Kubernetes Cloud Provider Special Interest Group, 19 Jan 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SIG Cloud Provider 2022-01-19

Description

New Chair, 1.24 KEPs, Kubecon EU

A

This meeting is being recorded.

B

All right everybody welcome to sig cloud provider. It is january, 19 2022., um I'm the moderator, nick turner. uh Please remember to respect the code of conduct and be respectful to others. Let's go ahead and get started.

B

So we can go through some project updates. I will, I guess I'll just briefly go through each cloud provider. So stop me if you have an update or I'll stop. If I see something written, alibaba, baidu, aws uh aws, we have uh we just merged 123 alpha release.

B

Release- um let's see, I think, that's all we have kishore did you have anything no.

C

I don't that's it. Thank you. Azure.

A

Yep just pretty much the same stuff, we got 123 and we're doing patches. The usual gcp.

D

So one interesting note on the gcp um and this may affect everyone other than azure. uh There was a recent uh pr from red hat. Thank you red hat, to upgrade the cloud provider gcp repo to be, I think, 123 compliant uh and I was reviewing it and I noticed that it was pulling in an azure library which I we still don't know where it came from. But I noticed it's actually in the kubernetes kernel.

D

uh So I there is some interesting something is pulling in one of the azure core libraries that doesn't look like it's actually, the micro, the the microsoft uh cloud provider, azure implementation. So we're we're going to be following up on that, um but yeah someone, either red hat or someone from google will try and work out. Why that azure library is being pulled into the main portion of kubernetes.

A

Great, thank you. Can you put a link either in the notes for this meeting or in the chat so that we can see if there's something we did, that we can undo um sure give me.

D

E

F

D

Yeah no no worries I'm trying to find the.

G

I send that to the chat. Please take a look.

F

Yeah look, I will put it in chat. Oh all, right.

D

Yeah awesome yeah, so we just need to follow up and work out why that library is being pulled in uh no offense to any of the azure folks, but it just feels a little weird having the cloud provider gcp uh implementation with the micros.

A

And I'm I'm looking at the recent comments on that um issue. It looks like a few hours ago. Somebody pointed out. This is like a go ansi term thing that just happens to be under azure's github repo, but we should definitely figure out why um an implicit dependency that you know shouldn't be cloud provider specific is listed like that.

B

Yeah, we see the same thing on ibm.

D

Yeah, like I said I did it, I I was assuming this wasn't gcp specific, but it wasn't also. It wasn't like. If I look for the ansi term inside of kubernetes and kubernetes vendor, I wasn't actually seeing what pulled it in so anyway. Yeah.

F

Cool all right for what it's worth like that, those users for docent and joel speed. Those are guys on my team at red, hat and yeah. This provided us with an interesting conversation today, so we were kind of stumped by this as well.

D

Yeah I mean like, I don't think they did anything wrong, so I I did actually mean it when I said thank you for for, for, for the the pr it's just right, that's where we we! I I wanted that answered before I approved.

D

A

Go ahead, oh I was saying I'm going to ask the folks whose usernames I recognize on that go ansi term repo. If there's a reason that we can't move that to something less um azure specific, because it looks like it's just a cross-platform anti-terminal emulation um for a library for windows users, but it shouldn't be azure specific.

A

So I will take as an action item to ask hey. Why is this hosted under the azure github repo.

B

Cool all right rest of the updates, um huawei ibm openstack vsphere.

B

Extraction migration.

B

All right, moving on to the agenda um walter, you have the first item: go ahead.

D

Well, you say I have the first item, but uh apparently, nick and andrew have have uh managed to to to perform a successful coup and I am being ousted as co-chair. Now um I'm teasing um so one of the rules for the sig.

D

I I think it actually may be more than just our sig, but it definitely is for our sig for some fairly good reasons is we do not want to over represent any one cloud provider uh in the chairs for the sig and, as a few of you may be aware, uh my co-chair andrew has recently joined google, so this does mean that at this point we have two co-chairs from google, which is actually against our charter.

D

So I will be stepping down as co-chair and going to being just uh one of the tl's for the sig. uh So only nick and andrew will be your co-chairs.

