Kubernetes SIG Cluster Lifecycle, 25 Mar 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: kubeadm office hours 2020-03-25

Description

https://docs.google.com/document/d/1ONcoy8bOw8SWPUwXxnKeRZST3lnUxYpUv4Y6466h9Ek/edit#heading=h.4qer8rrq8h05

A

Hello, everybody- this is the sequence lifecycle chameleon office hours today is the 25th of March 2020 I'm, going to share the general approach at.

A

Please, underneath the list of participants and your agenda topics, let me share my screen.

A

Alright, let's see what we have so the 1:18 release. Sorry, let's start with new participants new having been new I told anybody, hey Pierre, says: 180 was delayed apparently from yesterday sorry.

A

180 was delayed from yesterday because it was some sort of issue related to the Cooper initiatives, repository and I think it's already fixed and the release is hopefully coming today. Maybe we are good documentation. Prc everything is merged, so yeah good to go going to the general topics. I wanted to propose whether we should have the planning session for 119 next week, I already spoke with four Blitzer. He said that today he is not going to be able to join, but maybe next week we can have this planning for the cycle. Any comments.

B

Postponed for next week.

A

Okay, Rafa was also plus one Tim. Would you be able to join next week.

C

Next week, I should be able to yes.

A

Okay, so I guess we have enough people to conduct the planning session. We can prioritize some of our items for the cycle and see how it goes.

A

Alright, so I added this, because we, the next item by LEDs for also referred to discussion but I, also participates in the same discussion. So basically, we have too many comments in github about this PR that Rafa said and I wanted to finally discuss this over on a meeting and to give some context about the PR.

A

Basically enough assent appear to modify the way we do upgrades in terms of honoring the HCD tag version that is stored in the user config file on disk and is related to the config on upgrades, but this whole the whole discussion about like how should we manage density tag and I I locked some of the items that we might have to fix here and also yesterday in the sequence lifecycle meeting I spoke with Justin and with Daniel from HCD idea, and we somehow arrived to a conclusion that cops is already doing something like eBay.

A

Dm is also doing, but we are not doing it very well. So this is how we manage tags for GD, so cops. The way cops is doing it they. Basically, if the user applies a custom tag to a CD, they don't perform upgrades at all the quickly. This means that the user has to if they want to perform an upgrade. The user has to go into the coastal configuration modified their tag and a manually deploy a new HDD. So this is basically one of the items that I want to discuss here.

A

So basically, a lien sub code is here also. The first item is we also have to resolve this. We couldn't get in a warning because imagine that the GCR image has a dash zero revision, correctly cube ATM desert, condo potential revision upgrades to war. The fix further is fairly cheap. I think we just have to tolerate revisions, but yeah just open for discussion. Let's, let's go through the items in order, because I cannot mess this up in the introduction. So what do you think about the GCR revision problem.

B

So I'm thinking here that probably because we are actually using an image in a pet city- probably an image revision is always good to have and, as you can see, GC our images already have one and probably I'd show distributors such as society more or whatever are going to attach their own vision numbers to the image. So I think that we should tolerate those and at least try to cope with them. I.

D

Agree with that, I I think the the revision number is important, so you, you really know what kind of putts not only the butts level, but what kind of pad size do you have on top of that putts level? Maybe some cv is fixed things like that, or maybe some custom putts so I think that's that's useful.

D

The blender I see with our logic, is that maybe we are getting too much into a pill so, and would you intend Lumiere, but is that we are comparing something like this with the full contest with the metadata and so on, with what it city provides us, a runtime that date did see the front end doesn't have this information either know if this is a defect when building it city. So if it's city also reported this as well under runtime information, there will be.

D

We will be comparing the same kind of tax, but since this is a stripped on the it CD runtime at the moment, then we are comparing different things at the moment.

