Home
Contribute Contact Us Browse all meetings
Home Contribute Contact Us Browse all meetings
Kubernetes / SIG Cluster Lifecycle / 4 Nov 2019
Kubernetes / SIG Cluster Lifecycle / 4 Nov 2019
Previous Meeting Next Meeting
⏯
Add meeting Rate page Subscribe
youtube image
►
From YouTube: kubeadm onboarding 2019-11-04

Description

0:07 - Introduction
3:11 - The SIG Cluster Lifecycle SIG, agenda document, Mailing list
6:28 - The kubeadm release cycle, KEPs
14:38 - The kubeadm issue tracker (k/kubeadm), labels, deprecation policies
41:30 - OWNER files, sending PRs, test-grid
57:36 - The kubeadm source tree

A

Hello, this is Lumiere I work for VMware I joined the Kuwaitis project around 2017 in the summer of 2017 I started contributing to cupid year mass. My primary project I've been doing that since then, so this video is going to be about explaining to you what is the development process of qadian and how you can get started with it. Well, first of all, basically this project cube ATM. We have to explain what it does so cube. Atm is a coaster bootstrapper, also a kubernetes deployer.

A

So the purpose of this project is to facilitate the deployment of the kubernetes components on modes on your infrastructure, so that you can create a curious question and there are a lot of tools out there, but maintained by the sequence life cycle group. A sequence. Life cycle is one of the special interest groups that is responsible for all those great ease deployment tools like mini cube, cups, cubed e em chip spray and so on and cube ATM.

A

Ideally, all of those tools are going to use it eventually because cube ATM is a very low level building block it's a very, very low level tool. So to quickly explain what low level means here, a cube. Adm is a low level 2, but if you really want to understand like how is kubernetes deploying how to connect Cabrini's nodes into a coaster, maybe you should look at the KC high tower tutorial. That is quad core. It is a hard way and he this is an awesome tutorial.

A

He explains everything in very good detail, but on the other hand, tools like cuba didn't have exactly the opposite purpose, because kubernetes should not be bankable. It should not be hard cube. Atm has the purpose of facilitating this process. At the same time, keeping keeping it you know feasible to do low level development, for instance, you change something in the API server and you still deploy your custom-built colleges control plane using to succubae diem, and then you can join also some workers. You know we perform some conformance test tests, so this is.

A

This is a project that is comedian is a fairly fairly popular project. It's part of the kubernetes release. It follows the kubernetes release cycle and it also falls some of the principles in developing communities. So before we begin I wanted to do this brief introduction, but I'm now going to share my screen and show you some web pages that are important for the cube ad and Arobin.

A

All right, so this is the first page I wanted to show. This is the under Cobra this community, and this is the cost lifecycle special interest group page in here. You can see all the different projects this SIG is maintaining and one of them is a comedian. So we do meet every Wednesday.

A

This is roughly like 5:00 p.m. Universal Time. You can convert your time zone. You can see our agenda notes, so this is a basically a Google document that everybody I'm able to display who can view this, but you can join so this is a document that we use for pretty much orchestrating the our zoo meeting. So we are using zoom for the meetings and the way this works. We pretty much add a list of attendees like everybody should add their name here, and we also have agenda and agenda based in the agenda.

A

We introduce new participants, we show some PSAs like hey, there's a massive parking, courageous, let's fix it, it also excavatum and then within the general. So we have a list, a list of people and their topics like, for instance, for bridge. So here the square bracket is talking about something about NCD and the joint controlling command, and if you want what are the gen titles, you can just open this document, not your name and your topic that you want to discuss.

A

We also definitely do back back walk grooming, so this is like going through the pending issues in Pierce, and we just you know, assign people who want to work on that and we also like cause issues that are already fixed.

A

So this is the general document, but you have to also join our main emailing list. So this is the sequence life cycle, mailing list- and it's you know a regular google group.

A

So when you open this page you're going to see join button here, you can click it in your now following the seek activity once you have that I'm pretty sure you're going to be able to view and edit this locker alright. So what else do we have from this page? Apart from having all the meetings and agendas for the different projects, you also have a list of existing recordings. I believe this should be YouTube playlists, yes, so this is also the entire activity of the sick.

A

We have everything as a recording here, but you can also see previous recordings of cubanía office hours, I search for those and you're going to get an idea like what are we talking about during these meetings? And some of those can be really. How do I say this like technical, but once you start understanding the codebase and like what is the behavior of qadian? What is it doing, and also once you start following our issue tracker, you can pretty much start understanding what we're talking about. It's not essentially, it's not rocket science.

A

So let me let me call this page and I'm going to show you why cube ATM is bound to the kalidahs release. So if you go to google and type something like AIDS, 117 you're going to get information, but from this page about the current release, so cube, ATM follows the core industry.

A

Psychic abilities has a new minor version every three months, so this means that what Cory's mati is going to be released, there's also going to be 2 billion version 117, and this binary of Cuba diem is going to be included in the official I. Believe the note the covering is not tar balls for Linux and Windows sauce, so this is a page maintained by SiC release. That's another cig, obviously what they doing.

A

They are responsible for the release process committees, but I wanted to what I wanted to show you here is that there is a table here, and this is the schedule of the release. So the containers like we look at this table and see. Okay, we have to you, know, adapt our plan for what features and whatever we are going to add in this release based to this table.

A

So this table for the 170 release in particular, a very important date for us is like enhancement, freeze, that's the first one and I'm going to talk about what is enhancement freeze if I open and show you a certain upholstery.

A

So this is the greatest enhancements repository and it has a very interesting purpose. So kubernetes itself has become so complex over the years that whenever we want to change something, we have to submit a proposal. So the way this works is like. First of all, we have to log an issue in this issue.

