keptn Community, 16 Aug 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keptn Community & Developer Meeting - August 16, 2023

Description

Meeting notes: https://docs.google.com/document/d/1y7a6uaN8fwFJ7IRnvtxSfgz-OGFq6u7bKN6F7NDxKPg/edit

Learn more: https://keptn.sh
Join us in Slack: https://slack.keptn.sh
Star us on Github: https://github.com/keptn/lifecycle-toolkit
Follow us on Twitter: https://twitter.com/keptnProject
Sign up to our newsletter: https://bit.ly/KeptnNews

A

Hello good afternoon.

B

This should be the right screen to share.

B

Can you see.

A

You need to add your estimation board in there again, because last meeting I created one and I think I've left my link in.

C

B

It works okay,.

B

B

So let's wait a few more minutes for giving more time everyone else to join.

B

B

And meanwhile, I sure didn't check the link to the document.

B

Do we have a link towards the ticket.

B

Hello Stacy.

B

And hello rakshit.

D

E

Hey everybody, sorry I couldn't come off me. There I, don't know just a little delayed, no.

B

B

So I'm shooting again in the chat the link to this document. So please add yourself to that in this list and if you have any topics to discuss, just add them to the agenda.

B

So we are three minutes over I think we can kick it off. So, let's start with the first topic, which is the following.

B

So from the CSF Foundation there was this um request for Action to evaluate uh what would be the impact for all the different cncf projects about the change from the Mozilla license to the I forgot, the name business license of ashicorp and I start working on that today and for lifecycle toolkit everything is fine. All the dependencies are on the Mozilla license.

B

However, for the old Captain B1 we have one package, the CLI that depends on Azure core packages and in particular one console is now a BSL license.

B

um We are depending on version 1.18, which is still the Mozilla one, everything fine for that, and we don't depend on this directly, but it's a transitive dependency, so we depend in the CLI to the Viper. Library Viper depends on Crypt. Encrypts depends on the console.

B

So if we don't bump anymore this version, everything is fine for us and anyway, since we have now an announcement that we will deprecate at the end of the year. It's not a big issue.

B

Two never bump these slices.

B

And that's it for this topic.

B

Then from the agenda we have rakshit Google summer of code project updates. Please go ahead.

D

uh Yeah so the step and aggregation fields in the captain metric the PRS have been merged and then the Prometheus and dynatrase apis have also been updated. I've linked the PRS and the data dog API PR is still under review, telling that too so I'll most probably be done with the API thing and Metric controller Thing by this week and then from next week we can start with the like the showing multiple results in the status field.

D

So we can start with that from the next week and we can discuss more about that in the refinement meeting like I have to open a ticket. Regarding that so yeah perfect.

B

Thank you very much, rakshit very nice progress.

B

How was to implement this step part in the different providers.

D

uh They already had like I used the documentations, so it wasn't much of a work.

B

Perfect thanks: maybe you can do a demo also in the next community meeting.

D

Yeah, actually I was just waiting for the Matrix controller to get updated then I'll do the demo.

B

Oh nice, thank you very much. Then we have an item from BB I. Don't know who is.

B

um Penita but oh I guess but she's not in the call.

B

F

She's in the document right now, since this was not there before.

B

Would you yeah she's in the document, because this was added recently, would.

F

You like to try to write a comment. Maybe she joins later and we pick it up.

B

Maybe we can go over this.

B

If she joins later any other Community points to discuss.

G

Oh um I tried that actually, with the link that was in the email and yeah I was sitting there for a while, too, and so then I went back.

B

Oh okay, so we have sorry that I took up with the link.

B

What what did I post it? Oh.

F

This should we change also the one on the top.

E

I had a similar issue when I was trying to join, which is interesting but I just exited out and re clicked the same link and it worked yeah I.

F

Don't know what's going on with that um thing, it's not the first time that this happens.

B

Well, meanwhile, we are trying to bring her to the call um we can jump to the next item on the documentation. Time block and I see here that Yash added a ticket here, foreign.

B

What would you like to discuss about this.

H

So yes, uh recently I got an update regarding the issue I had made, I had started working on it and now finally drafted a Content. Around scheduler I also got some of the reviews from Megan Anna, so yeah we'll be working on that.

B

Okay, perfect, so let's add this to our to-do agenda: to provide a review when we have some free Cycles. So, yes, you can finish the pr.

