to say. Yes, we do want to accept this feature, and yes, this does not have any weird weird semantical.

A

Issues that are you'd have to have a lot of experience with, and the reviewers are intended to be there to kind of the first line of defense and make sure that someone takes a look at it just because we're constrained on folks who really understand the code base. So the real difference between those two roles is just an amount of experience and the difference between an algae and an approval.

A

Effectively, I think that's changed over the time, so there's some confusion over that it used to be I. Think the intention was that you could lgt him something, and then approval was different for some reason and I think that got changed to be okay. If the algae team, something in your approver without it's approved.

D

Did the LD TM I kind of trust them? Maybe I've looked at myself a little bit too personally so, like the the algae. Tm is like, if you lgt on something it should be like you've, you've gone through and you've you've read through and you think it's good and then the approval is like. Okay, all the people who algae TM did I think they probably did a decent job and I've looked through it at Oakland's myself as well.

D

I'm gonna prove it because I think the review that's a has gone on is sufficient to go and that's oftentimes. What I see happening in the kubernetes project.

A

Yeah I'll, just repeat the first part of what you said because you might not have been picked up on the recording, which is that the LG TM is a soft approval, but generally considered good enough on the code level basis. And then an approval comes in is kind of like assumes that if the LG TM is there that it's it's been reviewed sufficiently, and it's just checking the box to make sure that the direction more cross-cutting organizationally, maybe another metaphor, is like.

A

Sometimes we go into the dentists like you spend a lot of time with the djinn like the first person and then the dentist comes in at like the very last minute and just looks at the x-rays and they're like okay, yeah and then leaves.

C

So the CMD test is a mostly a feature test that is right in the shell script. It has required our lives clusters. So it's a basically it's a trying to operate our life clusters to trying to mock what people will do with this feature. Does that answer your question.

A

Is the command test? The test LSH test, yeah ma che we're.

B

Actually having to rewrite the test significantly, but here's how we can think about it. The unit test is literally white boxing. Whatever the command is doing, and you know whatever gets in and the output, you should be verifying that it is exactly as you expect it. The object that you, the the parameter is you put in, are actually producing the object. As you imagine, the comment can be thought as an integration test, because it is it setups, a limited kubernetes cluster by that I mean there's only API server.

B

There's no notes behind it running anything, so you can invoke commands they will be propagated in the API server as they would normally, but there's no pause running behind it, so the majority of them are just invoking tests. Invoking regular cube, CTL commands and verifying that whatever you intended is actually happening on the server and the e two E's are all written and go.

B

There aren't invoking the cube, CDL commands, but they're actually invoking against a life fully fully working cluster, where a pot will be spin up for you or a resource that you define and then you're verifying. If the actual output of the command, including the full scope of the command or full scope of the your invocation, is happening on the server, it's a limited ete like yeah yeah, but the CMD is the one that will be we're. Writing with this exist.

A

I think that's not a unique feeling, I think a lot of people feel overwhelmed. I think I feel overwhelmed with much of the time.

A

B

I hope I felt numerous persons with all different types of code within the entire cube code base that I'm involved in so the easiest way is just reach me out on slack and be persistent. I might sometimes not respond due to so many different stuff.

B

But if you're persistent enough and you get me there, I'll help- you go if you need a hangout or whatever will do so. If a discussion on slack is sufficient for you, that's perfectly enough. I've held a friend of mine with writing, create job command.

B

I I've heard I've helped Julia advance, helping up creating her first cron job fix PR and then again, a friend of mine from Red Hat is, is pinging me very consistently about some other set and command that he's trying to to fix I'm, not saying that it my advice will be good I'm, you know I I'm, just a human I make mistakes. There's one that I know there was one that was super vague and I was so ashamed, but yeah.

B

It involves the create job from cron job, but generally yeah. It's not a problem, just reach out and slack. This slack is our slug is not that busy. So it's pretty quiet. The mailing list is pretty quiet. Eventually, the I don't see any problems with doing something like that during the sick meeting as well. If there will be enough interest and actually doing some kind of a oh, this is how we work with stuff or create stuff.

B

We can do I, don't know a 15 30 minute, walk through either fixing issue or actually coding something within a sample command. If people will be interested yeah, that's that's not a problem.

A

Yeah I think that's a really good point, I'm available to on slack I'd, like to echo ma che's point where there's like weeks at a time where I, probably am going to be tossed slack I think is the best approach for everyone. The important thing is you want to at mention someone if you just put it in the chat, something like I, don't think, there's a lot of people lurking there.

A

Yeah, would it be, do you know a bunch of people, or is there like a community of folks looking for help that we're not aware of.

A

A

We definitely want more people and we're trying to understand like how we're going to scale.

A

The other thing is I. Think Paris has a onboarding program.

A

Yeah, so what might be best is if you're like really interested in it is to possibly have someone from six Eli to be determined could act as a mentor in that program, and then potentially individuals could get both the support of us, but also the other folks in the internship or mentorship.

A

Exactly I would love to start a Ponzi scheme controls a Ponzi scheme, yeah.

A

It's gonna go in some Mikey press release somewhere 60 lightly. It says coop controls a Ponzi scheme.

A

Cool how much time do we have I think what one more question.

A

Yeah, so I'm gonna go deep into this during the deep dive. That's almost what the whole deep dive is about, how we're gonna support version skew and how we're gonna support extensions and API server. Aggregation like this looks a lot like the core, so specifically CR DS, when you say working with CR DS. What specifically do you mean because some things work with cities like apply works with Sierra T's.

A

Like which verbs do in custom I, do you mean? Can you add a custom verb like logs, but for something else for your charity.

A

Yeah excellent question: I, don't think that's the lowest hanging fruit for us right now, I didn't get what your! If you want to just have a CLI command to do something on a c rd distributing your own CLI. I think we need to figure out what the benefit you get over having it in qu control. Is it just the visibility like you can run qu control and it's under it.

B

So we we want to focus so the server-side stuff that we're moving all the way to the server, so the get and the apply that will improve the user experience that people have with the cube CTL, because out of a box to get will print nicely. How your CRT looks like in you as this year, the author can define what will be the output of a get command apply, should work without a box. That's not a problem even currently for more specifics. You are the targeted commands.

B

Verbs then I think that our direction will be to to just go with the plugin. We will be focusing hard on providing the the library and the framework for people to build a plugin so that it's easier. Currently, it's not that simple, not that straightforward.

B

So this is where we will be more focusing on to provide the same building blocks that we use in cube CTL for other folks, and that would be the the core focus that we will have over the next couple months.

D

So I could see.

D

Today, where.

D

To add new sub resources, you could reflect back into and today.

D

Some resources but I, imagine that if we see that situation improving in CR DS to where the point where you can do more complex, sub resources and CR DS, then that provides another path for a cube. Ctl.

A

Alright, yeah yeah I have a there's other mechanisms that I'll discuss during the deep dive but yeah the putting running server-side code that actually does the stuff and then dynamically having qu control, discover what those verbs are and adding them is something we're exploring.

A

Cool all right, thank you. Everyone Thanks.