A

B

A

That sounds like less of a coup and more of a oh hey, look at that new friend who now gets to do the work.

H

Quit revealing my secrets: bridgette, that's exactly what it is. I was duped.

B

But also all your hard work, one thing that we're one thing that we're also doing is kind of more defining the the tl role right. Walter.

D

Yes, uh we will definitely be doing some and in fact that is happening across.

D

Across the kubernetes organization in general is uh most of the sigs are being asked to better uh better define their their tl role.

B

Got it sorry, I might have to mute in between um updates here, as my dog decides somebody's trying to break into my house um so great, so andrew you're next on the agenda with 124 cups? um Do you want to take this one.

C

Yeah, can you can you guys shoot me yeah? Okay, sorry, I'm having audio issues using the zoom web client, um so I'm dialing in for my phone as well um yeah. So in the uh extraction migration sub project, we we spoke a bit about like what caps we want to work on, and so I figured given the kept deadline is coming up next month early next month.

C

I think we should just um have that conversation again um with the updated with the list of caps. um Let me actually try to find so I think there were four cups we talked about. um Let me let me try to find this actually one sec.

C

uh So there was four caps, I'm just gonna copy paste. The agenda here.

C

It was these, and this is the priority that we had agreed on, then the first one being like um having the admission web hook integration so that we can do pv labeling um and like making progress in the last known good stuff for testing and getting the kubelet image credential provider to beta, which involves adding some ed testing in there and then the last one being the controller manager, leader migration and the last one.

C

That was last mainly because the work is done and it's just a matter of like hardening it and getting it prepared for ga. So we don't think there's too much work involved there, but those are the four that we had talked about was: was there any other caps or to or does anyone like disagree with this priority order?.

B

I agree with the priority order. One thing I think we should do is have uh we should nominate people to not necessarily do the work, but just be responsible for finding people to do to do the work.

B

So I know like me and walter are going to meet today about the the first one. um Should we should we try to find uh folks in this meeting that that are willing to at least you know, I don't know, try to find people to. You know in fact like some of these, I don't even know how much work is actually necessary, so uh cubelet image, credential provider to beta, says we need two tested, implementations, um aws and google. So maybe it's just tracking down.

B

You know the people from aws and google, who can provide the guarantee or the data that this is tested or something like that. So um what do you think about that?.

C

Yeah that makes sense so yeah. So maybe maybe just like gather a list of volunteers who want to take up any of them, and then we can have another meeting and just figure out who can welcome what yeah and I'll.

B

I'll part I'll start with.

B

Putting mine on wall stores on the first one.

C

B

C

Sorry bridgette, I totally missed your your first note in the agenda there um on prr. That's a good point. um I I find that like pr is like the thing that is most review constraint and the thing that tends to block kept, so we should make sure the pr questionnaires are filled at a time and looking good.

A

And I don't want to, I don't want to make it sound like it's a bigger barrier than it is it's basically just like a couple of pull requests, but you need to make sure you have them in place or the cap is not going to move from stage to stage and it didn't used to be that way, and so, if you're thinking back to a few releases back where it was much easier to just be like this is going to beta.

A

Oh at this point, if you want to move something from alpha to beta, there is work to do, and you got to pay attention to that. So I'll drop a link in the chat um as to what that looks like.

B

Okay, so is that note specific to this cup.

A

Now, literally, moving any cup to any stage now has a slightly different process than it used to have. So, if you're wanting to work on that, just keep in mind that there is something you should check right.

G

B

Cool, so um so does this? Does the uh mixed protocol, support of mixed protocols and services need um an owner as well? Are we going to include that in our.

A

I think that one's technically under stick network, but um yeah, but since it applies to everyone from all the clouds, that's why I bring it up here.

B

I guess yeah we're we're listed, but.

A

Yeah and yes, I am, if you scroll down you'll, see that I have a checklist of items I need to do, including the production readiness review.

A

As you can see,.

B

A

This is a maybe overkill, but you want to get your caps through. You gotta just make sure you've got all your items kept in mind.

B

Cool okay, um so last known good, do we want? uh Does anybody want to to sort of drive that.