G

And then yeah she said you're working on a graphical. Are you making a sense out of what you've got? Do you have the information, or do you need help with that.

H

So there is a architecture, uh visual diagram in the readme of Captain, so it's a complicated one, so I'm creating a new one, basically for the scheduler part uh to the captain, scheduler architecture, so yeah we'll be drafting that also, alongside working on the architecture, content for the lifecycle operator, there are two issues which and I will be working on side by side. Yeah.

B

Okay, okay, nice and once you're done with that post, the pr Link in the slack. So we can have a look and provide you with some feedback. Sure. Thank you.

B

ah She's trying.

B

Any other topics from the documentation time block for anything that you like to discuss.

G

Not really, we've got one coming up in refinement, uh but basically I think everybody um the we're waiting on PRS, but pretty much backing off from there there's a piece on the strategy that some ways about thinking about uh moving to klt, if you're on Captain, V1 and other than that we're sort of backing off on that. Until we get more specific questions from people who are trying to migrate. So we're doing some cleanup and uh things are moving along.

B

Oh nice, okie dokie, then I think we can jump into the issue part here. We have quite some issues and you said that you had one here: Mac I guess it's.

G

This one do the others more yeah I dug out. We started this a long time ago. um We had some contributions. We had some people who were going to contribute and then backed off, and so I decided it's time to go in and finish that up.

B

Nice, these are mostly good for tissue, so this PR already provide you with.

G

Great and then what I'm doing the I mean what we're doing right now we had information in the contributing.md docs, we had two of them and they were overlapping and stuff. I am basically moving that information in and then, when we're done, I would love to have lots of people looking at it, and we know we need some more materials and see what we need. That's not there so and hopefully by next week, we'll have everything in place where you can look at that. So.

B

Perfect and I think this is a good candidate for a good first issue, epic, since a lot and lot of these tickets are good. First issues right.

G

But you'll notice that a lot of them are already closed too there's relatively few that are open, so they gash did a couple of them. Actually didn't you, yes,.

H

Yes, I started contribute to Captain through that issue. Only yeah.

B

B

uh Where is the one from here it is fun.

B

Would you like to give us a overall.

D

Yeah, so thank you. So this is what I was planning like how we should modify the captain metric struct to like have the stored results field so for the spec I was thinking to introduce a stored results field that would accept the entry value so like a user would be able to Define how many results she wants to store in the result, then we'll have to modify the status field so for that I was thinking to like build a struct for result and then have a area of results in the status.

D

Now we'll have to create a maximum length of how many like maximum results. We want to keep inside the status. So I wanted to discuss about that like what should we keep I discussed this with Florian, but just want to take all of your opinions.

A

All right, uh I just have two uh comments on that. uh Maybe we can rephrase the second sentence, uh which is the metrics controller, with curator slide provider for stored results. uh I think we can rephrase it that, because the controller will keep pulling the API periodically, but the last stored results are gonna, be stored in the status, so but the metrics controller will still keep to to communicate with the API.

A

Also after this, this number has been reached just to yeah yeah and then the other thing is uh because then in the status, we would have a an array of results uh here. I'm not sure if we should keep the raw value of each item, because this will likely uh lead to.

C

A very large payload.

D

So should we remove the raw value entirely or how.

F

We have a way to store errors in case the row. Value will return this a failure.

D

I, like I, kept error message separate from the result.

F

So our message would be for one single error.

C

F

If we have multiple metrics failing, we would lose this information.

D

So we can keep error in the result.

F

B

F

Row value yeah, the.

B

F

B

Value if we need to keep it for all the different results, very easy to go over the 4K of a DCd.

A

B

A

The error would make sense to to keep.

D

B

So we can provide awesome, more try to be meaningful in the error message, so we can say this specific one is the one that fails.

B

um No I miss to understand what you like to change in this sentence. Sorry.

A

I would say, the metrics controller would store the last spec dot range dot, stored results in the status field.

B

F

How would this work, no yeah.

B

Exactly um use.

F

Because don't we have multiple metrics at the same time? So how do we keep track? Do we store all of them.

A

So within one metric, if you set this expected range stored, results to 10, we would keep the last 10 results in the in the status of that metric.

B

Now I see what you mean with the issue.

C

B

Three three times to understand: they look this. If you read, this looks like that every time according 10 times.

F

B

Sli Professor.

F

Is actually checking the status of the CRT is not querying the API multiple times, that's what human!

