Kubernetes WG KMS, 24 May 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: WG-KMS Bi-Weekly Meeting for 20220524

Description

WG-KMS Bi-Weekly Meeting for 20220524

A

Hey everyone: this is the sagad kms meeting for may 24th. This is our 16th meeting. um I don't know if we have an agenda or stuff for today, but I I think top of mind for me is what can we do to try to move the cap forward? So we've gone ahead and put it on the agenda again for tomorrow's sagat meeting.

A

Anything else we can do to maybe try to move this forward.

B

When is the kept deadline? Is it two weeks from now.

C

But now it's just three weeks: it's like mid-june, like the 16th or something.

A

We're on 125 right yeah, um so I'm looking at enhancement sprees is june 16th yeah, so they did push a little bit forward. So we have a little bit of time. um So that's that's good but like this is still a really large cap with a brand new api with barely no review so um yeah.

D

Does it happen often that um it's hard to find the reviewer and that? um Because of that, we missed that line.

D

Did you repeat your question kristoff, because I think I missed the first part? Okay, so does it happen often that um that the cap is not gonna make it because there are no reviewers available? Yes,.

A

Roughly speaking, I think most of enhancements.

A

Exception requests are basically like, hey, like it took xyz like a while to get into place but like by the way we've like already implemented most of the code. We just like finally got around actually getting the all the all the paperwork in order- um and I know it stresses out the enhancements folks because they they aren't following the code right. So they don't know the state of things and they really do sprout on last-minute things.

A

So yeah I don't, I don't want to get to the state where we have to ask for an exception right, like exceptions have to be in by I guess july, but I don't want to be arguing with the enhancements team that hey I'm making an alpha thing. So I should be allowed to go outside of the recycler.

B

Yeah, I think, there's a few outstanding. I mean few comments from mike and I think we've answered all of those and updated based on his review feedback. So there's nothing outstanding. So hopefully, if we bring it up again tomorrow we'll get more reviews on it and then we can get it to a state where we can merge it by next screen.

A

Yeah, like in 125, we can have the cap, you know set as implementable with you know the alphabets there like.

A

That means we can basically, at our own pace, start working on implementing the work in its own sandbox right and if, if nothing is in like a great state in 125 like it's, okay, like it's, not a big deal, it's the fact that, where we're able to make progress without coordinating with a bunch of other people- and that's where the value for me is- is right, because you know, like I've, been really busy the last few weeks, but sometimes I'm not in business or you know, I go through those cycles, so those don't necessarily line up with kubernetes release, windows and enhancements.

A

A

I think my biggest concern is like if someone is like basically like trying to fix the beta api in place. I just I really don't want to try to do that.

A

It's just because then you have to like get every bit of backwards. Compatibility just right.

A

Are there even.

C

People interested in doing that.

A

Well, like one of mike's questions was like what prevents a plug-in from using a key hierarchy today I was like well technically nothing, it's just that, like basically a hack, then right, instead of like a like an actual supported, feature you're like sort of forcing it in, and I was like, but you know when you combine that with a desire to change the storage format, our pc schema and key rotation bits like when you have all of those together like yes, I could try to very carefully insert those into the beta api.

A

I really strongly hope I don't break it and destroy people's clusters by encrypting their data wrong, or I could just do it in a sandbox on his own and very carefully build it out and test it.

C

So yeah- and I guess, like also only allowing this kind of features to be available in the new alpha api, also can make the users do the transition to the new api because, like if we don't bring any value with this new api, how are we going to motivate them to do the transition right.

A

Yeah I mean you know so, certainly that that aspect is there um there there's also because these are separate sort of encryption concepts within the encryption config. You can have both of them at the same time right. So it makes the migration story really nice, which is that whenever you're ready, you add the new v2 api. As one of your read keys at some point, then you can turn it into a write key after you've done your coordination and everything- and you know you get a really graceful path.

A

Whereas, if you were like messing with the beta api in place, we would basically be trying to do the migration for you over release boundaries. I just don't want to be responsible.

A

I I do think there is a certain uh aspect of you know. The beta api is already deployed and used, and so you know there is a there's an inertia to that, and you know we are creating work for all the current deployers of beta. um I I do feel like the v2 changes provide enough value to warrant that right.

A

And you know, maybe the answer is a very long deprecation window for the beta apis.

C

Yeah because, at the end of the day, you have to take a decision regarding beta, like you cannot leave it as better forever like it needs to be as a stable or deprecated.

C

That's the two direction that it can go in and after, like all the the discussion, we had it's clear that we cannot just make it stable because, like there is so many thing in the api to change already, so the correct path should be application, even though, maybe longer since it has been um already like in place for a while- and it depends on like plugins and stuff like that.

C

But I think yeah pushing the users to move to alpha is the correct answer.

