keptn Developer, 4 Feb 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keptn Community Meeting - February 4th, 2021

Description

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

Learn more: https://keptn.sh
Get started with tutorials: https://tutorials.keptn.sh
Join us in Slack: https://slack.keptn.sh
Star us on Github: https://github.com/keptn/keptn
Follow us on Twitter: https://twitter.com/keptnProject

A

Hi everyone and uh welcome to this uh captain community and developer meeting uh today is february 4th. uh We have quite an agenda here. As I can see, uh we got um two new contributions, also from the captain community, uh but as usual, um I I'm asking you to fill in your names in the list of attendees and I will just go ahead to give you an example just write in your name. um You can also uh add your affiliation uh if you want to, uh but the name is also uh just fine.

A

um If you are not yet uh owner or of this document, uh the captain team that maintain us will will approve your um your addition to this document and I will go ahead and also share this in the chat. So you will have this link.

A

Also here you go so this is the link and you can just uh add your name here. uh Thank you very much. uh While you're doing this, um let's start with the first agenda item, um that is our new relaunch or release of the of the new front page of captain.sh.

A

You might have already seen it. It got a little bit of an overall and you can see more. It's now more focused on the core use cases.

A

Instead of focused so much on the technology and how you get started, we still provide um some prominent uh points how to get started which will bring you to the quick start section and it's still the same uh experience. um You just bring your your kubernetes cluster and then get started. Just try it out today. I think the installation took about one and a half minutes, which uh I was surprised that it's so fast but um yeah. This is our new website.

A

uh We are actively looking for feedback, so if you have some feedback, please put it in the um in our slack channel or reach out to us on the various other channels that we have twitter is fine. Our meeting list is fine, um and uh but also if you already joined us like, we have a website channel there, and we appreciate your feedback.

A

uh Then uh he meant already mentioned it, and I will just put it in here. I thought I already did it, but uh we are now part of the lfx mentorship program, so captain has applied with two projects and I will just paste the link of this link.

A

Still here it's this link, I paste it in here and uh if you're interested in contributing to open source and if you're particularly interested in contributing to captain, we have proposed two projects.

A

One is focused on uh integrating uh to write a better integration for promethoice for captain and the other one is focused on working on the captain cli and providing a way to get to kind of a kickstart to uh to integrate with new services, to captain to kind of create the skeleton uh in a template where you can already start with your captain integration.

A

So this is our two projects. If you're interested, please take a look there and apply via the cncf and then the captain uh mentors, the captain maintainers will um will uh find the um or will accept the applications.

A

So with that, um I think I'm opening up a discussion here. We have an ongoing discussion um if we should hide implementation details in the api or it's around the naming of apartment points, maybe johannes or christian or andreas you can share something. I was not part of the last meeting.

B

um All right, um I will, I will share my screen because I have it. Thank you.

B

All right, yeah, I pulled in this discussion because the team is currently discussing how we should call certain api endpoints, because we want to hide some implementation details on the api and thought about renaming api endpoints and, as you already know, there is the configuration miner service endpoint. Then there is the mongodb datastore endpoint and these two gentlemen. We want to rename into slash api resource sys and the second one to slash api event store and the default endpoint of captain should be slash.

B

Api control plane where you can then um send events to, and you get also the end point to fetch events.

B

Now it was a little bit my concern whether we should stick with the name of event store for for this particular endpoint, because when we have the name store in there, we kind of um conclude that there must be something like a store, but in maybe in the future. This won't be the case and that I propose to just call it um events endpoint, because then we are pointing to the entity that gets managed by this api endpoint and not the implementation behind that all right yeah. This is currently ongoing discussion.

B

We have not find a conclusion yet, and this would now be the place where we should start discussing this topic.

A

uh Do you want us to do the discussion um basically, should we take five minutes for discussion here, or should we discuss it on okay? um Let's.

B

Please use five minutes.