B

Yeah- and my proposal here is to basically just always compare only tax which is easily implementable in the stock density case, since we can actually play with the static pods for the images that they are using and extract the tax from there. But it's not that easy to do in the external 8 city case, but the accelerator. The case is only used like when we actually do an upgrade plan and we in particular use a sort of like side effect of it.

B

Basically, the code there actually goes and checks all the eight City nodes and will error out. If there is version mismatch between bb-8 see no questions, so I think that we should probably stop outputting, at least that information about its external HDD doing a great pal. So what is that's my take.

D

B

D

Agree with that also on one question Lumiere before you said that you were talking with the folks from from the sick about how cops, for example, is handling this one. One thought regarding that, if you I mean I, understand what they are saying that if you only change it CD, for example, they will ignore that, so it's completely up to the user to change the configuration in a way in which you can then apply that. But my main wonder here is how do you handle an upgrade of Cuban Eddie's along with that?

D

So in our case, for example, we want to bomb Cuban that is as a whole, and this might include it see- maybe not, but it usually included city as well. So we want to bump everything in this configuration file. We break the whole configuration file for Covidien. We write it back for for the classic for the Cuban version and also the 8 CD 1. If this was to be ignored, then we would bump everything even Eddie's, but not each CD.

D

We would have to pass that on our own at one moment before track before trying to do the upgrade after it's kind of tricky. You know.

A

It's especially tricky if we want the bump to HD version 4 and they are incompatibilities between the API server that we just pumped an HDD. So the idea is to pop the control plane first and then pump in CD. If you are breaking down the process in that way, but imagine that they, the new API server, so uh so the the API server should be backwards compatible with the old HDD in this case.

A

So imagine that you will release 120 API server, you, but you bug to this version, and now it supports 8 cd40, which is like, maybe completely different, but the way cooperate is going to do that they are going to handle the backwards compatibility with the version 3 of 8 CD, as well so you're going to have a loaded, C version, and you have to also update HDD like a separate step to be for 0 at this point, but I I see the problem here is that we have to document this manual step because users are not going to know how to do it.

A

Apparently cops is completely delegated these to the users incubate them. If we do that, maybe we are going to get some complaints. People asking hey I have a custom tag. How do I, how do I upgrade a CD like when is the right time and things like that.

D

I mean from from my point of view, if you know we have this whole discussion, that is aside of these of this config flag, that we should probably remove at some point so, but assuming that we have the config flag, I would just keep the same logic in the sense that I will just start with the user does right and if we find reports of people you know using it CD versions that don't match our expected person. Then this is an unsupported setup right. So we cannot really support that you.

D

You went down the wrong way, let's say, but if you are using this in a way, for example, as we are using a different schema of tagging, but we are effectively using the same. It CD real version. The qadi more kubernetes expects. So there is no issue in that in that sense, because it's the same in CD version, but it's it has a different tagging, schema, let's say yeah. This is.

A

Where my proposal came around around basically, this is the logic I pretty much dubbed. Is this like a row out of if branch istic conditions, how can we manage upgrades for users if the version is a new we're just going to assume that the the same image? As you know, the metadata for the vendor exists in the image repository?

A

So how can I give a good example here so much of this so.

C

There's ways to do this by breaking up the well-defined version string into sub components and allowing you know the if you had templating, if you don't really have so that you can allow the variable override, have an entire fully qualified stream naming for every single parameter. That makes up the entire name of an image. Artifact it'd be possible to do that. We don't currently do that.

A

Yeah basically I proposed, like standardizing the tags we do for comprehensive, a and I said this to Justin and Daniel like maybe, if we standardize this, maybe we can. You know template the vendor template the revision, but they said that the cops is already doing something else, so we are kind of drifting into not standardizing things.

A

Cops is doing this complete delegation of you know. If the tag is custom custom, we don't do anything with it. So I was wondering whether we should follow the cops example or diverge. My preference is going to be to follow what cops is doing. You already have something like that: incubator, I'm, essentially.

A

Ignoring ignoring tax, if the.