A

Tracker so create an issue, and you know a issue looks like this and it has a template like it's an enhancement proposal and then we say like when is this going to be our fab, PTO or stable, and also there has to be something else and that's a so-called Cabrini's enhancement proposal. So for those you have to create a PR and there's a template for those as well. But if you look at the sequence lifecycle existing caps, let me have a bunch of projects here, like pasta, API at CDI diem in Cuba diem.

A

This is the list of our existing proposals. But let me clarify here: these are big changes in Cuba diem. Normally, when you start on the project, you were probably not going to care much about this yet, but these are big proposals likely changing the API or we are, for instance, adding phases for Southern, Command and I'm just going to let you know about this process. If you look at it, you have to you, have to open one of these files and try to understand what it does.

A

It has a pretty big template and it explains what are we going to change? It has a sections like GAO's motivation. You know proposal user stories, implementation details like how. How is this going to work? Even we have a test plan. Graduation criteria, examples it's a pretty detailed proposal, so a cap means covering this enhancement proposal. We do this as rubidium is still using this process of defining extensions or which can be changes or features in general, going back to the enhancement, freeze topic so by this particular date.

A

We have to be done and we have to be merging. We have to be done with merging all these new enhancements proposals, so we, you know, we want to change cube ATM to behave in a certain way. You have to submit this appear for this new file in here and then, basically, this has to happen before they has been. Freeze has landed, so this happened on October 15 for this particular cycle.

A

We already done with our proposals for hands cube ATM and now, let's look at the calendar even more we this is, you can also have exceptions for if something is not being able to land in the particular date. You can also have exceptions. That's the separate topic. This complicates our examples in this video, so another another important date is called freeze. So during during this time here during the enhancement freeze and code freeze, we have a period of time where we can develop this feature. The feature has to be sent as PRS.

A

Everybody has to be. You know able to review this, and the pier has to be merged before code freeze. This is about features, but if something is a bug, if something is critical, we can still send PRS for that after cold freeze. This is the code. The code freeze is more of an enhancements.

A

Look at more of a feature freeze so after this is like that only critical bugs can be sent to the code base until another date begins, and this particular date is called tall when this happens. This means that this means that the branch of 1/7 release 117 of the Cobra disk reduce upholstery is now not being fast forwarded, and it's a completely separate release, and in at this at this point in time we can. We have to start back porting fixes for 170, even if the release is not out yet.

A

We are already like backporting releases, sorry features or not even features with just backporting PRS to this old, like new branch.

A

So the next day that is super important for us is the actual release date and there's also this in between this cherry-pick deadline after this has happened. Well, there is a period of like a week where we cannot even send anymore cherry picks. So this is very time boxed, and this is the result of trial and error from many different. You know many releases of qualities in the past.

A

This is a process that is still evolving, but the release team and the sig release believes that this is optimal for the time being and keeping them follows this and actually I'm a believer that this is healthy for the project, because we had time box to work on this on these new features, we a time box to fix critical bugs and we are not. Essentially, we are not slacking on the development process of qadian the another day that miss Kier is a dog's dogs complete date.

A

So after we, after we did the cold freeze, you know we did a code freeze and enhancement freeze. There is a period of time where we have to send dogs, and this is actually doc tops deadline. This is for Friday than November the 15. We have to set our our dogs, the document new features, and then there is a dog's complete day that opions should be revealed. Our dog spirit should be reviewed and range merged by them. So we have a period of over like a seven days to complete that as well.

A

So this is pretty much about my summary of the release process of qadian, but now, let's, let's look at some of the actually. You know details about what we do so I wanted to mention like first of all, there's this cube ATM repository and the company on the kubernetes and there's also this folder cindy cube ATM under co, British communities, so comedian is still part of the main khorinis greatest repository and it's being developed as part of this main repository for historical reasons for him.

A

Basically, we have some import restrictions that we are trying to decouple. That's one of the reasons cube ATM still lives in this repository. Arguably still it's also easier to test this way, but in the near future. Hopefully we can break cube ATM outside of this repository, but now, like I'm, going to get back to this, to explain some like briefly. What is the construction of cube Adium, but I wanted to mention that we also have this repository, so this repository, if you go and look, what is the description?

A

It says that it's an aggregator for issues find fight against Cuba yeah. So this is like. Is this the issue tracker for Cuba diem? Well, the answer is yes, but also it contains some tools and also dogs, the main documentation for Cuba diem lives. In this repository.

A

So if you going here, if you go on under documentation, cubed M is basically spread in some of the sub pages. You probably if you have tried to bait them already. You might have seen some of these pages really going to tasks. You can go to administer course, I'm, going to find some of our pages that we maintain, but then why do we also have dogs in this repository, which is under aggregator? Well, this is these dogs do not really fit very well into the main query stocks.

A

These are more like of coring implement stationed in Australia, but we have a design folder and it is. This is slightly outdated, but it explains some of the concepts behind the original design of cube idea. Missive pretty interesting. You should definitely check these pages if you are interested in some of the original design. So also we have pages like managing and doing this like how? How do we do end-to-end tests in communities targeting cube ATM even like? How do we do it?

A

We also have some documents were like these need some updates, but we also have a release cycle document, which is pretty much overlaps with what I try to explain here, it's slightly more detailed at slightly outdated. We also have test failures like how do we? How do the blood test failures?

A

We also have testing privileges, did some interesting resources for you to basically read before even jumping to qadian development. We also have updating command command references. This is this is more like instructional documentation here, but also some design is covered.

