Kubernetes SIG CLI, 29 Jun 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes SIG CLI 20220629

Description

Kubernetes SIG CLI bi-weekly meeting on June 29th, 2022.

Agenda and Meeting Notes: https://docs.google.com/document/d/1r0YElcXt6G5mOWxwZiXgGu_X6he3F--wKwg-9UBc29I/edit#bookmark=kix.45y8fo13yjfk

A

Forward to the cloud hello, my name is sean sullivan and I will be your host for the june 29th 2022 sig cli bi-weekly meeting, and why don't we immediately get into some announcements uh today it looks like we will be looking at a couple of uh the six eli prs.

A

uh All of the announcements have to do with the 125 release, and so the enhancement freeze has passed um and there was one kept that we were attempting to get in and on to be honest off the top of my head. I don't know if we did get that one in that had to do with the coupe rc file. Did we get that in manchester yeah.

B

It landed yes, okay, great, we landed on the last day, great.

A

Thank you very much for everybody who participated and made that happen.

A

uh That is the cap on the kube rc file uh and let's keep in mind that uh our next dates for 125 they're a little bit in the future, but there's going to be some vacations going on, I'm sure so we've got the code freeze coming up on tuesday august 2nd, and the release is scheduled for about three weeks after that on tuesday august 23rd for 1 25.. Let's keep that in mind, okay, so in addition uh or sorry moving on to our next section.

A

Oh, let me sorry, let me ask: is there anyone else who has any announcements that they would like to share with the six eli.

A

Okay, so our next section is to uh to introduce those who are new or who haven't been here for a while, so that we can all keep up to date with our colleagues.

A

um And so is there anyone I'm looking at the participants list and it looks uh pretty familiar: is there anyone else or anyone who would like to to introduce themselves today.

A

Okay, moving moving on so, let's get to the topics and the first is marcus's pr on json support and cuddle config. So I had just a quick aside uh marcus. I had a fast look at your pr and it looked totally reasonable and- and we may end up asking for a couple more uh tests as, as you know, that's a big focus to ensure that we have a really really good test coverage, um and I saw that you had tests there so cool. So why don't? I bring that particular pr up.

C

uh Do you think I could share my screen really quick yeah. Let me.

A

Just quickly make you a co-host, let me know if you can share your screen. Do I have to stop sharing.

C

Oh wait hold on oh no whoops. Okay, I have to rejoin because apparently I have never set zoom to be allowed to share on this desktop. So I'll be right back.

A

Okay, and with that, I will bring up I'm going to share again and.

A

We can have a quick look at the pr, so this is the jsonpath client go code. We're adding a parameter here- and here here is where the json path is parsing.

A

And adding to the value text, it looked.

D

A

Straightforward to me, here's the tests that he has so far.

A

A

Then there was some some added help text explain text! Sorry about that with the code being mostly looks like right here.

C

C

Basically had to re-implement the existing one, um okay, so um I'm going to not share this panic.

C

um Okay, so this is my cube, config file and so just as in a quick example to get out of the way.

C

um So this is how the json padding would work for, for example, here for clusters like if I wanted to set a new cluster named test cluster, then this is like the path that I would set for it and then, if I want to set server, it would just be cluster and then server. So we have to make use of the filter to be able to find it um and if it doesn't find it and we're using a filter with a strict equality, then it says: okay: well, we gotta create it.

C

So we create it and then we keep on going down. So you can see, I had an empty config file, and now I have this test cluster with the server blah blah blah and we can change it to whatever we want.

C

C

The main thing, though, is that um we can set off providers because that's something that's not supported by the regular set command right now,.

C

Some of the pattern kind of feels awkward, especially compared to, um for example, how the old dot dominated formatting worked. um This you know it kind of feels weirdly verbose because we got to do, but we have to go through the whole thing like an actual json path.

C

So we're going to like users and then looking it up by name and then we get the object that we want to go into, and so then we have to go into user and then username, whereas before it was just like users.test user.username, because I just skipped all of that extra stuff. um But then also.

B

Yeah, but let's not forget that, even though the uh the previous way of specifying the keys in the cluster map- because that was usually either users or cluster maps- was only used in this particular setting or for this particular command. Whereas this approach, where we are using, um let's call it native kubernetes json path, it's something that people are already familiar with when they are trying to. Let's, for example, get pods and as an output, they will use the json path.

B

So that's the benefit that we are just reusing, the same mechanism across everything instead of learning them. I'm fully aware that initially it might, uh it might look like uh limitation, but at the end of the day, it's just consistency across all the cucutal commands, yeah.

C

E

Yeah, this is also way more powerful than it was so very cool, yeah.