A

uh So I can already share my opinion. I think it's a good idea. The the last comment that you made to have it just called events, not only because of the points you mentioned, but also that we can avoid.

C

The camel casing here.

A

That I think we don't have in other parts in captain, and I would also go to rename control plane to control plane, but with lowercase without the the the uppercase p, just to kind of avoid any any confusion. And uh then I think uh it would be api control, plane, api resources, api events.

A

That would make sense for me.

B

Let me just summarize what you have said, but the others are welcome to bring in their idea.

A

I also allow we have other members also joining uh hi chris. I will just allow you also to unmute if you want to share your opinion.

D

So- and you brought up one argument that we shouldn't have controlled per lane inside the euro and because it's not a resource right.

E

Yes, because now we have the other two which says events which are kind of a rest resource. We have resources which are also resources and another, and we have control plane which is, from a rest principles, point of view no resource so to say, but I don't know it's if it's a good idea to look on this that strictly because what are the alternatives so.

A

What what do I get when I ask or which entities can I query from control plane.

B

Let me give you the example: that's perfect, I'm opening up the api.

D

But it's mainly can entities project services.

B

This is what you get on: the control plane, endpoint, an end point for a sending event and fetching events and then the end points to manage project and service.

B

That's pretty much. It.

D

If I could ignore the technical implementation, I would like to have it directly in the root so.

B

Kind of that approach.

D

B

This way just to have it on the default end point there.

D

I know and the technical implementation will not that easy, because we we also have a swagger, ui or api, also for the for our api service. But I guess it would be the most yeah, the most logical implementation.

A

And we still have two different version.

D

Of events, we have an api v2.

A

uh Sorry, api v1 events and there is api events, v1.

D

So right, but um the second so api events we want is only in the store of the events. So a user should only use this for querying um events, which happened, and this will on, and this will contain a single or two endpoints for retrieving the events, but not for posting events. There.

A

And can I retrieve events also via api v1 events.

A

No because what I saw in this.

D

Regular ui there will.

A

Be api only one and then projects, services, events.

B

A

That's the trust.

B

A

A

um But the second one they get will will give me events. It says here get triggered events.

D

So good point: we also have to streamline this, but let's do this discussion on on a level up so on the on this level.

D

Not on the endpoint level, because otherwise it will get really complicated. Now.

B

Just to give my two cents, I like this approach to to have it on the default. Endpoint, all the control plane, relevant elements.

A

I like it, my only concern was that there are two event endpoints that are very similar and the one.

C

That first starts with api v1 event: api events. We.

A

Won- and you have to know.

C

Which one you you.

A

Have to take, but I like the approach.

C

That it's not the control plane in.

A

It because it does not, um you yeah, you don't have to you're, not clearing something from the control plane, and I think this also was from the community from adrian. I think he also mentioned earlier on the part that andreas.

D

Mentioned yeah, then we only have to find a technical solution for that.

E

Is it because we then need to somehow merge this wiggle uis.

D

D

I'll get rid of the api service- I I don't know so, but let's do this technical discussion on the issue.

B

Then I will take a summary and then wanna want to say that we agreed.

D

B

B

Thanks this was it from from my side discussing this issue.

A

Great perfect thanks so much uh I would just briefly share where the agenda is going and uh up. Next, we are already discussing um noteworthy issues and contributions and we start with our community contributions. uh I will hand it over to email. uh I think.

C

It's your first contribution.

A

To the captain project so really um very much appreciated. Thank you for that and uh we usually uh do a quick. um You should do a quick demo or just showing uh on the issue um on the pr uh what you have achieved.

C

Well, logan. Thank you uh I'll just share my screen.

A

Yes, that would be awesome.

C

uh I hope in screen specific.

C

uh Is my experience? Yes, awesome.