C

I think that's on me, or at least like I'm, willing to work on that bit, um but I think this one in particular could use like extra hands, um because I I'd love to like arrive at a um a framework for how we test across multiple providers and repos um and make sure that, like we have a good cross cross repo testing um so like at least personally for me, like I'm, going to be looking at it. A lot from the cloud provider gcp repo perspective, but it'd be good to get other folks.

C

Also thinking about the the testing matrix and whatnot and make sure this is, it looks good.

D

uh I I'm willing to sign up. I, unfortunately, I had someone working on it, but uh they they've moved to a different team. uh I will sign up for at least getting uh a proud job running in the main kernel that will expose uh which tests uh need work so I'll sign up, which should give us sort of a baseline of all the work that needs to be done.

D

So I can sign on for that case.

B

Okay, um cubelet image credential provider to beta.

C

I guess I could take this one too, so.

D

For context, I was going to say I just as a side note, uh but I would follow up with kermit on this one.

C

Yeah sounds good. I could do that.

B

All right and then later, migration.

D

uh So I think jared is sort of owning this one right now, but maybe jared and andrew.

G

Yeah and jared so for this one we only need the testing working. This shouldn't be too much too much features to go out. I think right now.

D

The big thing we need to do is define the ga criteria right all right, so I think right now, I wouldn't worry about anything other than coming up with a good ga criteria that we can all approve of.

G

So last time we talked about that saying, like all providers need to be need to be like this feature should be working for all providers, but you know it's pretty expensive to to develop the whole eqe test on one provider, not to mention all providers and uh competition timeline will be different. Let's.

D

Go ahead agreed so so I I just to be clear and I'm not the only one who gets a say in this, but my perspective is that for us to say it is general availability and we believe it it 100 works. We need automated tests on a system.

D

We need confidence that those tests are in enough depth that that we believe they work and we need um you know uh testimonial, I will say testimonials. We can work out how we want to surface that, but we need all of the providers who plan on using it to basically have tried it and said yeah this works for us. I don't necessarily think we need automation on any on every cloud provider, because certainly we have other features that are ga that aren't tested on every provider.

D

But I think that, based on what the feature is, we need confidence that it works everywhere and then we need a mechanism which can be one cloud provider where we know it gets tested on every release so that we have confidence that it doesn't get broken.

D

And I think that's this and for those who are familiar with the with the concept.

D

I think there are problems there may be other things that we should add as well, but I mean the ga criteria are not supposed to be an easy bar to me.

G

C

C

Did you want to take the action item of like making the j criteria's the kept clear, or did we already do that.

G

Oh yeah, I think I think I'm already on that. Okay.

C

B

Awesome um anything else on the cups that we have to get done.

B

Cool all right, uh so then, in that case we can move on to the next item. uh Bridget, I believe that's you.

A

Oh yeah, so some of you, and by some of you I mean walter and andrew, might remember that we were kind of on the azure side. At least we were chasing you looking for reviews right before the patch release deadline, and I thought gosh, perhaps in his exciting new role. Nick wants to also do some of those reviews, or perhaps some more people from this egg. Mostly, I just want a wider pool of people who can do the cloud provider related reviews so that we aren't always bothering the same people at the last second.

B

Yeah, I'm absolutely uh willing to take on more reviews. um I know that there were restrictions on like who can be a reviewer for the entry uh legacy providers.

A

Basically, I just went I'm fine, that's why I put it on the agenda. I'm like we should figure out how to remove any barriers to add, maybe nick and a couple of other people, eventually yeah, don't even really care who I just don't want to constantly bother poor andrew.

C

I think I think the context here is like we, um we shrink to the number of approvers, um because we we wanted to slow down like feature development and have people focus out of tree, but I think the result of that was like walter and I being the only approvers under the legacy club fighters directory, um and I think the problem is like walter and I actually are doing um full reviews of that. But maybe like the right thing is like we should just be like stamping the pr saying.

C

Okay, like I'm, acknowledging that this is a bug, fix and not a feature and then defer to the actual actual owners of that package to do the review. um So I don't know like I feel, like I, I think, maybe I'm causing myself more pain by like doing full review when I don't have to. But I don't know if we have the right, like owner semantics to actually to do that, because there's only like one. One of you only need one approver right for a whole pr.

