Kubernetes SIG Testing, 3 Jun 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: sigs.k8s.io/kind - 2019-06-03

Description

https://docs.google.com/document/d/1b9Ppm7ZT_tMWRs5Ph1zGJJKb5nF9c3ZHzMwg1olJIrc

A

So does not look like anyone had any particular agenda, those or anything, so we can probably just meet quickly and take whatever you all had. So this is a leading ender, a cig testing project under the C and C F. So this follows the kubernetes community code of conduct, which is leave the CN CF code of conduct, basically Dunphy jerks, and this will be publicly recorded and uploaded to YouTube.

A

Today there have you had anything you wanted to talk about.

B

I'm just coming along today, just say: hi I guess to be working on that dog sings reiterating on next yeah. Thanks.

A

For the pure yeah I think currently there's enough things in the air and there's like enough discussion. There are other channels that this meeting has kind of become a bit of a point, even though it's only by weekly we'll see going forward. It also got interrupted by things like cube con and we yeah.

A

You yeah so I. Looking at the agenda, the last last agenda set up for like basics,.

A

All of those things are being worked on, I.

A

Didn't have one no is going to discuss, but given that it's just the three of us I think I will make sure to discuss it by some other means. Anyhow, um do you have any questions or anything I.

C

Don't know anything in particular I, just kind of wanted to swing by a kind of lost, a kind of lost track and come up with some other things. They finally got some time back and now I'll say just wanted to catch up, see what everyone's working they are working on. Well,.

A

So we've got a bunch of work towards ipv6. Other small patches I've been spending time on trying to make things fast again. Urban haze 115 is about to come out and it's not as reliable speed, speed wise in time. I think I like that and then I think we're just planning like which things to drove the going forward.

A

There's a lot of stuff that people want mmm I'm, hopefully gonna ramp up one of my Google teammates who's, been doing a bunch of code review into sending actual code export and paid down a bunch of technical that like not really having unit tests.

A

But right now, there's like I, don't know 20 something open PRS to go through. What's.

B

A

Of them just need to be closed. Yeah there's, like infinite things to work on after come on. That's got quite a bit more popular a lot more attention, so yeah a lot more requests and whatnot.

C

So if anything is issue triage, reviewing a reviewing peers will be a good way of helping you get some load of um I. Think.

A

We at the biggest thing that's been super helpful for Oakland is actually the the kind channel.

B

A

There, because that one has actually way more of the like questions and stuff. A lot of people that are looking for help go there. I need to figure out how to go back through and, like politely, close out. Some of these, like we've, got one open for a couple days now. I guess I'll just leave it, but as someone's asking about deploying some specific thing onto it, it's like this.

B

Isn't this doesn't appear to be here like.

A

Something related to kind.

A

And similarly, we've got another thing where we have a lot of peers that are like they're kind of the right idea, but they're dot, they're, not really merged, ready and like people are just not following up so eventually we're gonna have to let them age out and then just open, more or less the same idea, but yeah fixed up.

C

Agenda slight so kind it's awesome, but at least time that they I'm not here I'm, not here, to try to add more feature say if anything, the thing that I was wondering is: how would you find new contributors, a wave? What would you say, what would you say is a good way for new contributors to help you up and don't be super busy without a with old things.

A

Like I, don't know um all the feedback useful, the documentation work has been super useful, like helping review. Peers is always welcome, but, like thankfully volumes not too bad, that most of them have been kind of reviewed, it's the figuring out how to like get them to a state where they're actually gonna merge like that back-and-forth.

A

A big thing that happens is just that people there's like an inkling of what happen, and someone goes on file as a PR, and then there wasn't any discussion about like actually how it's being done and then that winds up happening on the PR, which is really slow, I'm, not really sure that anybody can fix that. um Also, just like smaller fixes, like not sending I keep having to do this, to make big changes to fix stuff. But, like you know, really, small peers are much easier to work out and merge.

A

So if there's just like some little thing, that seems off, you know fix, but right now, um I think most of these that the code throughput isn't a big problem. It's designed, like figuring out. How should we make these things better, like people want to be able to access like node, ports and stuff, when they're on docker for Mac there's like a whole bunch of ways that we could hack something up in an afternoon?