A

C

So uh the first contribution to captain was- uh I picked up this issue where uh the cla basically didn't have uh support for shell auto completion. So thanks to these booth links uh and uh captain cli uses, so I had checked it out. I had learned how to uh write auto completions and um I I proposed a pr policy. It called much so a quick demonstration would be. If I just go to my terminal and to captain and just press tab, uh it will show you all the uh available commands. That's there.

C

So, even if I go off uh like, for example, you need to provide an epa token and an endpoint, so it can give a good description.

C

So that was the first beer and uh the second uh issue which I picked up was this one? Where uh let's say you have a captain, cli version of 4.7.1, but there's a new version of captain, that's out there and you want to upgrade your cluster so first, as the issue suggests that we did not provide any of like a like a command which they could uh the user, could you know, write just copy paste with their terminal and upgrade their captain cli right from there? They have to either go search. How to do that?

C

Also, that was the first part of the issue and the second part was uh when you do a captain upgrade and your cli version is lagging behind. For example, right now the cli is 0.7.1, but uh 0.7.3 is available. uh It would produce this message, which is not pretty intuitive, so uh I proposed uh see uh I proposed uh issue where you know we could.

C

We could show something like uh whenever the whenever you run captain upgrade, it will do a quick cli watching check and if there's a new cli version, it would prompt the user to upgrade their cli first, uh which would then enable them to upgrade their cluster with ease. So, uh as you can see, uh I think I have an old build that is standing up which is currently on uh 0.7.1, as the issue mentions.

C

So here you can see when I, when I put caption version it tells that a new uh version is available and it right here gives a command, so they just you, know, copy paste it right into my terminal, and I could upgrade with these and uh if, if I run like captain upgrade.

C

So it would first do a cli check uh yeah. So as a new version is available, it just prompted me the user, to upgrade my cli version first, so that I can upgrade my caption cluster. So that's about it. So both these two issues were related to more of user experience side of captaincl so that it's easier for people to use.

C

I think that's about it.

A

Sorry, I was muted, uh pretty cool, pretty cool uh for for the second part, uh I did not get it that it, so you were all already on the latest version or you kind of.

C

You know I kind of built an older version to demonstrate how it would work out.

A

Okay: okay, cool! uh Are there any questions from the community to he-man.

A

Robert, if you want to unmute yourself, um that's perfectly fine um and just.

C

Well, uh thank you robert. Yes, uh I'm a csh user and if I may, I forgot to share something so in case uh for the auto completion to work you can type in like uh you can type in captain uh completed and give a flag. It will tell you how to enable auto completion in your shell. So I am a csh studio using power level 10k.

C

So I can directly it tells me what to do and if you are a bash user or if you are a mac os user, it will tell you how to enable the auto completion for it to work with captain. So this is something which I missed.

C

So that's there one once you put all that uh configurations in your dot bash profile or dot or csh dot rc. uh You can just go captain and hit tab for the auto completion.

A

Cool, I'm really I'm looking very much forward to to give it a try that looks cool uh thanks so much for this contribution. um If there are any more questions, uh please uh just ask the questions. Otherwise uh I will be um asking adrian to share his screen and show his contribution uh on the custom, mongodb uh deployment and I think, there's also another one. So.

F

Yeah, I I just added another one, just to sum it up. Thank you. So let me show my screen if I find the.

F

Yeah, so actually there were two contributions. uh One was very simple and pretty similar to what has been just presented, but in a different repository. It's in this get captain sh and let me share the whole screen.

F

Yeah, so the contribution was related to auto detecting version, and I submitted some code that that would basically find the latest version and the the most tricky part was actually uh curling uh releases from gitlab and our proper grapping and using oak, because we run into issue with get shci tests where something that was working for me locally on mac was not working on.

F

I don't remember what you use for ci for travis right.

F

Yeah, that was tricky part because- and it was super strange because I was coding it on mac and and it didn't work on mac, but anyway we used very simple solution uh and I'm very sorry I don't remember. I think it was christian right who was helping me. So thanks for that- and I think also he did some some notable improvements on the pull requests.