C

uh I'm trying to remember the names of the stuff that like because before we couldn't set lists, we can set map cues and sorry. I was just kind of going and not thinking but like, for example, like before we were never able to set the off provider stuff, and even if we were, we were never able to set that like, for example, this is a map, string, string, um value or sorry. Config is map, string, string and then refresh token is a key within that map and that would have not been able to be set.

C

I'm afraid of opening anything does.

E

This do validation, marcus.

C

Kind of um so like, for example, if I put in exec and then fake key and then try to give it a value, it'll complain it should complain. Let's, let's see, yeah unable to parse path at fake key.

C

um It does, and I think that one of the things that it doesn't do that I should make it do, is it doesn't really? I don't think it validates bullying stuff properly. You know doesn't validate.

C

um Well, I guess the the slices is fine, I'm trying to remember the name of one of the uh fields that takes a slice that I'm blanking right now. um Let me find it.

C

C

Oh wait: yeah impersonate stuff is perfect. Okay,.

C

uh User ty has groups okay and then um lists will take comma, separated lists and uh split them up for you. um So like test one test, two test: three and it'll create that group for you and then you have. You know the full power of json pathing to be able to mess with them.

C

If you want right so like if I wanted to you know, get into this array and be like, I want to make um the first entry in that, whatever test for then it'll go in and update it for you, um and we also uh say we wanted a second user.

C

So now we have two users that have as groups and say we wanted to set all of our user names to the same thing.

C

Well, whatever yeah, so not only will it set both of them; it'll also create it if it doesn't have it over here, that's pretty cool.

E

C

That works with.

E

All the filters too so selectors and.

C

So I wasn't exactly sure how far to go with what I supported. um I don't know how it should work for all of the filters as long as the filters make sense. The only thing that's like really important, or the only thing that I wrote anything specific around- was the double equals so like. If the filter node has a double equals, then it will create what you're looking for for you um during the process of uh walking the node list.

E

Nice, so the reason I ask about that is: I know that gke is changing their auth binary and I think aws is also changing their off binary, so we'll have an easy way for users to patch in the new um client command stuff. For that that's great.

C

B

Yeah, there's no question that I had because we are basically changing how this works. uh If you use the old dotted notation, I will that gives you uh give you a warning that oh, this is the old way and we're getting away from it.

B

C

uh I did not include anything like that. No.

B

It would be good to to say that this approach is deprecated.

C

B

That we will be slowly getting rid of it, because if we don't, we won't be able to remove it.

B

If we start the deprecation in 125 and per the usual deprecation policies, we will be able to remove it in 128 if I'm counting correctly, because that's a year or three releases.

C

Okay, yeah: I can make a note yeah.

B

So it would be good to say that using the data, the dotted notation is deprecated and you should be using the json path instead, um but leave the behavior, as is.

C

B

That's another thing:.

C

um I I don't I didn't implement. I originally had implemented a flag to indicate that that you wanted to use json pathing but decided later to just like check the first character of the string. That's provided, and that seemed like it was good enough. I don't know if anybody has opinions on that.

B

No, if you're, if you were able to distinguish between one another easily and just like you said it's sufficient to check that just the first character. If it's a curly braces, then it's a json path. That's sufficient! In my opinion,.

C

C

I think that it, oh sorry, one other thing is that I.

C

I also I I pulled in the kind of like plus minus list, editing logic that I had from the the original pr that I made that just extended the dot dominated stuff. So um if users wanted to like remove just a list item without having to look it up, like that, I don't know if anybody cares to keep that that functionality.

E

I think for labels we prepend minus right. If you want to subtract the label.

B

Yeah, that's something that I started wondering whether we prepared and I think we do actually, let's, let's have a quick look. No, we do after.

C

B

Just looking at the cubical label help and the last example is label parts full bar dash after the bar, which basically means remove, and I think annotate should have something similar yeah. So it's good yeah, that's! Okay! That's a reasonable approach!.

C

um And then the last thing really is um originally. I was using sets, but it was messing a bunch of stuff up, um but I um because I would deduplicate things so. I implemented a flag to deduplicate stuff. um So, like you can see here, we have test one two three and then another two which you know. Maybe that's uh on purpose, but um oh, my god. Okay, maybe I didn't. I thought I did.

C

Is it de-duplicated.

A

Or de-duplicate.

C

uh Let's, oh it's: okay, you're right thanks yeah! So um basically just runs it through the sets or is it sets list something like that? I can't remember um strings, no, whatever the there's like a built-in thing that um the that we have and it just like adds it into there, so it um it both alphabetizes and um deduplicates. The list basically.

B

Yeah we have sets for that.

C

Yep yep, so that's that's what I did there's one other thing, but that's not really important to share.