A

Yeah and like one of the places where we've sort of left a beta api for at least for now, is like client core credential plug-ins. You know they went ga in like 122 or whatever, but the beta api is still there and we so far haven't decided on when to remove it, because it's an api basically implemented by cube ctl in that type right.

A

It's it's deployed to clients it's much harder to like go update all your clients, whereas this api is completely opaque to the outside of the world right, so we're all we're talking about is, like you know, doing careful, rolling upgrades of your api servers, which you have to do once per release anyway.

A

So if you give you a bunch of releases to do this over, like I don't think we're necessarily asking for a lot, uh this is assuming that our reference implementation is very easy to use or a reference library is very easy to build a production-ready implementation of the v2 api for the different providers.

C

um I would say it.

E

Must be didn't.

C

At some point, we mentioned that we wanted to kind of meet with some of the plugins maintainer upstream, that are still using, like the beta api, to see like kind of gather, their feedback about. Maybe the new announcement, as well as like their kind of maybe like their fear regarding, like the migration to the new api and like what that will involve for them and, if, like they are ready to do this kind of step.

A

Yes, we had so maybe maybe the first step there is to look at the contributors for the three flat provider source ones and try to tag them on the cap and be like hey. You know, you see, you appear to be maintaining one of these yeah, I'm about to create work for you. Would you like to read the campaign, and this did you work on the larger one yeah.

B

I maintained that so we have one. Yes.

A

We have one plus one because you're like that's fine like because I I'm I'm already ready for this. I love that oddly.

B

A

Okay, so we just have to find uh the gke and uh yes.

C

Folks, I think it can be good feedback because, like there will be the one in particular like that, even though like we will bring the library that they will be able to reuse a lot like they will still have to do some work.

E

I'm just going to leave myself a comment on this.

E

C

Should we decide also on the deadline to finally review this or escape like let's say um if we don't find any more rejuva by the end of this week, then we go with like make as an approver or something like that.

C

uh I mean yeah.

A

I think we basically have to make that decision tomorrow and say: oh that's, right. Basically, hello, you're, you're, the owner of this game, you're.

F

The only people.

A

I can I can do this anyone care to do it.

A

And and well, and if we, if we need like if we are having a hard time finding someone to that, has enough bandwidth or whatever, even if they're like willing, but they don't have bandwidth right now, we could try to say that um we won't go until to beta until we have a like a like, basically a very thorough review of code and kept that way alpha, you can just delete at any time right. That's that's kind of the point, so um you know that that might that might alleviate concerns.

A

um Yeah I mean I'm not trying to rush this thing into a beta.

A

My experience with beta is you better make sure it's basically ready for b1, because people treat it as we once once it has the beta attack. So I'm going to ask for a pretty high bar.

E

What else is on folks is mine.

D

um With the reference supplementation accuracy pro request and I'm slightly hesitant to continue until I get some don't know, plus ones or feedback or anything, um because I would not like to develop on top of something that might be completely uh kind of thrown away, um as we did with the pink one. So um yeah.

A

I I I I 100 agree: don't don't don't keep tracking on without feedback? um I do apologize. I have been very very over committed recently uh too so I have. I haven't, worked on reviewing code or even really any anything, whether it be the cms stuff.

B

A

B

Take a look at it by tomorrow, hopefully,.

D

Okay sure definitely, and um it would be cool to have a second pair of eyes on it, because sometimes when everything is in your head, it makes sense, but the other person's like what the is going on. So it always helps a lot.

D

A

Yeah, no, I I I I don't know if you all ever have the feeling when you read some really old code and then you realize you wrote that code you're like and what I was doing. Why was I doing this.

D

Yeah yeah this.

A

D

Usually, where, when you stop caring so much about get playing when your first time start complaining complaining to get blame, see your name on. Oh.

E

Anything for me to be.

C

The same like until kev frieza will have like barely any time to really focus on that, like maybe I will be able to focus on the game, but besides that it will be tough, but after that's done, yeah you'll definitely have more cycles to look into kms. I think yeah, that's mostly the case for all of us because, like I am being a bit of a special period right now, yeah I mean.

A

Like you know, personally I I am, I am okay if nobody does anything but try to get the kepners, because to me that is goal number one and I would be disappointed if we didn't actually get any code like mers in 125 related to the alpha, but I wouldn't I wouldn't be like upset about it.

A

It is what it is where we've got a lot of stuff going on.

C

Because we have like two months of development cycle after they get freeze right, awesome.

A

Let's see uh code freeze is right at the beginning of august, so one and a half months.

C

A

That's tight, yeah! That's why I'm saying like I I just I I rather have the idea approved and work started at you know our own pace than to be worried about any deadline other than the enhancements. Basically yeah.

D

B

D

Maybe if maybe maybe the cap is kind of big, the topic is not um very clear in the minds of the reviewers, so maybe it would make sense to have a dedicated session with them or something like this, where they can like in person, make questions or they can prepare some questions. We can prepare something, maybe more mermaid charts or whatever just to make it more pleasant for them to review it right, so sort of reduce the obstacle to even start.