F

uh But anyway we uh we used some functions for for the um like atom and atomic functions for bash script. That made it a bit clear, so we are basically getting either the latest version or the the pool of latest versions, uh and we are also verifying if the comments are cool or installed, and then we have some code for basically checking if the provided captain version is correct. If it's not, we are checking all the available versions, uh so there are basically few uh ways you can use this right now.

F

First of all, you can provide a a proper, uh a proper version of of captain, let's say: zero, six zero and we're gonna pull for zero six zero and we're gonna install it. The other usage is when you don't provide the version at all, then it should use the latest release version 0.73 and the last, let's say edge case. um I can think about this using the captain version that doesn't exist, then we will have the information that the selected version is not proper.

F

uh I don't know why the formatting here is a bit off, but I've I think it should. It should just list all the versions. So this is something maybe maybe to fix, but uh it's listing all the available versions for the user if they made a mistake. So this is the pull request. Number two.

F

So it was very nice to contribute on this one, and thank you thanks christian christian for help, and the other thing was during our own. Implementation of of captain uh in quitope would discover that uh using external, not the one that is provided within the captain. Installation is pretty low-hanging fruit because it requires only a few changes to the helm, charts and again. I was not alone on this one. So one this this time. Andreas was helping me with with this pr.

F

So finally, what came out is that we are able to use the helm chart and provide instead of using the uh the built-in, um the built in mongodb for captain. We can use the our own mongodb instance. So I have in my kubernetes cluster here I have namespace where I have my mongodb and basically this is from bit taken from bitnami home chart, so nothing special. But the important thing is that I was using previously this to deploy captain. So I already have some projects created here and I don't have captain.

F

I deleted all the all the name space so now, when I just use the helm chart with proper values, and I will steal the values here.

F

I need this one, so I'll use this one. So we can install the helm chart passing some additional values and the additional values are uh first of all, control play mongodb host. We are passing the host of external mongodb in in my case, it's it's service of mongodb in namespace.

F

We are also the most important part. We are also setting control plane, mongodb enabled to false, which will not deploy the internal mongodb with captain, but we'll use the external one. We will also pass username mongodb password for this captain user and others additional feature. We can also pass use admin password for for the advanced features and once we deploy this.

F

Of course, the demo.

F

F

So once we once we have our chart here, of course, for external users, it's going to be the chart somewhere on on our own home repository.

F

We can deploy this from chart and we were fighting with the deployment before the presentation, because I had my own fork.

D

You have to go one level up. Oh.

F

Sure, because I'm inside the captain yep demo effect as always so now we should deploy the whole captain uh control plane with uh all the require uh required applications, even a bit too much because we're gonna have this statistic service, I'm on master. So it's not released version. So as previously, we will just get rid of.

F

It so, as you can see, the the applications here are deployed, but there is no mongodb.

F

So, let's wait for them to.

F

To create okay, and now I will just use my port forwarding, use captain, and now I will just steal again for the big chunk of.

F

Comments, I need to use to.

F

E

F

Maybe the service is not running yet.

F

D

Yeah, so we had this a few minutes ago up and running.

F

um Yeah so they're my effects for sure, but.

D

Just let's try and to do again a put forward.

C

D

To kill it and reopen it.

A

Should it include the path here? um It just says, path, slash api, slash one uh not found. uh Should it include the the host name as well uh on the error message on the in the middle or is oh no, it says it could not.

F

That's interesting in this case.

D

So, did you update your cli no.

F

No, I didn't update it so like there is no changes. Basically, we have this one.

D

Can we try and to go to the swagger ui.

D

Localhost: okay, 8080 slash speaker, my most ui api.

D

A

Sometimes, when I open and kill port forwards and sometimes not working.

F

uh Maybe or maybe this is even simpler, maybe I broke something when I was having fun with get captain here.