A

So going back to the this repository, I wanted to explain that, like probably Cuba IBM is going to move from here, you know see Nikki baby I'm up it's going to move in here, but we also have some other stuff like we have kinder kinder is a tool that we use for end-to-end testing, it's a tool for testing to medium deployments. It's based on kinda very deploys doctor and docker quality setups, and it's a tool that we use. You can check the readme this too.

A

By the way, we don't recommend that you use it in Europe, I, don't know in creating your production questions, because this is a tool for testing it's designed for helping Canadian contributors below by allowing them to create, delete and modify local questions and also execute some actions.

A

So this is a tool that we also maintain in this repo. Something else is we have some logos and of the operator sole operator is a work in progress. This is something that Fabrizio is working on for Bishop and Denis from VMware. He also is working on kinder he's the lead maintainer above the operator and the kinder project the operator is, you have to look at look up in Google.

A

What is a complete this operator pattern, and it's probably going to give you an idea what we are working on here, but this is please mind that this is like working progress. It's experimental should not be used in production, maybe in the future we're going to have a release of comedian that is going to bundle this new feature, that is the cube ATM operator.

A

We also have some end-to-end tests, so those I mean it's kind of questionable if some of these should live here, but these we have Fenton tests for manifests, and this one is in particularly interesting because we use this entrance through the basically monitor. If all the docker images are the Google GCS buckets that we use for the point. Let's say a control brain are always there always up there. They are always supported so because we don't want to get all those users.

A

Reports, like hey I, tried to point out the latest, kubernetes versa, music to make them but I'm getting this missing images, error and the manifest. This is exactly for testing that we also have tests for packages, and this test is responsible for checking whether we have all those Debian and RPM packages also deployed to Google infrastructure, and this is pretty much the contents of this repository. But let's go to the tracker now. This is where, like basically most of the key beta and development happens.

A

This is where we monitor the state of what we have done in the past and what we have to do in this cycle. Well, we have what we have to do in the in the future. So, ideally, everything should be targeted with a mouse home like we have some milestones, let's request what we have so the correct one is obviously 117, and then we also have some like some from the past, like 150 116. We keep this. We keep these in the repo because we might tag something for backporting using these milestones but I.

A

Ideally, we should remove them. You know as soon as possible once like the new releases, come in action. So 117, if you click at this milestone, we're going to see some issues- adults, ideally issues that we should complete 117. But this is like a best effort. You have to understand that in open source, even if some of us are backed by companies like VMware so say Intel, we can only do as much so.

A

Basically, this list is a best-effort list. Like you know, we might based on priority. We might complete some of those items. Some of the items are super important. They are. We have priority labels that I'm going to explain in a bit but don't base achill. The priority here is that you know if something is high priority, we're probably going to do it in the 117 cycle.

A

You know, as per this particular milestone, but if something is a bit of a stretch like if something is with the priority backlog, so we might shift this into the next cycle because it's not a high priority for us. Maybe people don't have the time and we have to pretty much prioritize our work so I'm going to show you like the list of labels that we have quickly so there's a big list of labels, but I wanted to show you the priority levels.

A

So these priority labels are pretty much standard practice in the Kuwaitis project, don't buy. This needs more evidence because pretty much duplicate or for waiting more evidence, but starting from awaiting more evidence. This is like the low imagine that somebody looks an issue and we don't have enough information or like from the user. Maybe this doesn't really make sense for us to add as a feature. So this is the lowest priority for us. We're going to add this and pretty much keep the issue open.

A

Maybe, but we are going to wait for more evidence like that. You know the the guy who sent this ticket. They have to convince us that this is something that we might want to address. So this is the most priority. The next one is priority background. So this means okay, this we are acknowledging this problem, but we might do it in another release in the future. So that's pretty much the meaning of backlog. The Vedas also have descriptions on the side. So what else do we have here? We have important long term.

A

So important long term is like it's. It can be a bridge type of issue, because what I mean by that is, we might work on this in a 117 cycle. There is enough time, but if we cannot do it, it has to be moved to the next to the next release.

A

So it's pretty much close to backlog, but except that this one is important, so the backlog issue- hey, we might do this or not, but important long-term- is something that we have to do, and it's like waiting for somebody to get a site and work on it. The next thing the next priority is something that, like it applies to this release. Pretty important so means that this is not the highest priority of the list, but it's pretty important and we have to ideally complete this in this release. I come up very rare occasions.

A

We have important zone issues that get shifted to the next release and then comes the hospira will screech coercion. Critical engine is the type of priority that like if we have an issue like that look, we try to send a PR for this like in the next couple of days or maybe, if it, because there has to be discussion for us, it might be- might get delayed a little, but usually this appear for something as priority critical, urgent get sent almost himmier here all right.

A

So let's get back to the list of tickets for a mousehole I wanted to like quickly outline some of these items like the. By the way. There is a life cycle both in qualities. You might have seen this already, so there is a bot that if an issue is open for a very long time, you know, because we have so many issues in kubernetes qualities.

A

Let me show you all right: Cinco British colonies should have like 2000 issues and something like a thousand pairs. This is very difficult to manage so folks from sick testing. In particular, Eric fete develop developed this tool. There is a bot, so the bot is basically automatically triaging old issues is somebody so if nobody is touched, a certain issue for you know ninety days or something like that, the bot is going to close this peer, but it first goes to different stages and in the cube, a DM repository.

A

We use pretty much one of these stages and that's life cycle active, so the different options for life cycle. Let's look at those as well.

A

Alright, so there we have to look at these. The life cycle slash types of labels, so first of all its life cycle active. This means that, like okay, somebody's working on that we are at this label to cover it is because you know for new contributors that are not yet members of the kubernetes organization, it's very difficult to distinguish. If somebody, some of them are working on an issue you cannot get assigned.