A

That would work but figuring out something that's actually going to be good long term is just gonna, take more like testing stuff and exploring the options, that's way more time consuming than the code or the pure users, or also or like. Similarly, it's not very hard to write some half-decent entry on the site for like documentation, but the tricky part is going to be organizing it and figuring out what stuff should be added to it.

A

The low-hanging fruit keeps getting grabbed, it's been hard I've been trying to file like you know, issues that are like good first issues, but like most of the most of the currently available easy stuff has been tackled and a lot of the other stuff needs like bigger changes. First, like um like the unit testing, for example, it would be really hard to go add any right now, because things need to be mocked out more and like writing. The mock interface for these things is a lot more like opinionated and design than just like.

A

You know solving the code problem.

A

It's also not that much code currently so I've been able to pretty much stay. On top of that I don't know figuring out. That would be a great one figuring out what would be useful for other?

A

That's like one of the the hardest things are there some. There are a few issues that haven't been tackled, just because they need the right expertise like like, apparently a while back, we accepted Pierre to add zsh completion support along with completion and the CSH completion doesn't actually work, someone familiar with zsh and completion to figure out like what. Maybe we can do about that I think it's a upstream problem, but like we might have to do a workaround I know, that's a tough problem. I get asked a lot.

A

It's like for a lot of other projects. I've had a much easier time filing. Like here's. Some, you know good. First issues like.

B

If you look over intestine.

A

Bro we have a bunch of them, usually, but our kind every time I file a batch they're like immediately taken by someone who's actually kind of an existing contributor. Usually so it's really hard to bring on new people, because those those like good first issues, don't stick around they're, usually closed in, like under 24 hours, which doesn't give people time to find them. We only have like two or three right now.

A

A

Really don't know for that. One I was actually discussing this last week with one of the other existing contributors, and we didn't come up with anything. I think there will be a lot more in the future, because, once the were like when their codes merge, they born up it's kind of easier to go. Add some small things to it or clean it up. But when we're doing things like well, we're just gonna completely replace the networking to make ipv6 work.

A

It's hard to contribute to that, because the next thing that's gonna happen is some several files online PR. That just like replaces everything um that should go away over time.

C

A

It's not quite there.

C

We're good makes sense.

A

Yeah I think the probably the easiest one right now is the the documentation could use a lot more work and that's really easy to jump into I think we need to start breaking out some more details.

A

That would be useful for people to know that, like we don't want to make the QuickStart guide, you know, 20 pages, like a best practices guide, would be a good one with like you know, you should pin your images, so they don't change out from underneath you that's what I'm, even just a lot of things, just the activation energy of like getting a page like that started as really useful to figuring out like what would be a good category to have another page.

A

B

Also the whole known issues page and then known issues, kind of stems into best practices and like just seeing how that met up so yeah and.

A

There's probably more known issues that are not on this page.

B

A

That's that's hard to keep up with again. A lot of the discussion is I actually happening in the slack, and then it doesn't tend to get like really captured.

B

A

Be better about that I've been juggling too many things, but so the biggest thing I'm doing is just removing some of the other things, and so I can focus more because that happens the answer we answer questions or something, and then it's real, quick and then we forget about it, and then it has been relying on people that, like saw those questions, answered and then answer them again. More of these things should probably make it and do it doc. I mean it's still. Gonna have to go both ways.

A

People are still going to wind up, referencing it, but having it written down somewhere. That's not like searching through slack is also like. That's a that's a big thing. We should do.

B

Totally agree: cuz, it's like you know what I don't like I'm gonna go check the wiki open, slack yeah, but like I, maybe maybe like during the week, you guys need their. There needs to be some sort of capturing the hot take some kind sort of thing like that. This should be in a dark light, and then we can't come into like the meeting every two weeks or something in there like these were the best hot takes. Did we get them in sort of thing? Yeah.

A

The other thing that's gonna happen is that we're gonna probably want to our game on this in general, because this is gonna start to become standard for all the projects. Yeah.