D

But still this vega apa should be with or yeah we should reach it.

D

Let's try another port 8081, maybe.

F

Do you remember which is which one this one is this one is, I think, from the host right, no wrong.

D

ah It was correct.

E

And uh let's give it a try.

F

No, it's not working. Oh! It's not working.

D

Are you getting the the bridge so just delete apa in your.

F

Browser just like this right: no that's not nice should be working.

D

um Nevertheless, like yes, we do not have to fix it now, because it's not related to your change.

D

The feature you have implemented is awesome, because this now allows users to simply use their own existing mongodb.

F

G

But it was wrong.

F

For wording, it should be forwarding from port 80. The service of ap gate when engine x is not hosting on should be hosting on 80. Okay, so now it should be working documentation. Is this incorrect in documentation?

F

We have to fix this. Actually, I I have it stored in my big bag of okay, big bug of anything. So maybe I made a mistake: it's a demo effect. Okay,.

C

F

Anyway, we should be able to get here. After all, this beautiful fight.

F

Really cool feature yeah: you need to pay the price and yeah and we have the the thing that we have in have in external database of so I was trying to create project from here, but actually I have the captain cli from 073, which is not compatible with master anymore. So, uh let's just stop here that we can use the external db. That is basically here in namespace mongodb uh and I think if we have some, if we have some logging, then maybe we can see something here in the locks.

F

I don't know if it's locked or not when we open the project yeah, there is something.

F

Yeah and we should we should have some old artifacts so, as you can see, it's using the external mongodb too thanks, not the smoothest, but it's working.

A

Really cool really cool. uh I I know that uh we already had some captain users asking for this uh to have an external mongodb somewhere.

C

On the same cluster, but.

A

With um in the different name space, you also had it on the same cluster below just in a different name: space.

F

Right, it was not a different.

A

F

Yeah, it's it's just yeah, but the what andras fix in pr and it's a very good idea to have the basically host and port together and we are just using one host one host and port variable. So you can point to any network available service.

F

Thanks guys and thanks for help again.

A

Are there any questions for adrian.

D

A

uh Then I'm uh handing over to uh airmen um he's also. um He has prepared some something to show us today. Thank you, yeah.

H

Thank you very much. uh Thank you. I'm going to present uh three pull requests today. uh One is the sequence: style refinement um the second one I have already um presented that one last week, but there have been some more changes during this week on that one.

H

So that's why I wanted to present that one again and just show the details here and one uh thing that was implemented by clau uh plus, uh but he is not uh with us on thursdays, so um I'm going to present also that one uh that we now highlight the selected stage in the timeline. This is also coupled with the second pull request to the sequence. Details um yeah, as uh you can see here in the uh pull request the screenshot, but I will also show a live demo.

H

The sequence star refinement is about this tile here on the left side, where we showed the currently ongoing sequences and details to this sequence so when it started how long it took and what was the recent task and also the artifact and the stages it went through.

H

In the second pull request, I will show it in a live demo, because here I also have only a screenshot.

H

We added more details to these tasks, so when you expand them and yeah the the stage timeline uh now also highlights the selected one so that you can see uh which state you're currently looking at and yeah coming to the live demo. So when you click on the sequences screen, you now see the list of sequences that are executed or have been executed in your caption installation.

H

As you can see, I have here artifact delivery running. It started yeah a few minutes, two minutes or already three minutes ago. uh The recent task is uh release and uh yeah. It already finished successfully in the dev stage and currently it's in the staging stage. When I click on that one, I can see on the right side the stages as you can see, the stitching stage is currently selected here. It's bold and yeah a little bit larger than the other stages.

H

You see the progress indicator also here and that the release is currently going on, as we can see the evaluation finished and when I expanded that one now, I can also see the details about this evaluation step, so you can see that the lighthouse service started here and finished with a result pass and yeah here on the right side.