D

Artist I'm missing something here, because what is the point of providing an image stuck if every time you try to configure that it gets ignored and then it doesn't do anything I mean shouldn't, then be a boolean like don't handle. This then handle it. City I think that's worth yeah, but.

A

The problem with the cursive tag is that, if it's way too custom, if it's not following a standard, we don't know how to upgrade the version inside the tag itself. We can upgrade this part, but we don't know what the user has added here. If it's not standardized.

B

That this is come. You can do I believe that it provides.

B

Its back but at least we can compare releases by.

B

The same solution that this in this case three point four point port, then don't one-and two-and. If we have those two tags, we are going to be able to compare them with what we already have in place. The problem here is that we are basically using the EDD client API, which abuse inversion, which is going to be three point. Four point: four and yep a missionary wiper is pretty much going to treat that as a final release, while the vendor appended parties usually treated as a pre-release, and we get into the situation.

B

Where always he already deployed position more like as a stock density portion is going to be the way they spawn as long as the symbol shown on it is the same yeah.

A

I think everybody agrees that we should stop comparing to the requester version. This is subtly, huh so maybe we should define the actual items here. Let me go back to them. Is here so this one is pretty simple: we just sew our same brush and library is going to tolerate these. As long as we don't compare to the HD questa version, we're going to have good comparisons between 0 and whoa, so we first first action item is to stop comparing to the client, I think the client server version I.

A

Guess then what we should do, then, is this question of what do you want to change the existing behavior which tomato said in the system here? Is we don't upgraded CD? If, in the coastal configuration we have a custom tag, is this correct.

D

I'm kind of I, don't I, don't have a good answer to that. I would say from from my point of view: you could we come. We parse the version. It should be a seminary person. If it's not assembly person, we just ignore it. I mean it needs to be assembly base, and then we compare major minor and patch and if doesn't match somewhere expectation, then it's fine I mean I would just ignore the metadata part of it and then I will just go ahead and write whatever I was passed, I mean if I.

D

If, if I'm inferring this from our own logic, then we compared to that. If it's new we're what we were supposed to run, we upgrade otherwise we write a wording or something. If something was passed and is not cember I, don't think we can do anything, maybe just move forward and break whatever we were told, but we can write a message.

D

This is not Sandberg, we're gonna, compare anything just go ahead, and- and this is your warning and then if they are cember again check the new version bit with the previous one and compare the major minor and patch level and just ignore the others.

A

Well, I just wrote something similar here. To be honest, it's slightly different, but it's a proposal for managing the operates based on the existing image. We can still decide whether this is supposed to be upgradable.

A

The issue I have here is that we change the behavior of comedian with these with these assumptions or what you proposed like right now, but also something to note here is that we are already using metadata, because this 0 is the revisions that we have in GCR are already metadata. So we cannot ignore metadata to me.

B

And the funny thing here is that if we actually decide to accredit for pushing 2-1 in some beliefs, this is not going to great XD.

A

You mean, but this is going to download the latest one image.

B

Only to prefetch it but I don't think it's going to be upgrading it some because if it won't, it won't see it some if it will say basically it's an older version that you're trying to upgrade to and undo that.

A

Okay, I think we should stop here with this discussion. Maybe I can draft a PR and just discuss these two potential issues from the PR I. Don't think we're getting anywhere this, so it also prefers PR, which is adjacent to this Rossi wanted to get your feedback should we allow the PR? What are your concerns? We are having separate topics here, and this is like a different change. It's just like in.

B

A like what my concern here is: basically the we are sort of divergent between config and the comfort that we actually get from the coffee table and we base key in the conflict that we get from the complete map. We overwrite the image type of that city and conditionally, with whatever made them thinks, is the most appropriate XD version and we used to upgrade, and then we store this in.