F

So, when I'm calculating things and I want to use more than one result, I just check the status of the CRT to get the old ones.

A

Or exactly yeah, so this is basically on history of the history of the last couple of okay.

G

A

That have been retrieved then, and then, when you're done with adapting this, can you scroll down to the to the results array again.

B

And start the number.

B

I will fix it later. What do we like to discuss flow.

A

uh Can you please go down to the results array in the example?

A

I think here would make sense to indicate the start and the end time of each interval.

D

Inside the range or like inside the result, shaft.

D

I think we should do it in the results truck because then we'll have to modify the range struct, which will lead to the like. We use that struct in the spec too, so I think we should do it in the result. Struct.

B

F

Not really because you could update the CRT way after you have done the query. If the query is a bit taking long, no.

B

If you have a last update date like this one, will it be this minus the interval being the start? Yeah.

A

I think you can derive it this way, yeah.

B

C

B

Will create the sort of provider in an interval of at the top and keep the last.

B

Amounts off results.

F

Does this mean that it also needs to remove the oldest and put the newest.

B

B

D

What should we keep the maximum length like the maximum value for the store results.

B

You mean what would be the default value.

D

No, no, like the maximum value a user can give so that way. Yeah.

B

That's zero point because we cannot really exit much, maybe 255.

F

uh I would test it out as part of the ticket.

D

So that means there would be 255 results in the crd.

F

Yeah try to store a CRT with the such a huge thing. Let me see if it works.

B

Yeah, let's start with two five five, and if it's too much too few, we can do a follow-up.

D

And then uh I'll have to remove like the air error thing inside the result: struct right. Okay, yes,.

B

You provide it here, mix total sense.

D

B

You, thanks to you, done: let's see whether we need a boat who managed to join. Oh.

E

No I was trying to walk them through multiple ways of getting on and they weren't able to join. So I really requested that they create an issue in GitHub or join the slack and ask us there. So they said they will.

B

Oh perfect, thank you. Stacy yep.

B

I forgot to add the ticket from Mike. I will do that later. Then any preferences someone wants their ticket to be discussed.

B

Otherwise I can go ahead with mine, which is a freshly created ticket today, because kubernetes 128 is finally out was released yesterday and looking at the release notes, there might be two interesting things for us, but the validating admission policies now they're in better and they provide a more customizable way of validating crds. So this can be an interesting thing for us, since we have quite some CRTs and few validation workbooks and also there is matching conditions for admission by books. Right now.

B

We include some logic to decide if we should act or not inside the rotation Web book itself, but maybe we could use this new API to let the Playbook color between the heavy lifting instead of doing that by ourselves. On the code side, So This research is to play with these two features and see if there can be any good or bad for us and if so, we can create a following up tickets, but also over the release, notes and have a more in-depth look. Not just this briefly Overlook that I did.

B

Maybe there are more coolest cooler features that we can play with.

G

um Giovanni, do we know, uh does klt in its current state, run on 1.28.

B

Technically, yes, because we have 127 as a cube client, therefore, we are compatible with 128 by default. Okay,.

F

We will soon have 127.

B

Didn't we already immersed the PR2 bump since the FED client is fixed now.

F

um No, we discussed to do this after.

F

um But if, if you want to this could be done, yes,.

B

We should do that and also another cool thing about. 128 is major backboard and forward Port compatibility between cubelet and the different controllers.

B

So all the runtime controller can support more versions and not just the previous internet. Following up.

B

B

Interesting things I removed them. Yes, then, let's start top to bot.

B

Improve looking in controllers by adding relevant context, Road, you want to give up overview about this ticket. Yes,.

A

Please, uh yes, essentially, this is about improving the locks and streamlining the uh because recently I was debugging, something in the livestock toolkit operator and I found that in several places like this first example that I've linked here is that we just output a lot message without any additional context. So, for example, this searching for a cabinet version, but without any information about uh which particular object and which namespace and so and so on.

A

And this makes it uh yeah not really usable or that usable for debugging purposes, uh because you cannot filter, for example, on a particular resource, name or namespace. If you do, if you have a lock filtering mechanism in place and therefore this ticket should be about adding.

A

These are properties like in the example, if added here below, where, whenever we do a log, we add some common properties like, for example, the the kind of the resource that we're reconciling in the controller. That's generating this log line, and the name of the resource and the namespace, for example, to be able to filter on those properties and and only get the relevant logs for troubleshooting something.