A

If you are not a member, so I'm going to also I guess talk about what is membership quickly and all not files as well. But let's look at the lifecycle labels first, like life psychoactive. This means that somebody's working on this and then we also have life cycle still now. This is when the life cycle bought from Eric fated comes in. It starts marking a certain.

A

He should make a stale issue like nobodies, look working on that and then, after a certain period of time, it removes the life cycle, scale Mabel and it lies, that's the life cycle, rotten label, and this means that after another period of time, the issue is going to be closed to prevent that and keep this option. Sorry, this issue open regardless of the boat actions, you can add the life cycle frozen Mabel, and this means that the issue might live for.

A

You know a very long time because it can import them for us, but we don't have a good decision on how to proceed. So these are the life cycle. Labels like, let's, have a look at what else we have here whenever something like that is called tracking issue like occasion where you have these tracking issues. For instance, eggnog, like we have a tracking issue for windows support now. This is this is not an easy topic to work on incubate them, and this is going to take a while for us to develop completely.

A

So it's a tracking issue. It contains a number of PRS linked in there. Like have a look at this. We have so many different pairs. People working on that- and we used use this as a tracking issue too, as a hub for all those peers that we sent, and we also have a project board for by sick Windows perfectly okay.

A

So now we have all these different labels. So what I suggest you just hover over the labels and every label has a information for you. That's that's one way to learn the meanings of all these labels that we have, and you know this is github this low, like magicka way for us to distinguish between different issues apart from naming and labels and also the milestone. So we are users of you know a descriptive name.

A

We add everything is that it to a milestone if it's not applicable anymore to this milestone is going to be moved to the next milestone and also we have you know all those labels that try to qualify this issue to a particular area, but we have for these issues, but you don't say that hey all of those things don't make any sense to me. Well, that's when we have special labels for new contributors and those are a couple of labels. One of them is the help. Content idea was a good first issue.

A

So let's read the description of those by the way. I think you can google kubernetes good force issue now so kubernetes help audit and there's some web pages developed by some kind folks from the greediest community that basically have a very good explanation of what qualifies as a good first issue. What is a Help Wanted label as well? So, let's risk, let's read what we have here. So this the good physician denounced an issue ready for contributor according to the Help Wanted guidelines.

A

So this is kind of vague, but I want to explain to you that good physician means that this is something that it should be the simplest possible issue to tackle, and in this this type of issue, we have to provide you with a guideline like it's the responsibility of the maintainer, to give you a very good guideline to work on this issue.

A

So if you see a good first issue type issue- and you don't see a guideline in there- that that's it's not correct- there has to be a guide or um you know mini tutorial in there. So the other one is Help Wanted, and this is kind of this is a bit of a controversial label. We had you know little discussions with the sick contributor experienced folks help what that is.

A

If you look at the description exactly this, it's exactly the same as the good for a see show like what is the difference. So, according to the same Help Wanted guidelines, our Help Wanted type of the issue is something there is more complex from based on a good first issue, but it's more complex and you don't have that. You don't need to add this- the tayo tutorial in there. But you know still you want help from someone, and this is where the cupid image tracker is slightly different from career discovery.

A

This and we say, like hey, we are going to use a Help Wanted label if we think that that there's nobody assigned to this, but we want somebody to be assigned, so we are going to add a Help Wanted label and see if somebody assigned assigns themself, like let's click on this. This issue from yesterday I read the Help Wanted label, because I thought that this is a pretty easy change that somebody can work around.

A

You know, as somebody just immediately assigned themselves and that's already a coppery dismembered, because they can type a slash aside and you're the created CI robot Amelia size them. So this is an example of how the Help Wanted labor works. We have something that is important. Folks, come by a saying, okay, I'm, going to work that they might ask some questions. We are going to explain to them. You know some details if they need them, and this is how cube ATM contributes and also communities.

A

Could you just get on board that they start working on certain issues? You know they start looking at the code base like okay, I have a link here. Let's look what we have here, so this is a. This is not the best example, but this is like hey ad person.

A

Has this admission ponies like okay, maybe I'm going to Google, like the kubernetes talks, to see what is a lot mission podium and then like what is the meaning of this node restriction mission program and what do I have to do here so they also after Mike's a short explanation: I basically explained that we have to add a new admission control now, by the way we haven't decided should do this yet. But let's see, let's look at this as an example.

A

So what is what else do we have in the issue tracker? Well, we also have some peers, but dosa, for you know, dos these pairs here are pretty much fixing the end-to-end tests, the kinder and operator tools that I showed you earlier. These are not important. You shouldn't look at these. These are like internal. You can't have this internal occasional development PRS for this repo. But how do you send your like?

A

First PR for cube a DM isn't like before you send the PR I strongly strongly encourage you to log an issue first and I'm not going to stress a lot on this topic, but if you send an issue you're going to get a response from the qadi maintenance within the couple of days and if you're not sure, if you should be sending a PR for this, you know you find out found a bug, but it's kind of controversial you're, already saying that this is touching way too much like send the issue.

A

Send your you know, send a link to your poor request in your Fork of cooperatives and basically, let's have a discussion first, if something is complicated to be going to what go to the sequence lifecycle meeting.

A

And you can book that you can book time during the you know: the sequester I've cycled meeting, sorry, the QP diem office hours meeting. You can have an agenda item, we can discuss the issue and maybe it's only then is going to be appropriate to send this PR, because Cuba diem is a very mature project at this. At this stage, everything we tried to touch my defect. A very large group of people and breaking backwards. Compatibility is something that we like. We don't want to do every three months.