C

Yeah, I just wanted to show it off and if anybody could um there was a lot of stuff. I tried making sure that I tested everything, uh but I most likely missed things, and so if anybody would be kind enough to like give me a quick like beta test- and let me know that, oh I forgot, there is one thing that I don't know how to fix. um It's uh god. It's it's one of the types, that's kind of weird, and I can't remember what it's called now: the exec interactive mode.

C

I don't know how it won't. I can't figure out how to set that.

C

But I don't know how important that is to everybody.

D

B

Yeah that looks awesome um thanks a lot for uh for the patience and sticking with us. uh I know that especially myself. I've been pain in the ass uh with regards to reviews on that one, uh especially that after eddie did the initial reviews.

B

I came there and completely changed the vision of the approach, but um looking at what you did, I'm positive that the overall outcome uh for this command is just uh mind-blowing and functionality and the possibilities after your changes um are significantly superior than with just building on the old mechanism that we would have there.

C

That would be nice to delete that whole navigation step. Parser.

B

That's that's one of the reasons that I uh uh that I'm pushing to have that deprecation message, because the sooner we start that deprecation message, the sooner we will be able to get rid of the uh the old craft.

C

Well, I wish that it was bye, bye, reflection, dragons.

B

Yeah but it let's call it by pi reflection, dragons from sexylize code it'll be more in a uh contain client go code only.

A

Okay, great, shall we move on to arda's prs.

A

Okay, great, um so I'm actually going to start with the last one. First, since I think it's the smallest, the shortest one is that okay, arda.

D

A

Again so I think this is only this particular small change that fixes uh two flags which are incompatible. Is that correct.

D

Yes, exactly this is happening in replace command and I just added this validation to fix the issue.

A

Cool, so we'll we'll probably uh ask for a test uh if, uh if that's appropriate, uh but that actually seems pretty straightforward to me, anybody else uh have any uh uh anything to add.

D

By the way, I myself, I tried it and open the back, because I thought that driver and force flex simultaneously has no meaning okay.

D

If there is any.

A

D

We can even close that yeah.

B

Okay, but the bug will be fixed as soon as the pr merges. So I wouldn't worry about that. One.

A

Okay, cool- let's go on to since that was pretty straightforward. Let's I'm gonna go to the first one, which is uh introducing a new environment variable on where the caching is. Is that correct.

D

Yes, this was uh an issue with triage uh two or three months ago, and someone tried to configure default caching directory and we thanks to that pr. We will introduce a new alignment. Variable users will be able to change caching directory via environment variable because otherwise they have to pass flags in each comment. Requests.

A

Okay- and it looks like brian's already giving you some some feedback.

D

Yes, thanks ryan. He spent some time. uh We couldn't hear you by the way.

C

You're a mute right you're on me.

A

Mike so so I think yes, as artist said, we we've already triaged this and accepted it so appreciate you taking the initiative to to address this one and it it's looking pretty reasonable to me.

D

Thanks, I have a confidence that I'm not breaking any parts.

D

I can just say this.

E

Can you guys hear me now.

D

E

Yep- okay, great sorry about that uh audio strikes again so um yeah I reviewed. I I think I probably owed arda a review on this again. uh I I think it's probably good um I'll, take a quick once over on it, um but I think it's probably good to go.

B

Yeah I came through it. I I didn't see any any additional addition that we would have to go there. I would just tag it, as is.

A

Thanks cool: let's go to the the last one here which okay, so this one I had a question about, uh so we're adding new flags for the alpha events. Is that correct? Are they.

D

Yes, to promote alpha events to control events better, we need to be compatible with the control, get events and I added a lot of flex to just be similar with the get events and recently I added a lot of uh integration tests and I'll run this pr with these integration tests and it passed.

B

Okay, yeah- that was one of the problems that we had when we put together a cap for promoting the events to beta. um Unfortunately, we don't have almost non-test r for this particular pr. So.

D

By the way, I'm going to open a new pr new integration test follow up here to testing new flags. I just did not at in this piage, because that makes the pr bigger and bigger.

B

No, that's fine, a separate pr is okay, we don't have to do it in a single pr. Everything, okay,.

A

So I'm gonna add uh triage accepted. Does that sound reasonable.

B

I'll make sure to review this particular one tomorrow morning and leave the comments with regards to.

D

A

Okay, so we may, uh we may have uh some extra time uh for all of us this morning. uh That was all that we had on the agenda. Is there uh any standups or anything else that we'd uh like to address.

A

Okay, so we had a group uh kind of pair programming or peer reviewing session uh today, uh which was pretty cool. That's a good way to make sure that your pr gets attention, bring it up in the meeting. uh So we appreciate everybody showing up today um and I'm going to close the meeting and uh and hope that everybody has a good day.

B

Okay, thank you all.

C

Thanks everybody.