B

So then, is it correct to say we need more reviewers and possibly not more perverse.

C

I'm not sure, to be honest, I think in general, like more approvers and reviewers. um No, I think I think bridget is right, like we could probably use more approvers. I think we have a lot of reviewers already they're, just not officially approvers now because of the bug fix change that we made um or like. Maybe the right answer is like.

C

Maybe we put the we put the owners back like put the original approvers back in there, but like we really want to make sure and have guardrails that we're not letting like uh features in so like. Maybe we can have a bot integration that just blocks the features instead of like having people review the pr's and make sure they're not features um yeah, I'm not sure. To be honest, I'm just rambling at this point.

B

Yeah makes sense um uh somebody, I guess me, you and walter should probably take an action item to figure out what we can do, whether it's at a approver or you know, change how we do approvals.

A

B

All right, um steve, you're up next.

I

I just wanted to call attention to the fact that the kubecon europe cfps are open for maintainer track. If this sig wants to do one again like we did for north america, the cncf is still saying that this conference is going to be physical only so it would require somebody to be there to deliver it, and I'm not sure whether you know there's the group perceives there's a need for one of these.

I

Whether we'll have news beyond what we did at the last one or not, but just wanted to bring it out there, because the deadline for submitting is about a couple weeks away.

A

Wait I I am confused, because I definitely saw something at some point that said that they would have um in-person and uh virtual options. Are you saying that you found something to indicate that it's normal.

I

I've heard somebody say recently that they looked at the website and said it would be a hybrid. So I don't know I was I'm on one of the program committees and in that role I was told it was going to be physical only, but perhaps they changed it without sending out a word to the program committee.

I

Iman, so uh don't take me as authoritative if you find something that says something different. It could be that this would be eligible, and I don't know that whatever the rule is means that the presentations are also virtual or whether only attendance is virtual. So we'll have to see even in north america you they could not do a hybrid. You either had to go 100 percent.

I

A

I

A

Yes, it's a hybrid conference, but yes, you either submitted a pre-recorded video or you came and talked in person, but they didn't have the bandwidth to have you and someone on a stream giving the presentation. If you were in the room and they were not. um I think this is set up the exact same way as that. So, yes, you are correct. We need to commit to either doing a pre-record or sending a person.

I

Yeah or I guess a third option is not even doing one, but you know, and I'm not even aware if anybody active in this group is based in europe or not because somebody is maybe it's a good opportunity to you know, pass the baton over.

I

Just for the record in case anybody didn't know, I'm not I'm based in the u.s, but maybe somebody here is in.

I

Europe, so I guess next steps are maybe to just do an informal poll on whether people think that the group wants to do one. If we do, um we want to have some kind of a proposed subject for the presentation.

I

um I am willing to work on helping prepare a cfp and a deck, but I don't feel that I've got. I don't have a topic to nominate, but I'm willing to help like I did at the last one. If that's of interest, I'm also willing to step aside and let somebody else do the whole thing if they prefer, but just putting it on the table, because we by the next meeting in this group it's likely too late. So I think we have to make.

A

A decision, though this is for the sig update right, because the cfp is already closed, while.

I

The regular cfp is closed right and I think people have been informed by now whether they've been accepted. That might not be true, but I think so march.

A

I

It won't find out until.

A

March, 7th reviews are going on right now.

I

All right, but the maintainer tracks work on a different delayed schedule, so maintainer tracks are open. Now the sig chairs should have gotten an email uh from the cncf. I think only the sig chairs get those, not everybody else in the group. um I got one by virtue of uh being in leadership on a different uh group, but I'm assuming that that the chairs in the mailing list for this group got one. Maybe I'm wrong.

B

It does sound uh familiar.

I

If you give me a minute.

B

I can look.

I

Up what the subject matter the subject heading is: if you want to search for it,.

B

I yeah my opinion on this is just that like whether or not we want to um it's probably good to do some kind of an update uh just in terms of what we're doing like I, I doubt very many people know what last known good is and and what the proposal is there, um so that plus the status on the various caps that we're trying to push through, um I mean it's useful to some people. I think uh yeah.