H

I can also have a look at the payload of each of those events that happened in this sequence, yeah and, as you can see that now the progress is already in the production stage and I can switch by clicking here on this timeline between stages, and I can see the steps that happened inside that stage.

H

Yeah, that's actually uh from my side. Are there any questions.

B

I mean, I think I added a comment on the pull request and just wanna make sure when you zoom out now a little bit is then this part of where you're, showing that oh.

H

Yeah sequences.

B

Now a hundred percent.

H

Yes, this is uh filling now the screen so um uh yeah. I cannot demonstrate it now, because my list is here pretty full of uh um events but uh yeah it would or it will always expand. 200 percent, uh at least here is a padding, but it will always have the full height. So I guess, when zooming out perfect,.

D

You can see that it's.

H

Always full, but uh yeah, zooming and big screens are not always the same effect but yeah. In this case it will always use hundred percent of the hate.

B

What I really like is that you can see in the tile already the status of the different stages. Therefore screen staging is green and uh screen as well, and I assume that it will turn red when there is a problem. Yep.

H

Cool, of course, that's what I wanted to show if one of these stages is broken uh or one of the stages in a sequence is broken, it will show that sequence or that staging in red. Yes, okay, for example, yeah here. For some reason, the deployment was not working.

D

Yeah, oh, that's really cool. um Can I go back to the error.

H

To the error message: oh yes, uh that's just the message that I get in the payload, um which is, of course here also uh showing up yeah.

B

Very helpful, I think.

A

A lot of users were.

B

Looking for this to get an understanding what happened uh for certain tasks, I I think.

A

That the captain team is always happy to see error messages.

A

They are not that frequent, but when they happen.

H

I, like those messages, more finished, release, yeah.

G

A

Already, based on the new shipyard format, so um or yeah cool.

H

Yeah, so this is already the new shipper with uh yeah that enables actually uh that we show this ongoing progress because uh yeah when the release step uh is triggered, we have a started and the finished event, uh which is here kind of grouped um into one tile and yeah, showing us the progress, also very nice yeah. Thank you very much. That's from my side and I'm handing over to um christian christian.

G

So um I have a lot of things to show or to tell actually I won't show everything so I'll. Just start with this one, um I've introduced the pentabot, which is uh well technically it's already there, but you can configure it to scan, go modules in the mono repository that we have. Can you make it clear.

A

G

G

um Essentially, this: oh, it's it's very slow, so uh essentially it's a config file and what it does is. It creates a lot of pull requests. um If you are one of the 17 people that are watching the projects, I'm sorry because it created roughly 100 pull requests for updating a dependency, but that's basically what it does. So it's fine.

G

We have a lot of prs that we need to look into. For instance, upgrading go open. Api string format will break our build apparently, and that's also one of the nice things now. So we get one pull request for each update and we immediately see. Oh this one breaks our build this one doesn't so that's a nice overview and we immediately know if we need to pay attention to this upgrade.

G

Regarding code changes.

G

All right, yeah, I don't want to go into details into the aprs. You can look at them up. If you want to I've. Also added prs for captain code trip repositories. They still need some reviews, but uh that's a feature that we will have in in all our uh repositories, so we are always up to date with third-party dependencies and actually, if we release a new version of go utils, you will also get a notification. Why I depend about that? There is a new version of goyotil, so we can use this on our own.

G

Any questions regarding this feature.

G

G

Okay, if not I'll jump to the next one in our go utils library and in our cli we had the xzip io resolver, essentially back with captain 0.6, we had some code and I I just looked up the wrong one here or no, it's this one yeah. We had some code that uh just checked if the domain that we were trying to access for accessing the captain api, for instance, contains zip io and if it did, we just applied a regex and extracted the ip address.

G

The reason for that was that cpu was unreliable in a lot of cases and it caused problems when installing captain it caused problems for workshops and even for our integration tests. um Nowadays, we don't use xipio anymore. We focus on nipio, which is a lot more stable, and actually we don't even need nip io anymore, because we have now so many other install options where we just use the kubernetes load balancer setup, and then we just use an ip address to directly access the captain api, so yeah.

