Kubernetes SIG Apps, 5 Oct 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SIG AppsBi-Weekly Meeting for 20201005

Description

SIG AppsBi-Weekly Meeting for 20201005

A

Welcome everyone to the monday october 5th, 2020, kubernetes, apps and I'll start by dropping in.

A

To chat the link for today's agenda, if you don't already have it.

A

And the announcement uh there are- and many of you may already know this- uh the two new labels- um the needs, triage and triage accepted label using the accepted um or triage slash triage accepted to accept uh is this up and running. I haven't looked.

B

I could see those labels, but only about two pr's have them so, okay, I assume it's just for the new prs, so we still have to trash the old ones and then switch over to two days. At least that's my understanding.

A

Okay, I like this it'll, I think it'll be helpful in triaging things, uh but that was the the one announcement that was janet's. Unfortunately, she didn't make it the next one we'll get into the discussion. There's the enhancement slip and there's a spreadsheet of enhancement slip and one of them is for sig apps.

A

Actually, there's a bunch for sig apps, there's the pod disruption budget, the max unavailable for stateful sets and the execution hook, and unfortunately I have not been following these: does anybody have any details on what's going on with the slip on these.

C

I can talk about the pdds to ga, but that's a separate topic, so we can do together. First.

A

A

All right I'll follow up. I think ken was the one who was working on ken and janet knew. What was up with these, and I expected janet here- I'm surprised, she's, not so uh I'll follow up with them offline and we'll see what comes of that then. The other item we have is the pdb to ga. Do you want to take this away or the next item.

C

B

D

C

D

B

C

This is something I was working on a while back. um I haven't looked at it in a while, and I don't like to have a time for it, but it's also we're getting close to time where it needs to go to ga or with dedicated. So we need to make progress on this.

C

I've started to look at it again. uh Let's see where we are. I think the issues theme seems to mostly be the same, and I have to oh or I reopened them now, because they had, you know automatically closed into. um I think those two covers the two most important open sort of issues that we need to close before you can go to ga um the project and the link to the doc has a list of the different open issues that have been sort of deemed critical to close before we get into ga.

C

um So, if like, if someone else had time, I want to do this. I'm happy to let someone else. Remember this. uh If not I'll do my best to make progress on this, and I think the first thing is those two to the cap.

C

They need some updates to to allow me a new format for caps I'll get to that, but the issues discussed there are the same. We need to figure out how to deal with some error cases and how to deal with especially pulsar crash looping, which can lead to some crazy situations.

C

B

Have you actually looked at the timeline? uh Where is it going away? I'm just trying to figure out how much time we.

D

Have so pdbs are on the same timeline as cron job, which basically means the deprecation will start in 122, which basically means we don't have that much time? um Martin, I have a one question. I actually I have two questions and I left them in the doc.

D

First of all, uh I was talking with boydex some time ago about pdbs and um the most important question is: do we need to um modify the api of the pdb to be uh to be able to ga or it's only functional changes in the controller code that needs to happen before vga.

C

So I think there might be some small changes to the api. That's mostly around how to handle particularly pdb's now doesn't expose any conditions, so it can be hard for users to detect when when something goes wrong and the controller can't for some reason can compute a lot of disruptions and that will block all connections. It's not easy always to to see when that happened.

C

um There's also an open issue that was opened last week around how the discussion controller handles empty selectors it won't. It won't directly require us to change the api per se, but we need to probably change how a controller interprets um an empty selector, which is a breaking change. So we need to figure out how we can update that in v1, because it's probably not acceptable to change it in v1 variable indirectly,.

D

Okay, so follow up a follow-up question.

D

I doubt we will be able to push the cap for 120, because the future freeze is tomorrow, so it's literally impossible, but I would still try to put together and we can sink either.

D

I can refresh the cap or you can do it because I would really like to see the uh get the cap being updated within the 120 release range such that it is ready for 121 to be merged as soon as the gates open for 121, and we can cross check at least one bit and at the same time, hopefully within 120.

D

We should be able to modify or fix the currently open issues, at least the ones that do not require api changes.

D

What do you think about it?.

C

I agree, I don't think there's any. We need to do anything for a deadline tomorrow. The important thing is to make sure that we get this done before it gets duplicated and, like you said, there's not that much time left. So, if that's gonna happen, we need to get things moving now.

D

Exactly and if you're, if you're saying that, there's one I can't remember which one was it there's one particular uh fix that will require api change and that will be a breaking change. We can probably address that when gaining the api- hopefully, let's say in 121 and then we will just um the v1 beta 1 will be broken or we will provide an opt-in, whereas in in in the v1 we will have the selector fixed properly.

B

D

We actually gad.

B

Api, or do we have to do another beta.

D

If we do it before 122, we can ga the api, if we don't make it before 122. uh The api goes into deprecated and we need to have a second beta to be able to progress to ga. So, if possible I would prefer to.

D

I would prefer to save people from too much of a hassle, switching from one beta to two to second beta and then finally to ga, but rather um uh squeeze that in within 121 and that's theoretically, that's doable.

D

If not, we can it's going through.

C

Going via v1 beta2 just means more work for everyone.

A

Okay, that sounds great, um we'll try and keep an eye in sig apps in the meetings and with everything on uh on dates, especially with regard to 121. So we can try and stay a little bit better on top of this one yeah definitely yeah.

A

All right, uh then, we have a.

A

I haven't looked at this one, yet a damon set mac surge cap.

A

I haven't looked at this one in a while. This is from march of this year.

D

So clayton put together this cap um quite some time ago. We we talked through this several times during sick apps in the past.

D

um I think, as it stands today and both thomas and I had a look at it on friday.

D

Overall, it addresses all the concerns that people brought up brought up. The only missing bit was the prr section which is required in the new cap format. um So current, I I haven't looked at the pr, uh but I have uh faith in clayton, so that should be pretty uh pretty ready. So I'll have a look at it after the call, and we probably need um either you janet or can put a final approver before tomorrow,.

A

Okay, I'll take the action to chase this one down and.

A

And see if we can approve it by tomorrow,.

D

Yeah, that would be cool thanks very much matt.

A

All right is there anything.

D

Else I can share that the crown job cap got merged and the ally who's on the call we are working together on the new controller implementation.

D

If luck allows, we should be able to start merging the initial implementation this or next.

A

A

All right, thank you for that update.

A

Is there anything.

A

Else all right and we'll give everybody 42 minutes of their time back. Thank you all for coming today have a wonderful day and we'll see you back here in two weeks thanks. Thank you.