A

We release a new version of key medium, and this release should continue working for all those forms like no 117 is released for 116 folks, you should continue to work. You know the same manner. We have to keep backwards-compatibility so everything we touch. We have to follow some deprecation cycles. We have to follow the guidelines for API changes. So let me try to open some of those pages to quickly explain.

A

Okay, so there's a page in here that is strongly encourage you to have a read of like what what this is about. So kubernetes is a project that has to make some guarantees. So every every three months we release a new version, and what, if is there, is a feature that nobody wants to use anymore? What if you want to deprecated this feature? Well, this is where we have developed deprecation policy.

A

So if something is basically a feature that want to deprecate, we have to apply the deprecation cycle on it, and here we have deprecation for api objects. We have deprecations for the COI flags and also we have deprecation for features and behaviors. So you should definitely read some of those, but the T of they are here. Is that for other features we can disable and remove them. You know, without any notice we can say like hey. This is alpha. It was on. It wasn't a good idea. Let's remove this, as you can see like hey.

A

This is alpha is being removed in several releases, but for GA and please mind that cube ATM is a ga command-line tool. This means that every time you want to remove something that is a COI element from cube ATM, we now have to follow the g8 application cycle unless incubate them. We have this. Basically subcommander is called alpha and we have this for the purpose of basically having like a island of features that we can use.

A

But this island of alpha features that we can use to make it easy for us to deprecate stuff without following everything, if everything falling under the g8 application cycle. This means that we have this alpha command, a subcommittee of Cuba diem. We can remove something within zero releases, but we still conscious about it. So we might have, even you know, apply at deprecation policy of three months for for an alpha feature.

A

It's it can be a judgment of the maintenance of how to proceed, but we are very conscious of what we are doing with the deprecation of different. You know, parts of Cuba diem in Cuba the most has an API. This is the so-called comedian config file, and we, if we it's currently in beta, this means that we have to keep some of the our beta objects for nine months, and we are very conscious about deprecating the visual fields or adding new fields to the API, and this just gives you an example.

A

When you send a PR to modify that uranium API, it has to be sorry, it has to be scheduled, it has to be planned, it has to be discussed. This is very serious. No it's not that Cuba diem is a massive project is just the user base is very big, so we have to pretty much comply with all those folks that already using a certain feature from let's say, the API of Kapadia.

A

So, let's, let's see where we were before as giving some examples on what we have in terms of deprecation policy. Yes, okay, I, remember, so we please logon II should before sending the PR. That was my basically suggestion, for you know if something is a typo or or if something is a small documentation, change or no fixing renaming a variable or some it's fine. You can set the PR there's no need for an issue, but if something has changed the behavior or if something is a bug, but please lock the issue.

A

First, let and wait for a couple of days. You can have your PR ready in your fork and you can just send it to the community square disposal because I like I mentioned earlier. Cuba DM itself is still part of communities communities and you can go to CND Kapadia and now, let's have a quick look at what we have in this folder Cindi cube, ATM, so happy. We have. First of all, let me explain. Others and I gave a talk at some point about the kubernetes.

A

Called is like the governance of the project and kubernetes has many different structures and also different levels of membership.

A

So, first of all, once you let me let me find the page for that.

A

Okay, so this is the community membership page. So if we look at this page, we see a table and the first like membership role. You can basically receive in the communities community. It should become a so-called member, the Selective contributor. This is a person that has sent a number of PRS already for different projects and you can apply for membership. This page is going to instruct you how to do it and then you will be called a member.

A

The benefits of that are that you can now pretty much apply the LG TM level, which is slash unity, M to any PR that you like, which means that you're, a member you're a trusted person. You can review Pearson. You can say that you can get a PR one step further into the merging process that the LG TM label definitely does that, then you discover, if you're, so.

A

The difference between a member and the bureau is kind of is formal, but if you're is a person that is a distinguished reviewer for a project, this person can be present in all our files and I'm going to show you an honor. If I win in a minute and all our file is something that basically, the other file is a list of people that may be reviewers and approvers, so these are distinguished people that they have already participated in a certain project for a long time.

A

They are participating, they're, lopping, they're, reviewing their know like helping the contributors.

A

Let's change the PR in the right direction. That's like the main responsibility. They can also hold the PR if something is not right, so this is a reviewer. They get automatically pinked by bots for review like hey. You have to review this because you are in the owners, fam and the next row is an approver.

A

The prover is someone who can do the slash of proof label, and this is label that when, when is that it, and also when there is an LG Chem and a / approve as well appear, is going to be merged, but without an approved the PR, it's not going to be more gender. Usually the list of approvers is much smaller. These, like the false Fox, were like the most trusted people in the project. Those are the maintainer.

A

So, no maybe not the materials, but those of the folks responsible for the changes to the project so, and there is also the sioux project older, and these are the folks that reside in space, a specific yama file. You can click this and check this out, and these are folks that are pretty much owners of a sub-project.

A

So some project in live in a separate repository, for instance, like Britney, slash cops when supporting older. In these cases, justice at the barber justin is an owner of the cops project in his present in this owner files that you know distinguish him as an owner, and he such people can decide that they can even overrule a rule. Doctor was like hey. This is not a good feature like, even if this pair is already approved, like let's stop this cause, the PR go back to the drawing board and decide how to proceed.

A

There's also, this page also has a war of information like how everything works and should definitely cover it. Like there's, some interesting details like the members can do also okay to test. If you're, not a member of the communist organization, you have to wait for somebody to apply this command and only then your peer is going to be tested against the CI, but we can also members controls, issues and peers.

A

So let's have a look at the cube: a DM Orleans files by owner file. We only have one in this particular location, so we have a cross on the top of document. These are the folks that are responsible for the prove of appeared against this particular section of the communities, courageous people that is Jubei diem and also we have a bigger list of reviewers. So if you graduate to a reviewing cube ADM, the bots are going to ping you for review and then at some point you have to assign an approver as well.