G

This code essentially was not needed anymore and I have removed it because yeah, it's old and not needed.

D

Any okay but yeah, but still if somebody uses xepio and xepo, is in the moment unreliable. He will get an error.

G

But we don't recommend it. So it's like saying we we're fixing dns for the world and we're starting with zip io, uh but that's not our issue. To be honest, that's an issue of clipboard on the end user.

G

D

G

Windows bugs or mac os or linux, bugs in our code.

D

I agree yeah, but we had lots of users who were using zip io.

G

Because we recommended it, it was our default setting in zero six and before, and I I actually double checked. We use nipio in the tutorials and I've never seen an issue with nipio and in fact we didn't have the code for nibbayo, so it should be fine, um yeah next thing. Regarding the dependable updates, I did some manual cleanups of some dependencies because we didn't want to have each and every single pull request that needed to be merged.

G

So we did some manual cleanups and I had some dependencies that I didn't know where they're coming from so I double checked, and I I found that it was auto injected by um the ide and I've reverted that change.

E

How long will it take to clean up all the new pull requests, how many of them introduce broken changes.

G

um It's a good question because some of these open api prs introduce some breaking changes that I need to look into, but I think it's one or two functions that we need to look into and then we can do a mass update once we fix them. But I will write an issue for that and I will not do uh the pull requests one by one, because that would take too long by.

E

The way those are hopefully just minor changes.

G

Yeah, I think the one that are a little bit um worse to look into are these, for instance, upgrading the helm, um the helm, library that we're using or uh cube ctl or kubernetes api. So these are also not necessarily libraries that we need to update to the very latest version.

G

So we can do incremental updates here, but if we do the full update it breaks but again we'll need to look into it and if we don't do it right now, we can still do it in one month for two months: it's not mega urgent, but it's nice to know that we have something to do.

E

C

E

Do it now, can we somehow uh put those pull requests further down? Also, is it possible you.

G

Can filter, if you really want to uh so, for instance, you can filter for user dependebot.

G

And then put a minus in front of it that will remove all the prs so minus order app dependable. I can share the link with you, uh but we will close the prs if we don't update them at some point.

G

Okay, next one is a pull request for captain's bridge. A minor bug fix essentially um I'll, actually just double check the issue yeah if we send a configuration change or nowadays with captain 08, an artifact delivery event containing a configuration change payload without an image bridge was well crashing and printing. This error- and this was enough actually for me to look into it. The get short image name um basically had it to be adapted that it works with an empty, empty values, object it's nothing too, special to show, but it's been fixed.

G

Another thing that I looked into was code coverage because we had a problem with code coverage not reporting properly. I converted our pipeline to run the code coverage for every test if the test is executed so that that was our problem.

G

Actually, we are on a mono wrapper and we only execute the test under certain conditions, because if we would execute the test every time for every sub directory in this repository, then uh yeah readme change would actually take six seven minutes in the continuous integration system, which is not what we want, but that introduced the problem with code coverage.

G

So we do now send the coverage report after the tests are executed for each test and we flag the tests with the name of the module or of the sub-directory we test, for instance, here we're testing api here, we're testing the openshift route service, checkmeter service and so on, and this in addition, with a carry forward flag.

G

It's a setting essentially leads to. I need to show you on an up-to-date pull request. Actually, uh let's just take this one from one of our contributors, so uh this, in addition with the carry forward flag, now leads to prs that carry the last coverage from the last commit and only recalculate the coverage for the commit that this is about.

G

So, in this case, the contributor has updated the tests for the cli or has introduced some code for a cli, maybe with some tests and the coverage for the cli has slightly changed, but all the others are unchanged. Don't know why distributor service is having an issue here. I need to look into that, but most likely the flag is is wrong.