B

My image stack is typically not in the API documentation that it's basically a fruit, the cube alien promises it want to. And furthermore, if we use it goes in and qualifies the codec directly in the entrance cube a game okay to buy the his image tag is going to be overridden by this version, so the they'll pay to actually apply a different image. Stack is to base qs- Quebec I.

D

Agree with rose concerns I completely miss that part that I'll break it over right and that it's over in overriding the the image stack on the configuration and uploaded configuration, so maybe I just should close this request, and this should be fixed with the solution that we will provide to the previous discussion that we had. So maybe we should do something global for for to fix this. This issue, all together, so this was just a purple cell, but I think it. We can just close this one. Okay,.

A

So a question here: how do you upgraded Sydney in SU, say if, for instance, we completely disavow this pair to March? How are we going to upgrade the hcg on the controller's.

D

So right now it depends. It really depends, for example, if if we are, if we are bumping kubernetes just a hole, then it isn't much of a problem if we only bumped about if we only want to bump it C D and the change is only on the revision, then is when we have a problem, because we we cannot really upgrade only it C D, for example. This has happened so right now we would have to pass the cluster configuration around qadian upgrade node on on all the nodes.

D

That will be the solution right now, so.

C

That kind of falls into a category of problems is a little bit different like we never really intended to do patch wise for business, for, like say, CDE things directly from comedian. That might be a scenario for the Canadian operator to do the cluster wide passion but like the comedian operator, is very nascent I.

C

Think expanding the scope of doing patch updates for convenient is something that we haven't really talked much about the user stories. Interesting because it's definitely a legit user story. You have a CDE. You only want to update one individual component, your entire environment. How do you do that at the toilet we have today? And the answer is you need to read everything and that's not practical, necessarily for some people we're the only one update is individual component.

A

Yes and apparently we already do not support the Google revisions, so we have to. We have to implement a fix for that.

A

Yet the operator is interesting. I think it's going to solve a lot of our problems, but you know there's also this. We have to document it and we have to get it ready, but in the meantime, I think we should also get cubing them itself to be a tolerable, for these upgrade issues that we've seen.

D

Maybe I'm wrong with that I mean the chameleon operator is going to reduce most of these logic, if not all right. So it will probably have this issue as well or yeah.

C

It will like it would it would wholesale upgrade things, though. So, if you did like a one 1702 one, seventeen one, all the componentry would be rolled out as part of that release, which is probably a better scenario, anyways to be honest versus doing an individual patch, wise modification, but it would do it in a staggered fashion, right so do in a controlled fashion. So I, don't necessarily think that that we'd have that scenario, but maybe more holistic than doing a dividual patch update.

C

Because, usually when you do a quoi in typical, even upstream, it calls the whole stack, because one minor change in a component could ripple across the stack.

A

D

Have another question regarding that in order to fit the current it CD tag, do you mentioned that you have a paths for taking that from from the bulb from the pot list, and then also we recommended that we can fetch that from the local machine, the it CD local manifest?

D

Maybe do you think, maybe it's better to read the local manifest, because we are not open to API server being down or the list you know not being yet reconciled or mirror read.

A

To by the city, the rostov's or proposal, pretty much means that we don't have to fetch any versions from anywhere because only from the coastal configuration and if it's custom, if it's a custom tag in there, we decide to upgrade, we don't have to hide it. That's.

B

Only like one part of the thing so right now the check is like we get the current version. That's basically, what cube am thinks is most appropriate to hook up and they need to query.

B

That's declined to get it's. The portion that's selling right now, so I'm proposing is to basically from one side the version that Cuba name is trying to upgrade to should be taken by image tag field or if the image tag field is empty from the built in version inside of beta. That's the diversion to upgrade the version that is right now running almost 80 nodes is, should be taken from like the static port and more precisely from the that is a skin.

B

The HD container, so basically yeah I just did a solution that is just using the client right now to just fetched the tagger 408 City instances. That also allows me to replicate the behavior of where we actually compare all of the text.