A

So if you assign an approver, the PR is going to be merged best based on the LG TM and approved labels, and now, let's have a look at some Pearson and then I'm going to get back to some of the you know the structure of the cube ATM code base, but only briefly, so let me close these I have a bunch of useful links here and I also want to show this grid I guess at the end, all right so I have a link here. So this is this. Bookmark here is pretty much.

A

Has these filters, so is the PR open and is the area of the PR that is being in progress, Q, medium soul area, /g medium? This area label is being applied automatically by the bottoms, so everyone somebody sent the PR 48 name of a cake and also it for its religious, I click. This link and I check for new. You know new items and even if I'm, not an approver I'm, sorry, even if I'm not assigned to review this PR can still checked it because I have the list.

A

If something is interesting, I'm going to comment on that so.

A

All right, so these are some PRS that we have bending and pretty much. If you want to check some of the existing PRS, you can just you know, use this the same filters that I have you can see some of the discussion that we have in the past. You can see some of the peers are a problematic. You have some old peers that we don't touch, but overall we try to address stuff pretty fast.

A

On the other hand, Kapadia, like I, said earlier. It's such a mature project that you know everything that is being touched can cause problems.

A

So we try to be very careful about what we do and we try to be careful in the changes and when you send a pair for this repository that that's one of the reasons I told you like, you have to have the issue first and let me give you an example of a PR that I yeah I said this okay. This is a peer that I said. So when you send a PR and against the same applies to login issues in the Kapadia repository, please please always follow the template.

A

So template is important because it provides us with the user information. Why do we need this peer? Like? Is there a related issue like I suggested, always long dish? First, special notes like a hey I, want to explain something important to you like you cannot specialty here. This is also very important. That's this is filter has been automated but by our release tooling.

A

So if you add a release, not under here and you say, like cube ADM column- and you explain what this change is doing like in the 117 release lock, this particular sentence is going to be present and remember to always have this on the user facing change and also they recently added this.

A

This is you can link from here to an existing cooperate, this enhancement proposal, so you can link to the cap from this particular field, and then you can, if you, if you know which folks are competent in a certain area, you can even assign those folks, okay, I'm, going to assign Rafael for this there's a priority backlog, because this is not a super important problem and I can you know I can add some of these labels that I think are appropriate, so I'd label it open I, can add well the backlog, because I can establish the priority for this myself and then folks are going to participate in review.

A

Something I want to add here. If you see some failures, have a look at these red crosses. You know the failed jobs. Look indeed if something doesn't feel related to your er boost, pipe or a pet.

A

So this is a problem that we have in communities. We have some flakes. Some flake is basically a test is failing, but it's not really. A problem is just occasionally, you know connectivity between a Google cluster and the Internet can fail. So this is a temporary problem, so the way to solve it. When you see some failures like that, you can type the test. You know and the bots immediately starts retesting this.

A

But if something is really failing in the details, you can see like hey I, have this ppm problems and I sent a pair of Kapadia? Maybe your PR is really affecting the build process, or you know that Antoinette's signal that we have so your change might be disruptive. So please have a look at the details and see if this is problem.

A

Oh god, going further in the testing I wanted to explain some of the tools that we have for inspecting test results, and this is just the informative for you. If you are interested in you know, if you have a background in testing, you should definitely contact me and we can get you on board into like the whole development process of endpoint bits, but also, please go to be the repository.

A

Let your conduct again. If you go to the aaam repository, we have this. We have these docs in here about like how do we manage into intest- and this is a great page for you- instead, how it went missing before to make them, and this is pretty much the result we have. This is so this page is called case- slash SAR case tez grid, so this is test grade test very much. Our visual resource for inspecting end-to-end test results.

A

So in here we have like a lot of different sub projects and one of them is cube atm. If we keep clicking here, you can see some some current running tests, so it something is red, is connected. Failing because of this music app, you have to have a local base like what is the reason for this failure, like we have some tests that are in blue, so those those are pretty much passing.

A

So you shouldn't worry about those, but if something is red and if something is important to look great, this village, we have to follow, follow back and see like when this this started failing and we look, there is a magic trick here. Sorry, if you click and drag and release between the red and green, the there's a pop-up here that tells of a search for changes, and you might you know you might get a you might get an indication this.

A

This basically compares some shots between a cup of shots in the cub rediscovery supposed to itself is something changing comedia making castle. So why isn't there anything here? This means that something may have changed. So it's not completely square this. So it's not cube Adium itself. In this case, I'm, I guess, isn't something change in kind. So kite is one of the tools we use for testing. Also, we have to have a look at this, so let's go.

A

Let's go back to this of the cube ATM tree I'm not going to know I'm not going to talk about entrant. It's a lot, but this, let's say quick me quickly quickly show how does a test failure look like?

A

Maybe you might be interested into this particular area, but you know, arguably, this is probably one of the hardest thing to do when you get will get some failures in in the entrant, this you're going to get these red boxes. If you click on on the red boxes, you're going to get some the test result- and this is there's a massive founder. It's a blood work indicates nobody eats thousand lines.

A

But if you pick here on the failed part, you going to see that something went wrong with this particular command and if you click the mute walk you can, then you know have a look at the whole lock of the end-to-end test, and maybe, if you understand like how this works, you can figure out like. Where is the problem? And you can search in this particular a lock for special words like fail or error, and it might get you some information there's also a very important link here. That is artifact.