G

Any questions on that cool um yeah. There are some smaller changes for github actions and integration test workflows like caching, some more debug output and the possibility to draft releases which enhances our releasing process a lot. uh There's nothing really special to show or talk about here. So I'll just continue to one thing: that's a little bit more involved, uh that's our mini shift setup for integration tests and, as band has troubleshooted and posted here.

G

The integration tests for mini shift fail, and I think, with the help of florian, we found out that mini shift or actually the dynatrace sli service on mini shift is not able to access the yeah there. We go to access the dynatrace environment that we're using for monitoring, and I also did some more troubleshooting and I found out that, apparently that on mini shift, regardless of what I do like setting a name server or I think I did something else here- uh yeah setting a name server and other things. Regardless of what I do.

G

I cannot access a website or anything that's outside of the cluster from a mini shift cluster. If I use an alpine based image. So we've seen this before this is actually known issue that we've posted here on the troubleshooting guide for 0.7, where fully qualified domain names cannot be reached from within the cluster.

G

It's a combination of alpine based containers and the fact that you might have a search domain set that doesn't respond with a nx domain, but responds with an error when querying for a certain domain and um yeah, I mean that that would be the domain that is set on our on our mini shift container and we cannot influence it because this is predefined by github actions or actually, in the background, it's running azure.

G

So right now I had to close this issue with a won't fix, because we cannot fix this other than disabling the integration test for minishift.

G

But why am I talking in that level of detail to you? Because maybe someone of you has an idea how we can fix this problem or has seen a fix for this problem other than upgrading or changing all our containers from alpine based containers, to let's say debian or ubuntu-based containers or even going for scratch.

G

G

Okay, if you don't have anything to add or if you want to comment on the issue- maybe uh that's fine, um maybe just a quick look into the round. This is okay, that we disable the integration test for mini shift, then, because right now it will just keep failing until the end of everything.

B

I think it's okay to disable them now.

D

And um just doing this, this is a like very not for the mini shift. Would this be an option.

G

Yeah, I would disable it for mini shift exclusively. All the others will still work so for google. It will still work for k3, as it will still work. It's it's really just disabling the quality gates. Integration tests on mini shift.

D

But do we have to disable the complete test or just this sli query.

G

I guess we could um disable one of the steps, but that would be really ugly because that means I have to add into our code for integration test. I would have to add an if platform.

C

G

Shift or otherwise, I could just say uh to my integration test runner, which is github actions. If we are on mini shift, we will skip this test.

D

D

Anymore, this was never working before so.

G

All right, travis ci, it worked on github.

D

G

Good, so not not everything with travis is bad, uh just the fact that they killed open source projects but never never mind. um Okay. Last but not least, all right, I feel christian sorry to interrupt. We.

A

Only have two more minutes left. uh I have a heart stop because I have to jump to another meeting. I will give the host to uh to you christian, so you can still present this or we can.

G

Move it to the next meeting as you as you want all right. It's just a quick word: uh I'm working on multi architecture, support for a cli only, uh for instance, for arm based, builds, the pr is open for review and the end result. The artifact will be that we have more cpu architectures, supported with the next captain release for the cli, not for the docker images. Docker image is a different story, just making this clear. Okay, that's it thanks. Okay,.

A

Thanks so much for sharing this also christian, uh I think um if we have some cli or docker image specialists here, that also want to uh change the architecture or um make a broader variety available. Please let us.

C

A

uh I just want to thank um especially uh the new contributors to the captain project, uh thanks so much for your contributions, uh so you now most know most of the captain team. A couple of folks from the uh core team uh have not been able to join today. So um please also join next week on thursday um to to also uh get to know the others and we're looking forward to more contributions.

A

um So yeah thanks a lot everyone. uh If there are more questions, please continue the discussion, but uh unfortunately I have to leave.

C

And have a nice.

D

Evening, everyone.

D

A

D

B

D