B

The instances try and arrow out. If we actually have it's much somewhere but yeah we can just use the file and just the serial is nothing get the image tag from there. The.

A

Question here why why should we even parole static? Why should why? Should we even probe pots for the HDD version, if we, if the user has customized the tacking cosa configuration this means we can assume that the running version is inside this tag. If they haven't customized this tag, we can assume that the HDD version is inside the constant map.

B

We cannot be sure that the the thing that we actually downloaded from the config mode is the same thing that we are actually running, so you couldn't actually come through and we'll just change on the image tag and.

B

Great note to just update.

B

Just to create city varieties are.

A

We checking the version of the running API server before assuming we it's the same version. We want to know if it's the same version that we want to upgrade from I.

B

Don't think so, but they make secondary.

A

That is my point. We are not checking the other components and we are doing this special checks for HDD.

A

It's kind of it feels redundant to me, I mean it feels like the right thing to do check what what is actually running, but we are already not doing this for the other components.

B

Yeah, okay, then the solution is to just upgrade to whatever discussion is like the weather, like image tag called what's inside of you baby, but I think that the only the only thing that this check is doing its to just skip. It's not.

D

Okay, so what one question from my side do we agree despite the specific implementation and so on? Do we agree that if you pass a config flag with a certain configuration, qadian will apply that in conditionally? Maybe writing some warnings may be doing you know something, but do do we agree on that that if I pass a config Covidien should apply the odd.

B

Yes, it should be the same like config and the thing that we actually stored in the coffee table. These things should be the same. So if we just ask the image tag inside of a config file and by that with topic, you should be upgraded to that version or, if you choose to just edit in place the complete map, then game should be equated to that portion.

B

It's a basically it's a user who have gone through and did this change poster confirmation that is going to be applied and like the image tag and being shared post, refutes they're they're, intended to be used by users to overwrite qadian defaults.

A

All right so I'm going to draft appear. My question here is: should we close Rafa's, pure reto,.

D

If we want to do that, I can also adapt it. So we we don't change the image stack but rely on the defaulting if it's empty I can try to to change that. But if you want to handle that the vermeer I are fine clothes in it really I wanted to have the.

A

Virtual checks, mostly so the config beads I, mean probably better for you to show what you want. You mean in this fear, if you want to adopt it, I wanted to draft something related to the parts, because this is what the users are seeing right now, most of them this, but.

D

Okay, okay, so what I? What I will do is I? Will these cousins I will try to update and think about it and I will bring you back? Thank you. Alright,.

A

Let's move to the next one, we spoke a lot about this issue already misleading warning, conjoint control a so. This is a low priority issue, but it's getting annoying because the coastal key I saw it and I got a report from a user the other day outside of consequence. Well, the problem is that so backing that I mean where do I start with this one. We have a joint configuration that can include control, playing control, playing field and the control playing field.

A

You can add some bits, but if you do cube, ADM join without the control plane, you're going to get a warning that this is the warning. It means that the settings are going to be ignored, but that's actually not true. This warning is false because later in the code, we actually rewards the configuration from disk. So it's kind of the sequence of events is very confusing in our source code and I. Also I guess I should send a PR for this, but I made some proposals velocity. You had some objections.

A

Basically I proposed the following in a commit somewhere. Let's go something this commit I wanted to get your opinion again quickly on this, which is not around. Ideally, I should have gotten his.

A

So basically, you proposed that we should have a separate substructure, that is the joint control, plane sub structure, and you should populate the frogs inside of this. Instead of populating the flags inside of the main joint joint configuration right, yeah.

B

At least that's, we are in several other places inside of qadian, it's much more easy to do that way. For example, you can see what we do with like the joint configuration with the guards the discounter substructures.

B

We have a couple of like friends upstairs.

A

And we attach the one: that's yet I see the problem there to make. The Civic is that we have to manually default all the flags. We cannot use the API schema to default. These sub structures, yeah.

B

Then again with the junk control plane here, this basically don't require you to default anything to actual dutch once you attach it to the rest of the.