A

So every time we run a test, we after the job fails or succeeds with dump some logs in a artifact in artifacts folder, and this is where you can check. You know the the contents of or of every note in the coaster like here. Obviously we have three control plane nodes, three worker nodes, so you can click on some of those and you can expect like hey what are the logs for the couplet from this particular node.

A

You know this is doing very deep into the development, and you know these central tests are complex, but they also provide us with complex signal so that we can catch problems in you know the developing process. We also have a bunch of these like we like. We have some 20 recipes, so let me go back to the cube ATM source tree, so this is something that ideally, you should start looking at yourself. You start looking yourself like: where should I even begin contributing? I first woke the baby.

A

You have to say that you medium is a COI tool. It has some parts that are a front end and some parts on the back end and some parts are like they are like mostly utilities, but they can apply to both, but maybe they still can be considered back-end, but, unfortunately, Canadian work has evolved in such a way that it doesn't really has a clear distinction of where the back end is, but we definitely have some of that. It's not completely our spaghetti mess with multiple.

A

You know front end. The back ends portions mixed into the same files. It's not like that, but it's still pending for better organization, but for now I'm going to quickly explain to look how how does this project even work on the source code level? So, given this is equipment him as a CEO I to it obviously needs something that is a command-line section and common practice in the kubernetes is to name these sections as a folder that is CMD.

A

So if you're going here going to see the CMD folder and now, if you're, already a user of cubed M you're, going to see all those files I, firstly have joy in emails reset talking person, these are some commands of cube ADM, so you can clicking those files and you can inspect what is going on.

A

We also have some sub folders in here, so we have alpha. I spoke about this earlier. We have this alpha command that we care about is it can do some alpha features? We have some options. I. Believe me sonica. These helpers, yes, are constants things like that. We have phases all right, I'm going to get back to this in a minute. This is very interesting.

A

What's come public because upgrade by the way, this is being reworked right now, as we speak, but upgrade is a collection of more sub commands because you know upgrade has so many sub commands that we decided to create a separate package for that. We then also have videos, and these utilities very obviously applies only to the same department. So now, let's look at the symbol phases. So if you have been following the cube, ATM development, you've probably noticed merit phases for pretty much one of our commands.

A

You click in it join read, set the phases are a way to separate a certain process of the coop of the cube, ATM work fall and the goal of the phases is really to give you more power. Give the user more power. You can execute a certain phase of the needed process. I could give an example, for instance, one of the phases in the meter starting the colon. You might decide like hey I'm, not going to start the culantro. So let me skip this face and I'm going to execute all the other phases.

A

But then you have to understand how the you know the whole init process even worse. So if you skip a face you have to, if you still skip the start, couplet phase like maybe you might have to start the complete manual ease yourself. Otherwise, during in it you're not going to be able to both the control plane, static, pods, because that is going to be the responsibility of the coolant and then we also have phases for join, which is pretty similar.

A

It also starts a couplet, but it's more reliable on the existing control playing note that only you know the question is already formed. We can join using this command and there is also a set and upgrade north phases, and we have these these phases. We have this, this basically module in comedian. That is called the phase Runner I've, also developed by Fabio panini.

A

This is the module that is responsible for running phases. So it is a common back can be between you know, all those commands so in it has uses the in it uses the phase renderer joined. We said everything and it worked for can actually see some of the bits and pieces of this particular aspect like it'll kill is the rather here here is how we define a face, and it knows have some documentation here alright. So this is about phases and team II.

A

Let's have a look at what else we have here so API key beta has a configuration file, and this is our it's both for API object, the buffer configuration file. So kubernetes, you know follows this module of defining api's, so we have a llamo document with abortion and also it has a kind, and it also has a group.

A

So this is pretty much a API, but at the same time this is a config. So how do we call this? The name for that is a component config. You know there is a whole working group in communities called working group component standard. You can google that the goal of this working group is to define how to compress the cube, ATM define properly their API objects, but API object. The objects that are really configuration related their component config. So this is first of all, I want to explain.

A

If you already a Cabrini's operator, a comedian configuration is not being managed by the API server as a direct API type. This is something that people name manages it loads this. It writes this to a config map, but also the user has to feed. You know same values in a file in qadian then reach this file loss nor loss. This in memory creates a cost based on that.

A

So, if you want to check more, you can click here and pretty much look at our dogs. There's a very detailed explanation of how this works. Sorry, you have to click in a one of these. You won't be the one of you on video. These are our active, actively supported types checking going in hearing, you can click doc. These are essentially go dogs and it's pretty detailed. It means a lot of examples like how or like explanation. So what is this? How do?

A

How does the cubed in component config even works, and you can definitely check this out I'm not going to talk about like how do we do? What is the meaning photos generated files and registering types? This is really deep stuff, but if you want to understand more about this, just just try to you know have a look at.

A

Actually, this is a pretty complicated, but this is API machinery. This is the area of communities that is responsible for generating these convulsions defaulting functions and everything this is when, when a new API type is introduced, you're going to get basically let me open. Let me open the register, okay, so a player machinery is really the area of communities which is probably one of the more complicated ones. So the cupid EDM api type has a group- and this is the group it also.

A

This one is, the version has a version that is V 1 beta 1, and what what this whole logic here is doing is it register a schema with a grouping version, and then it enables some convulsion, so I can feed any version to dispersion of any any API version. To this version of cube, a diem in Cuba diem should be responsible for converting and understanding. This version now I mean that's all I'm going to say here, because it's this is digging really deep into some details.

A

So my my proposal is that you definitely should not stop with start with the API area of comedian, but you can still use it because that's we are making it easy to use but kind of difficult to maintain. So we also have comparative exacta medium, also the quality proxy at the couplet- and this is a package that this responsible for this particular area. We have plans to, or basically change this slightly in the near future. Hopefully this package is about constant. You know we have different constant BTU, UNIX and windows.