I

So well, I think we got pretty decent attendance, considering the overall attendance of all of kubecon for north america, so it it it is likely it would draw some audience.

I

But I also have the opinion that if, if the news is exactly the same as the last time, there's no point doing a repeat that we would want some new material. But if we're likely to have that then yeah. My vote would be to do another one. Assuming that we've got uh one or more presenters.

F

I I think, there's still value in continuing to broadcast the messaging around. You know the deprecation of the entry and moving toward a tree like even if some of our materials, the same from the last time.

F

Like I mean I kind of agree with you steve, like you know, we don't want to present the same stuff, but I think there is value in continuing to have a presence and just like putting this in front of people's face that this chain, you know the old stuff, is gonna, be deprecated and we're gonna move there, because I think I mean I'm, I'm not convinced that the entire community has gotten the message yet.

B

D

Agree with that, I will also point out. We have and I think, had reasonable traction doing things like uh in-depth on particular pieces of the solution. So I know that in china we gave an in-depth on the ssh tunnel replacement api server network proxy.

D

uh So there may also there's also an opportunity for us to help people better understand things like how the cloud controller manager works uh and what it does or how the credential provider does what it does and how it works and how to configure it or even uh what the mechanism behind the leader migration, what it does here and how it could be reused for other contr, because it is more than just a controller manager, migration.

D

It's actually a specifically a controller within a controller manager migrator, and so we could not only discuss how it works for controller manager, but how what has been implemented could be reused uh for other controller from to migrate other controllers, from one controller manager to another.

I

Okay, well, it.

D

I

We've got plenty of subject material, then an interest in doing one, so we should probably put submit a response to this. I pasted the email header there if some one of the chairs wants to see if they actually got an email like that, and that has a link of where you go to submit it. Like I say: if no one else wants to do it, I'm willing to work on the cfp with under the direction.

I

If, if things like what you just proposed walter, if you can lay that out or put that down in the notes, I can put together the cfp as a shared document like we did last time and we should be able to hammer one out the other thing that we could do as some of the material.

I

I personally think it worked pretty well last time, but having those little lightning talk things from individual cloud providers, I think maybe we should have done an audience poll to find out, but I thought they went pretty well at the last one. As you know, a way to get news out related to cloud provider.

A

Yeah, I know we were one of the cloud providers that submitted a slide, but did not have the bandwidth to prepare a short video ahead of time. So I would say, if you do want to do that, that sounds great. Maybe give us a specific deadline and length of time. You want the video to be be like hey. We need the video by this time and yeah.

I

I think it was a good technique too, given covid and people not um available to attend physically to actually have a little bit of the equivalent of what you might get by just having people in the room or in the hallway track. You know giving them a forum to get word out on whatever. So um I don't know my personal opinion is that doing that again might be a good idea. If we're gonna do one.

F

Just for what it's worth too, a colleague and I submitted a talk to the next kubecon about doing the migration from entry to auditory and kind of like what we've been going through a hat too so I mean you know. We talked about this, I think before the new year or whatnot, but it's not related to the sig but topically. It's kind of related, okay,.

I

Maybe what we can do is not duplicate that content, but if the sig's talk goes first, we can give a referral to the audience to go catch. The other one too.

F

That'd be really cool. I I think, given the topics we've mentioned here today, I don't I don't think they'll be overlapped, but um yeah shout outs are always appreciated.

B

So steve, what is that? Did you have the deadline.

I

B

It's in that header that.

I

I put in the chat so it's february the 14th, so maybe we will have another meeting before then um it's a little. It's out there a little further than I thought not quite a month away but close to it. Okay,.

B

Yeah so we'll have another meeting. So why don't we start thinking about topics, uh especially like the? If we want to do any deep dives on a specific specific topic and then um next meeting we can iron out the details on who would speak and whether it's in person or or not,.

I

Okay, how about if I'll this is what I did last time, I'll open a shared google doc with just a template of some of the things we know are in there like if we're going to do lightning talks for individual cloud providers, and I think we always close with how can you get involved in this group with our meeting times and things and that's kind of boilerplate and an intro in terms of hey?