B

Perform at the 14 unit, I.

A

See so it's also only the bind support that is defaulted there anyway, yeah so basically you're proposing that they're quickly going through this fear, because I want to adopt it to get rid of this warning.

A

You're proposing is sort of creating a copy of the whole config to basically create a joint controlling instance.

A

Yeah I think that there is yes and then compare compare basically the custom substructure to the tomorrow empty one yeah and then I guess I only have to default the the pint port from our constants. That's it.

C

A

This is, this is a quick solution. This is going to work. Alright I want to send a paraphyletic so because the warning is confusing for users, the config, the mold. Apparently we don't have that many users that use this usually just pass flags all right.

A

This basically I had a discussion with the user. Yesterday, to my understanding, he was fairly unhappy because we still don't have upgrade apply phases and I know that Barak was probably busy with internal work at this point, so we couldn't deliver. The 118 upgrade apply phases. I just wanted to mention this that maybe like, because we have been dragging this for too much time already.

A

Maybe next cycle we should just I can take it, but with the Q medium move, if it's going to company in 118 is going to be very helpful, but I want to mention something that we ignored at some point. So basically, the user said hey in our discussion here. The user said hey. If a Q proxy configured is not exist, why don't you just ignore the upgrade of proxy and I said? Hey we already discussed this.

A

It was proposed in fact I proposed this, but it was rejected because the cubing company, the bride, faces we're going to solve that and I think we maybe because it's some much less work. We should just revisit this particular bit.

D

So I can say the Vermeer is that I know. Marik is working at that solo. He told me so, but besides that, once we have the base, we can split the work on the faces. I, don't know about this specific one. How about doing this, but we should probably split at work.

A

Yeah, it was to be before the phases we can just apply. This fix. It's not going to be too complex. Roasty do a bit about the details. There I think we have a problem with deciding whether the config map is existing, something we embed the corporate config inside the other config. That was the issue. There was no way to check if it's new, so from this, this was the user is trying to skip q proxy mode on update, apply, oh yeah, so.

B

Yeah, there is no possible key right now to tell whether actually no right now it's going to try and download the config. But if the continuities missing things it's going to just do some altered proxy and at the end of the applying face, it's going to just reapply a full cube poxy with that default. So.

E

Part of cube dns and it has a container and you have to deploy it as a pattern. That's it unless you're using your dns, no.

A

The Shepherd coverage. Okay, sorry, so honestly, do you think we could compare similar to the previous ticket we were talking about? Do you think we can compare an empty proxy configuration or something like that to just say: hey the user? Actually we cannot do that so again, I ask you the question why if we fetch the coupe Rock secure, think about if it doesn't exist, can we assume that you proxy is not the point again? The same question.

B

Depends actually if we actually try and supply the normal fetching mechanism, we have some sort of a flag.

B

It's going to be pretty easy to just don't generate new config, but only on a great, because we use the same code during say unit and drinking it. We are actually relying on generating an imported config without if it did not supply a new version.

B

So what we are we're actually kind of like we just split the codes slightly so and also another thing is that almost nowhere inside of a cube ATM, we do some like new chicks about missing two parts calculation. So we basically even in the case where we don't have a key proxy installed. We rely on having a difficult relation of the carrier and just.

B

D

Have a question here: maybe it's a crazy idea, but just let me fill this in what if it's not a name, I mean it's not adding something to our API, not a boolean. Nothing like that, just a notation that we can give an answer to our users now and this annotation we get we'll get rid of it when we have the faces. So it's like you can't user notation, so we escape the proxy add-on if the queue proxy. If you have this annotation and then do we will just drop the annotation.

D

If you have it and we'll just I mean you will have to use the faces well, the problem will fall day.

B

Is that one circuit for validation, it's pretty hard to get deliverables at some point, so it's temporary. It's going to be like too hard to get rid of another possibility to tackle this and I think that in the future we are going to document who this is using like a list of add-ons that we supply to be enabled by cube, ATM or yellow manager during uni.