B

A

Groundwork or like all turbine Eddie's subproducts, to be able to get sites like this, so everyone's gonna have one soon and I I'd love to you know kind of stay at the forefront of having one of the better ones, because I need to think that working on those shows that, like you know, we actually care about having users and whatnot. Whereas, like you know a lot of our products, don't have to be Doc's yet at all. You can kind of see a prioritization there, and this one has willfully fallen pine a bit on priorities.

C

Also, since we're talking more blocks, I think I'm wondering this for a while a day. What would you think about having reference of where they actual document the a somewhere in the actual docks to tutorials or talks that people have given of all kind yeah.

A

That would be a really good page. That would be like a like a resources page know and like embed, like the talks and slides and stuff. That's another one things like I thought about: it. I forgot about it after this little general, adding categories like that would be and like making them separate pages would be really useful.

A

Like we have one where men are on private registries, because I was only people had questions a lot about, it actually needs a little updating the even that granular like so. If you can, if you can pull together, if you think we're gonna be able to pull together a page of content on it, it should be a page and I. Don't think there should be any big restrictions on what we put resources would be create. Actual tutorials would also be really good way when cute tests kind of gets sorted out.

A

I really want to have the tutorial I'm like actually the oh workflow, for working with kubernetes I'm, really written one down anywhere yet because you probably the best one right now is do just use kind directly speaks for itself, but eventually we should have an easier one. That includes like the actual test running and stuff.

A

It might even be worth writing up now.

C

Also, we're talking about gonna press a worthwhile to give it a try, even at some way they just give it a try somewhere somewhere within the project. You.

A

Could but you'll find that almost all of the PAC news right now involve talking to docker and until we break that into a somewhat more remarkable interface, most of the packages are not unit testable, like they're, just not because they're gonna turn around an exact stuff on the host yeah.

A

We we need to finish breaking the boundary between the host and the code way too much of it doesn't have that boundary. Well, it's sort of in place, but it needs actually fixed the like they configure. Something also can too, but like that's, not super interesting right now, because it's so simple I think, like tests are like best used to enforce that. Behavior doesn't change as well and, like the other things most behavior is supposed to change, but.

B

A

I'm expecting that to change in the next couple of releases, sorting out some of this network stuff would be big and making the docker and like ho stuff decoupled will be big. Once we have few things, I think the code will be a lot more stable, hopefully, and then that is like immediately. We've got to turn around like identifying we're good winters to run and have more unit tests or, like we've, got one out right now, like we have bash code, it's not going anywhere. We need to shell check it.

A

I'm sure, like other other quality things like that I know the build code for like building a kubernetes node image is kind of horrifying right now, I was like I'm like I, don't know 20 branches in it or something it's kind of a mess. That's another one! That's like it! It's not so much about you unit, testing that it's like rethinking and modularizing it, and then we can eat it test. It.

A

But once once those things are more modular, it should get to where it's gonna be pretty easy to get a like, adding a test and test cases, and that will actually just be time-consuming and that would be super helpful to them, but I'm pretty sure almost none of the back itches will work for that. Right now,.

A

It's a major deficiency of the project, I'm actually planning to try to get like an objective here at work that covers like paying down technical debt. It's like a leaguer it out it's useful! Now we need to make sure it doesn't like become spaghetti code.

A

It's also, if kind was very branching, but every time we add like a config nog, we add like another branch somewhere in a little bit different behavior and we're getting the point where we're not covering all of that in the CI. So yeah.

A

Anyhow, thank you all for Jesus, a super boring meeting I had we've definitely had some exciting ones in the past, but I think right now, there's like enough things in here that are like ongoing people are not worried about coming in, seeing them because it's been integral I P v6 work is huge so like that has a couple most trying to tied up I'm sure we'll be back discussing new features going forward like some of those like big ongoing ones, Land.

A

Yeah and asleep and yeah any any creations are definitely welcome. If you're trying to make big changes, though maybe you know sending it somewhere to discuss but yeah happy I guess. Is it still? Your probably Tuesday now happy.

B

Tuesday yeah yeah living in the future.

A

B