I

What is say cloud provider um and I'll share the doc and put the link to that shared doc in the slack channel and people who have other ideas? Let's start this out, where anybody who wants to suggest something in there throw it down in that shared dock. If we have to do triage and prioritize it, because it's too big we'll deal with that later, but does that sound like a reasonable plan?

I

Yeah sounds great to me and then maybe we can also have people indicate their ability to uh present or attend uh in terms of how we want to do it. Whether it's you know. Last time we had three of us who were there physically and plus playing back the recordings and that worked pretty well, but if we had to go 100 online we could probably do that too. If that's an option tentatively, I am planning on being there, but I don't know if anyone else is and also I don't know does anyone know.

I

Is anyone in this group based in europe.

F

El migo, I think, you're a no.

F

I was just gonna say this meeting is extremely late for our european colleagues.

I

So maybe not okay.

F

Not because of the.

B

F

Time maybe they would be if it was earlier. I I know for a fact: there are some red hatters who would attend if the meeting, if there were a better, you know, meeting time for them but like for them. This is. It starts at 10 p.m. So it's just not feasible. I I I actually.

D

Since we brought the subject up, I think there are a few googlers who would like to be a part of this, who would also be much happier if we could move it to an earlier time so that it was more europe-friendly.

D

So, um as someone who isn't a chair, I I am going to petition our chairs to to maybe consider uh doing something like either moving it earlier or making every other one of these meetings earlier.

C

I I'm cool with earlier as well, but I guess like how early is okay for the west coast. People because, like I'm cool like 10 11 my time but like that's, seven, eight your time. What not.

I

G

Got a suggestion having.

I

Done meeting changes with other sigs and working groups is that the probably the better way to do it is to make this a poll that you put on slack because asking the people here is just not a correct sample.

I

You might get more representation, because obviously the people who don't like this time aren't here to speak for themselves. So maybe a poll with a number of selections that people would vote for either yes, no or prioritized, would be the way to go here.

C

Okay sounds good and I I think just for whoever puts the poll, I think maybe some folks in asia might also be interested. So maybe we should put some time slots in there that are both asia and europe friendly.

C

um But, okay, I guess nick or I will put the pull together yeah the.

I

One assumption you might have to make to make the pole reasonable in sizes decide whether you want to have split times. You know like one meeting at one time, one at the other. That seems maybe a little more inclusive, so nobody gets permanently shut out. But just one person's opinion.

B

All right, um just taking spouts um so okay, so next meeting we'll we'll iron out the details for the talk and person versus virtual, what subjects we want to cover uh and who's talking and then one of us will create a poll to see what meeting times people like, and we can do that in slack. We should probably also send it out on the mailing list as well.

B

Okay um looks like we have 15 minutes left. Do we want to do some issue triage.

B

Yeah, let's do it is that a resounding yes.

C

B

C

Shift on the topic of like meeting times and whatnot, I do wonder if maybe we want a dedicated meeting for issue triage also or like just something to consider. We don't have to decide now, but just something that came to mind to talk about it.

B

Yeah, I think that would be a good idea, um because we never know how much time we're going to have for it um yeah. I like that idea.

A

Network so does sig network does issue trash right at the beginning for a time boxed amount, which is one thing to consider.

D

As I say, the the other is api machinery and they have a much higher load of issues than we do, but they actually do issue triaging twice the week in two dedicated meetings.

B

Got it um yeah? I don't know if we have enough issues to fill up two meetings worth.

B

B

Do we want to look at kk.

A

Sorry, what yeah? That's better thanks! Okay,.

C

There's a filter link on in the agenda, I think got it.

B

And this agenda.

C

My bad, I forgot to copy it the template. um Let me copy it real, quick.

C

Okay, I put it at the bottom.

G

You see it there.

B

You put it at the bottom, the template or the bottom of the.

C

The bottom of the today's agenda today.

B

Yeah right there got it. Okay, uh I guess, needs triage.

B

All right before detached device should ensure there is no pod use. It.

B

This is in kubernetes package volume, so this is cubelet volume stuff, I'm guessing.