A

Yeah I mean I I'm totally for high bandwidth reviews. I think part. I think the only reason the psp v2 stuff got off the ground is because tim and jordan ran those meetings for basically months and just forced the people that had opinions to share their opinions, and you know slowly work through all the things.

A

I don't think we're working on something nearly as controversial, but um uh you know if mike or whoever is that's, you know willing to sign up to do review and approval. I would prefer that you know I'm I'm totally for that. I find those to be much easier than trying to like write massive paragraphs right, because you know mike asked a question and I tried to answer it with every bit of detail.

A

I could come up with like there's like a small essay written as a comment now uh I I don't really find those fun to write just because they take a long time and you know sometimes you fail to actually answer the person's question because you're answering the wrong question um yeah. I I think those types of things are great. It's just it's hard, it's hard, because we all work at different companies to find overlap in our schedules.

A

I do want to be careful to make sure that we- and we have time but, like my my desired state for the cap, is that it stays as close to the code as possible over time. You know, obviously it's going to drift one way or the other right.

A

I don't want to get into the situation where people read the cap and like believe, the system behave in one way, but it behaves in a completely different way because we changed our mind. Yeah.

C

We can always change the cap afterwards, like yeah.

A

C

For to better, for example, or even like, even before that update the cap, while doing the implementation.

C

Like unless the feature is like stable, we have like some bandwidth to modify the game and even after I guess like, if you want to update the cup to reflect, what's really implemented, then I think it's always good to do so.

A

Yeah, I know I I I did that with like time for prudential plugins, where I I very carefully made like I. I can say that one is that cap is actually done because it one for one matches the ga implementation yeah, it's a nice feeling other than the cap and review on the reference implementation. You folks have other things you want to chat about. We still at least yeah. I forget. Is this an hour-long meeting?

A

Do you folks have other things they want to talk about? Do you want to take about half an hour back.

B

I was just thinking like so I have that uh branch in my fork which implements the v2 stuff. So I was thinking um I will create another branch where, based on the changes that's being proposed in the cap and all that add the new proto, so we can have like a work in progress. Pr open out there like, if you want to add the changes in bits like I was thinking, I can get started on the changes. Now we don't have to wait till later. Like do you think, that's fine.

A

Yeah, I think it's fine uh in terms of like code stuff, without what I basically figured we would have is something like we described as um a shared branch on one of our sports. That is open as a pr against uh the main kk repo, and then we would either directly commit to that branch or make prs against that branch. Whatever is comfortable, and you know, as those get funneled into the the pr branch you know, ci will run against it and we can see what happening um and.

A

You know we should try to hopefully merge something in the 125 cycle. That would be. That would be nice to have something worse than um so.

A

You know if we feel, like you know like as a for example like if we wanted to have an initial pr, that's just like the proto and maybe that shell script to help keep it up to date. That would be a totally fine thing to add in and then just kind of start layering stuff on the top. um So, but I'm also totally fine like if we want to go the whole lily cycle and just have one one branch that we're just all working on together. That's fine! I I don't!

A

I don't think what we want is a bunch of separate pr's open, because I think we would just be fighting each other right now.

B

Yeah but much conflicts, yes, but yeah I can get started so I was thinking at least I'll start with the feature, gate and the proto. So I'll add those two in one pr and then we can decide how we want to do after that.

E

Yeah that makes sense to me.

B

uh For the maintainers, I think for gcp it was all alexander, but I don't think he's in google anymore right.

A

Alex um immutable tea, yeah yeah, he yeah he's gone he's he's, I think, he's at like a hedge fund. Now.

B

Okay, I think, after that it's been seth who's, maintaining it, so we can probably tag him set fargo. I.

A

Don't think I know this person so sure yeah yeah.

B

I think we can tag them also, do you think it makes sense to create github issues on the aws and gcp repo, saying, like hey, we're working on this v2 stuff? Here it is for review. Let us know if you have any feedback.

A

Yeah, that's that's! That's very that might get uh that might help. They get broader attention from, for example, they're like more product leadership. You know you know they might be watching those to see if there's any feedback or whatever uh yeah. So I think that's totally fine yeah. Okay, like the issue, that was the issue and we'll say sorry.

B

Yeah, I think I can open one for gcp and aws.

F

Okay, cool. Thank you.

E

That was exciting.

A

Anything else anything else.

A

All right we can give each other a half an hour back and see all the moving like. Oh, I don't know who I'll see tomorrow. I think unless you're probably are you back in u.s time? Yeah, like your background, looks like you're. I think your apartment right, yeah, it's a different different place.

A

At least I'll see you probably tomorrow, yeah, that's more cool.

B

Cool. Thank you.

D

See you see you bye.