A

So ideally, we would have, for example, in the in a common package in the controllers package, we would have a method generating the keys and values for these properties, and then we can add that to the to the lot lines.

B

This house very much a good first issue. Indeed,.

A

And maybe to keep the PRS as small as possible, we can uh to individual PRS per operator, so one PR for the metrics operator, one for the life cycle and so on.

B

B

If you have something for the scheduler.

B

Anything else for this ticket.

A

That's about it, nice.

B

Then Anna a ticket about collectors.

F

Yeah, so we had already doing uh research about adding collectors to our cattle dust. Basically, the collectors in case of failures uh returns you whatever you want in a form of a could, be skill, comment, and in this case a very good idea would be to have a log so that, basically, in our test log, we got a log of the Pod of the crd or whatever. It is needed, according to only what fails and we don't have to go through an insane amount of files to figure out what was wrong.

F

The pr is linked with the example of how to use this and I just listed all of the tests and what could be added there as a log, in a hope that maybe multiple contributors wants to do this.

B

And they will be attached to the Ci or.

F

Yeah, if you look at the conversation, I put a screenshot at the end of it of how it will look like.

F

B

F

So this app is running is the log of the container, and the cubicity unlocks is what the attached collector was supposed to do. So we can do that with crd. We can do that with the default lifecycle operator and so on and so forth.

B

A lot of things to be added, yes,.

B

I already wanted to nice.

B

This is also a very good first issue, as you already marked it nice nice then another one from Florian.

A

A

Because we recently discovered that adding a new custom resource definition is uh quite a task, because it involves a lot of manual steps right now, due to the way we generate the docs and whatnot, uh and this ticket should be about uh improving the quality of life developers, adding new custom resource definitions or changing them by basically uh providing a a script and a make file that consolidates these commands that I've listed here.

A

So basically, the developer workflow should be whenever we change something in the the custom resource definitions. We do that in the uh go structs which are the source of proof for those, then we execute one make command and it should take care of all these manual steps.

B

Could be defined.

B

What should be the name of this entry.

F

F

Update Helen chart.

B

It's not just a chart, it is more than that because.

C

It regenerates.

B

The auto generated files.

C

B

Creates the new manifest you copy, those new manifests into the charts and updates the documentation of the CRTs.

B

Your muted flow.

A

Sorry make update crd resources.

F

F

Role, we will also have makeup date are back, and then we will have a common code maker update I, like it.

A

And then we'll have make resources that executes all of those.

F

B

We call it already make resources.

A

C

A

G

For the documentation should that go into the contributing.

B

Yes, I wanted to add the link to the file.

G

Or to the contributors guide, we're kind of pulling the content out of contributing.md.

G

B

B

I think I need to wait for you all.

G

Right there you go contribute by Ian right under Ian is contribute uh perfect.

G

um Go under software.

G

B

Okay, it works anything else to discuss for this ticket or all clear.

F

um Just to point out that there is a little Helm label template part to be added at the top of the crd, just as a it's specified.

C

F

But uh maybe it was not super clear.

B

I marked it in bold Maybe.

B

um Warning yes,.

B

uh Two: we have an example for this.

F

Any existing Security in the home template.

F

Like the first one.

B

That's the very very first one which is.

F

uh There is the release namespace line, seven and uh include side labels, 912.

B

And kind of select the right one: okay, so 12 and 7.

B

B

Let's keep the full URL, otherwise,.

B

Don't okie dokie.

B

B

This says this is an interesting one.

B

um Speaking with bread, we discovered that there are a lot of use cases for having a configurable service account for Captain dusk, because by default we use the captain default service account.

B

But since we are following the least amount of permissions principle, we cannot do much things in the cluster and now that you have the container runtime, we might want to do something like collecting all the logs from all the different pods workload that have been generated, and we cannot do that now with our service account.

B

So it would be nice to have a new entry in the captain task, CRT, where you can specify the service account, name and name space, so it can be mounted and the task can run in higher or lower privilege.

B

And that's it any question for this ticket.

B

This sounds pretty much a very good first issue, because it's one line in the crd.

B

Let's add a better definition of done.

B

G

B

Generating the.

B

B

So this should be a good first issue for the community.

B

And then another one from Florin very descriptive.

A

uh Yeah this is, it could be refined um yeah. This is basically about uh providing an example in the docs, for the new analysis feature that we're currently working on.

F

um Can I say I think we also had something as a safe for adding the analysis inside the example folder?