C

C

I think this is I've known. This is a known issue. That's been lingering for a while. I think that has come up in multiple issues, but basically there's an issue where like, if you have a pod that was attached to volume and then the node, the underlying node goes down or whatnot.

C

um Your controllers can end up trying to attach that volume to different pod that was scheduled on another node, even though the volume is actually attached to the old node that went down and that causes like um data loss or like race conditions, where two things try to write to the same disk.

C

At least I've seen this a lot in other yeah. I've seen this come up quite a few times.

B

um So do we want to accept this.

C

Yeah, I I'd say, accept it. I think it needs a cup to be honest or like it needs more thought on like how to do it right because, like you, can definitely do it wrong and then cause more issues.

C

B

um I think this needs I mean unless anyone happens to know so they're saying like there should be a check for device path to ensure the path exists. But I don't see any explanation of like what the failure mode is.

B

And looks like.

D

Yeah, I think, there's something. The other thing is this.

D

We might want to work, I mean this. One is assigned to both us and sig storage, and my feeling is. This is probably more of a sig storage than a cloud provider.

F

I recognize this username jan safrane is one of our storage, guys so yeah. It could very well be a sig storage thing.

B

Yeah, should we defer to them to accept it or not.

D

I I think we might want to try removing ourselves. I mean part of this is going to be that the device path, I think, has to be done on the node and right now we have nothing that runs on the node we well, it's not true. We have the credential provider, but the only thing we have that runs on the node from this sig is the credential provider, and that is not it draw.

D

So this seems like something that would be better done by the csi volume plug-in or possibly by node itself, but neither of those have anything directly to do with us.

B

Okay, let's see what uh sig storage thinks about it.

B

All right walter looks like a uh storage, a google related storage.

D

Don't know whose cylinder is, can you mouse over cylinder or cyclinder in the assignees yep.

D

Okay, does that.

C

D

Well, a little, uh can you try michelle's gonna hate me for this? Can you try assigning this to mso.

C

Now now it just looks like it just looks like nick signed it to her.

B

Without asking yeah well done walter, you better take uh take the blame for that. One.

B

All right, so it looks like all of our issues were, uh came up from the from one person going through and trying to do something um use zone from node for topology, where volume creation.

C

Yeah I'd say this is storage too because, like it sounds like the topology, information is already on the node. So, like the cloud provider part did its thing, but then the storage part has to pass along.

B

Yeah, so it's already.

B

B

Is that gonna work.

B

Remove sig is that a command.

B

D

B

D

B

B

Yeah, okay, um next one.

C

Seems like this one already has people assigned to it from the sick network.

B

Yeah, this might be another one. We can remove ourselves from.

D

It seems like they just need to use a finalizer.

B

Yeah, I don't know.

D

Would there be? I don't think this is cloud. I don't think this has to do with the cloud provider. This is between net working and api machinery, but my guess is: they should just be using a finalizer.

B

Yeah walter did, you did a sign t-shirt.

D

um I think at one point it was it was being looked. There was some indication that it was uh amazon only, but I'm not actually seeing where we got, that uh it was the provider aws. That's where okay.

B

Yeah, okay, I'll leave.

B

Kishore do.

F

You want to take a closer.

B

Look at this or excuse me.

D

I don't think he shows in the call.

C

I think he was earlier, but um it might be worthwhile to actually leave the cloud provider um thing in there because it it does seem like the implement. The implementing ingress controller could do something so, like I I know, like the endpoint slice api, we recently added a a terminating condition.

C

So there could be a case here where you can update the ingress controller or the aws load, balancer controller, to watch endpoint slice instead and then like delay, the removal of an endpoint, while it's terminating as opposed to just removing it right away when you delete it like that. That should be a issue in like the aws controller, but maybe we reference this.

B

B

B

T-Shirts assigned.

C

I'll just paint him again, I think we can just like triage, except this one like it's an actual issue um and like the right people are assisted.

C

B

D

He grabbed me, just as I was saying, bye uh loops forever in case of metadata cannot be reached. uh The right people are assigned to this one. So I would say we just triage accept, okay,.

B

Okay, that is everything, and we are out of time.

B

I'm gonna stop the recording thanks everybody for attending.