B

So this is going to live somewhere in the population and we, like probably just a list of all so whatever the postal Adams folks actually come up with or their corporation. But this is quite literally to think about, like the exact details there so right now, the only feasible option to me is just to like implement, update, apply basis and users just keep the coupons you DNS or whatever face they like.

A

I'm not convinced this is the only option. It's it's the right option. Definitely, but I, don't think we we have that much of a problem with the queue proxy configuration.

A

We had a discussion about this. I just have to try it I guess: I'm, not convinced that if the kabocha config map is missing, I would assume that does not keep Roxy so I'm, not sure. If I, you know, if I wrap this in a client call I'm, not I'm, not seeing a problem here, I don't think it's going to pass.

B

It's going to pass simple because we are using the solution that is going to just check if the complete map is interesting. If it's not existing, it's going to Germany sort of defaulted to proxy component, config and you'll never know about, and this is basically the same question use.

B

Upgrade or veto whatever we actually try and component context well,.

A

Yes, I, try to say if I wrapped this call inside another check, not not the contents of this particular function. I you want to basically just check for the copy party business outside outside of this call.

B

Yeah, that's like sort of a solution. That's probably going.

A

Yeah, that was my points because we have been dragging this from so much for a very long time. Oh I started thinking that. Maybe we should just add this at this point, so people are not skipping Cordilleras, but they are skipping to proxy different and uh I think we should just inject this crack at this point. Until we have the faces, the faces are not clear to me: I, don't think if one and if one nineteen is going to still be released for them.

B

Yeah we can actually be done, but let's ensure that it has a massive to do commend. This is just a simple resolution and should be removed once we actually can find them pieces like the list that I'm.

A

Talking about sure, so, by discussing these topics is firstly, I already created like three action items for me. So a question here is: does anybody want to take this particular one.

B

So I think that I can certainly come for me. Xd stop simply what they have done should call them very tested. Last week.

D

Okay, I can I can take it. No problem, I mean if somebody else wants to take it you're free, but if nobody else wants I can take it.

A

Okay, so Rasta baby, okay, the work on the CD bits that I promised to set the PR for that. Maybe you can also handle the revision to work and I want to set the PR for this one for the Q proxy and what else we spoke about so I guess. I also have to send the PR for the other one, which is the misleading warning. Okay, I'm going to take the misleading warning in Q proxy and you guys can work on date, CD bits, I, guess.

A

D

Topics: I'm, sorry, Lumiere, are you taking the one about checking the confetti Q press? It config not as well yeah.

A

I'm going to take the Q proxy issue, the misleading warning about control, plane, joint, okay and you guys can tackle the HCD issues. Okay, perfect,.

A

Agenda topics for today we have eight minutes I.

A

Don't think we should refresh there's not enough time. I just wanted to give a quick update on what I'll be working on.

A

Rosi, you admitted: do you want to comment I.

B

Know I cannot see pigmented myself.

A

All right, I been working on the keep IDM release and test whatever tools from the beauty, automation and I'm, going to show a demo of this at some point. I just wanted to give a quick update about this, but this has been a successful experiment as far I also implemented a way for every time. Kobori discovery this releases. We can also have a walk home release with artifacts, so it builds artifacts for different platforms.

A

It can also have release notes in here, but I forgot to send PRS with release alt but yeah. It has been pretty good thus far, today, I'm going to wait for the release- and hopefully you know here- is going to appear a new release, automatically managed by the tooling I'm, going to demo this in front of similar in texture, cig release and also this group, but this is going to happen later in the cycle.

A

I just also us will also have to update the cap, because it's it has become a bit of a mess too many comments, oops. So that's it. Anybody have last-minute agendas.

A

Alright, we can end five minutes earlier thanks. Everybody see you again next week when we are going to do the planning session right.