F

Was it this ticket or the other one? Maybe I'm confused.

A

uh Yeah this was this one, so not in the docs, but in the examples next to the to the other examples, we also have also need one for this feature. There.

B

uh Is this a documentation, one or not, I lost it. No.

F

G

That's an example: we.

F

C

Documentation for.

H

G

Crds and all of that in the how-to chapter that I'll reference, the example.

B

Yeah because we have uh documentation ticket ready. Let me refine this uh cool that uh PhD our Docs.

B

Explain this the new feature.

B

F

That the samples are already available, those we already added, oh.

B

B

So only in the example folder.

B

B

B

And we're doing a great job in good first issues, since we were running low on them and now we have 20 23. nice.

B

Now we have any other topics added to the agenda or should I continue with.

B

Taking this item.

B

Or maybe shivang rakshit Yash? Do you have any other ticket that you like to talk.

H

As of now, for me, I'll just.

A

H

Taking all these issues that you just opened.

B

Okie dokie, then, let's continue with implementing the conversion function from Captain, V1 SLO to analysis definition, hello. Please go ahead and talk a bit the details for this ticket all.

A

Right uh yeah, this is basically to uh allow or to provide support in the migration from Captain V1 to the curtain. Lifecycle toolkit by basically translating the the SLO yamos that are used in Captain V1 to the new analysis. Definition resources- and this is yeah about yeah- writing the the conversion Logic for for the SLO yamos.

B

Where this should be introduced,.

A

uh This should be part of the metrics operator and yeah, where exactly to put this.

A

I would say under the controller's common analysis.

B

uh But my question is towards more like how can I trigger the conversion I run the operator locally as a command line tool, instead of being an operator, is basically a CLI.

A

That I would see actually out of scope of this issue, because this should really be here only about uh implementing the logic and then wiring this up and uh providing the migration uh the usage for that would be part of a follow-up ticket. So they should really be only about implementing the conversion logic to keep the scope smaller possible.

F

This would be go code, not a bash script or anything like.

A

That no not not bash, no go code.

G

And how will the user implement the execute this.

A

uh This will be part of a follow-up application. I would say that so the goal of this should be to make this component uh usable either by a CLI or by API endpoint. So how it is uh is going to be invoked should not matter that much for this component.

B

Yeah, maybe we can, you know parse and if a flexors conversion, we don't set up the wall controller part, but we simply just convert and exit.

B

Could be a way, but that's a decision by the implementer of this ticket.

B

And similarly,.

A

Exactly this is uh pretty much yeah very similar to the other one, but this is about mapping the cap, V1 SLI resources to the analysis, value template Securities that we recently introduced.

B

And same as the other ticket, just implementing the logic, not wiring the top.

B

A

B

A

Yeah, uh this is about storing the status of a concrete analysis to the analysis, custom resource definition, uh but I'm actually not sure if we need a separate ticket for that, because uh this will most likely be a byproduct of of the implementation of the analysis, controller.

C

F

But maybe we can link it in the an Isis controller, since there is a suggestion on status.

B

I, don't see a teaspoon, because the scoring function shouldn't be part of this ticket was part of another ticket.

A

uh Exactly this coloring function is part of a separate issue that owner is currently working on, but storing the result would then be a responsibility of the analysis, controller.

B

F

Yeah, there is actually a dependency in the last part of the development of the analysis, controller, where you need to score a function and the result out of it. You will store in the status. Yes,.

B

E

B

Ticket linked now there is.

B

Perfect any last moment ticket that we should discuss. Otherwise we did a very huge backlog of items.

B

Any other topics for the community or documentation- yes, rakshit.

I

uh Yeah I just saw that we have a label for operator I, think we should rename it to lifecycle operator now that we have renamed it in the code base, like just a suggestion.

F

We don't have the label already. We just need to.

I

Check yeah: we need to rename it.

F

Now the problem is that the label is existing already. Oh, we need to rename them manually, I believe to return them manually.

B

B

F

uh We have some tool that automatically tax operator tickets no.

F

B

I won't destroy too many things with this change.

B

If it will take a while I guess ah perfect so now.

B

Yes, no more tickets assigned to the operator and from 18 or 12 that we were now 27., so delete it. Nice thanks rakshit.

B

Okie dokie: if there are no more topics, then we can finish seven minutes earlier than usual thanks. Everyone have a nice rest of your day and see you next week.