A

We have some unit tests as well. Here we have discovery. Now this company is an interesting package. This is the area where how how do we discover a joining mode? How does a joining no discovery discovering there is an existing posture? How does it validate if the question it wants to join is the correct form?

A

Oh, this is the package for it and we support different types, which support, burning, building up chip, config file using HTTPS and also using a token- and this have a little bit of an overlap here, because you know you can still have a pretty much boost up tokens inside a cube country. But if you want more information on that, I suggest that you go on the page. That is about Canadian coin.

A

Alright, so this page is going to, hopefully I think it was here all right. So here are the different options for how do we do different types of joint using cube ATM and they are advantageous advantages for all those one of the some of those are less secure. So if you're interested definitely have a look at this, this is what this particular project purpose or package is all about. So what else do we have here?

A

We have up in that folder also, we have features.

A

So this is pretty much. You know. Kubernetes itself supports feature gates, and if you talk about what a future gate is, you should definitely Google feature gates and communities in the same search quayle. It's going to give you an example of what is a feature gate, and the story here is that the cube ATM project did not find the featured need backing back and sufficient, so we created some I believe this is pretty much a wrapper at this point.

A

Yeah, we wrapped the existing feature gate support too. We extend it a little exactly as you can see. We have support for a couple feature gates at the moment. It's for core DNS. We also have one for ipv6 doors like I, believe this you should be removing colonists the colonists one at this point yeah, it's hidden all right. So this is my expression for this package, whereas we have images. So this is a package that is pretty much handling, basically the formation of all the images that comedian needs.

A

So let me clarify, like you, medium what it when you create a control plane, node cube ATM is going to pretty much start pulling images from the internet, and you know you can use a private local register as well, but cube ATM by default starts pulling from a currently format GCR bucket. So this is, you know, Google infrastructure, where all those images for the API server, hypercube or the controller manager recite- and this is this basic package- the images package in Cuba diem is responsible for the formation of the exact URLs.

A

For those images, that's how we basically pull these images.

A

All right, so, let's cut some unit tests here. What else are we having this the code base? We have phases now this is this. Is we have so? We have CFD phases and we also have no root phases, so this is the back end for the phases. So if there is a phase that is used during the command line execution, this particular phase might call something from the back end at the back. End is pretty much packaged.

A

For instance, if we we have a cooperate face- and this is this back end from the cooperate phase- contains some helper functions and utilities that we execute, and you have to look at this list like what are these phases like? Why, like? Why do we need parts of it in this back end? So some of the names match the sub phase names, because we we also have soup phases for the phases and the crab have a look at this. This is how you are going to learn with how cube Adium works. Pretty much.

A

And then we have pre-flight like pre-flight. These should be probably one of the things I explained at the beginning in this code base walkthrough, the pre-flight is something that we execute every time we run to bathing in it join or commands like upgrade. So basically we perform a bunch of checks. We check if the coastal is in a healthy state, in the case of upgrades, for instance, or in the case of anything, join we check. If this note, the machine where we run Q medium is a shelf estate.

A

Like we check for a you know a same version of talker. We check if the port's that Q medium is going to open by you know by starting the controller manager, for instance, is something if it's, basically, the machine is ready for this, and these are pre-flight checks. There are also different pre-flight checks, like checking I, believe we have a pre-flight check nowadays for how many CPUs, because cube ATM or control plane knows requires, like 2 CPUs, two CPU cores. So this you can also have a look at these.

A

These are the pre-flight checks we do. What else do we have? So whoever you choose you just package- and this is this- one- is pretty big. It contains so many different details. We have use for our connectivity to an API server. We create this wrapper of an API client. We have outed log support, we have certificate utilities, we have utilities related to the cube, ATM configuration format we have. This is I believe this some utilities ready to cryptography abstracting.

A

You know ecliptic the ecliptic algorithms, with RSA algorithms, so we also have dry run. We have EDD, we have innate system, init system is abstracting, I, believe the windows services again, and also the system D on Linux machines. We have keep configure till it is customized utilities, but these are so many different utilities that we have for different purposes.

A

So, let's pretty much it. This is my explanation of the existing code base. This is a subject of change because we have a tracking issue in the quality. Such could be the embryo that we want to organize this better. So this is going to change in the future. Maybe one day we might have only like apic-em DM p kt g, like the whole package, is going to move there. So this is a significant change again, if you want to contribute to keeping them by the meetings.

A

Follow our issue tracker check the peers that we do occasionally check the educational check, the peers that we send like the discussion that we have I strongly encourage like that. If you want to do something like just look for Help Wanted the tickets, if you want to learn more about the project, join the the zoom meeting, so that you made him officer also, we will we try to you know we have to.

A

We have to have a time box discussion, so we use some complicated terminology, but also we try to you know: oftentimes breaks break down the problem into simpler terms for folks who are neutral project to understand, so you know feel free to interrupt and ask the question like hey. What are you guys talking about kids? Can somebody explain to me like this topic, I'm familiar with EDD for the past five years? I can help in this area.

A

You're talking about HCD Park, so this is my I guess my whole, like in for informational video about the contributing to comedian. I, really hope that this is useful for you. If you want to get on board again, if you want to come backwards, you can also join the mailing lists and also our sequence, life cycle channel and then feel free to ask questions in particular, for if you want to join discussion on how to use medium, you can also join a parallel. It's called like hashtag Q medium, that's like all!

A

That's where all the users of q, p, DM, reside and most have some experts in their shopping, so yeah. This is pretty much all I have so yeah. If you have any questions, find a source like join this amazing, a song. Alright. Thank you me later bye.