keptn Community, 25 Jun 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keptn Community Meeting - June 25th, 2020

Description

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

Keptn accepted for CNCF Sandbox 🎉
Good first issues: https://goodfirstissue.dev/ or start here
Keptn mailing list / google group: https://groups.google.com/forum/#!forum/keptn
Noteworthy Issues & PRs (Review)
For each PR, please tell us if you updated the docs, or in case it was not necessary please say so
Ermin:
Display service out-of-sync info and allow to approve/decline deployment
https://github.com/keptn/keptn/pull/1877
Andreas:

Christian K.:
KEP 18: https://github.com/keptn/enhancement-proposals/pull/18
[Discussion]
Todo: Take a look at redhat service mesh
Todo: Follow-up issue on keptn install --platform=...
Manifest Installation:
https://github.com/keptn/keptn.github.io/pull/570
https://deploy-preview-570--keptn.netlify.app/docs/0.6.0/reference/manifest_installation/#access-the-keptn-api
Christian H. (excused):
https://github.com/keptn/keptn/issues/1843 in progress
https://github.com/keptn/keptn/issues/1781 PR created
Zohaib: https://github.com/keptn/keptn/pull/1922
Roadmap → KEP / SPEC / Issues:
KEP 18
Work in progress:
https://github.com/keptn/keptn/projects/29
Backlog Refinement / Next Sprint:
see https://github.com/issues?q=is%3Aopen+is%3Aissue+user%3Akeptn+user%3Akeptn-contrib+label%3A%22next-sprint%22
Planning Board
https://github.com/keptn/keptn/projects/29

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 and welcome everyone to another captain community meeting today, June 25th. We already have a couple of folks here in the meeting thanks all for joining and yeah. First part, it's a little bit of that's a small celebration. Again. Last week we awarded our first captain community rock star, and this week we are happy to share with you that a captain has been accepted for inclusion into the CNC F sandbox community or it's a sandbox will.

B

Be listed as a.

A

Sandbox project, so on this link you can take a look at the process on the PR. It's not yet listed in the landscape on CN CF. It was just accepted for inclusion.

B

A

Think on June 23rd, so on two days ago, which is which is awesome, which is great. We are very happy to see this and yeah. If you want to support us on this journey. Please follow us on the different social media channels, but also make sure to star us on github through its github.com, slash, captain slash captain and if you can give us a star there, we we really appreciate that if you want to contribute to captain, that's also really easy.

A

I would say- and we've already got the first contributors by our good first issues, thanks Christian for always providing good first issues and we are listed on good first issue, dot death, but also, if you click here, it will bring you directly to our github repository where you can start captain, but also where you can start with a good first issue, and those issues are usually the ones that are kind of. They don't need a lot of knowledge about how captain is working technically in detail, but you can.

A

You will have most of the description directly even issue anything. It started really easily.

A

If you have some questions around an issue or just on the project itself, we have our various communication channels, like slack like Twitter, but we also have since two or three weeks we have our mailing list, which is a Google crew, and you can also find us here, and you can just yeah basically most of ask your questions here and all the the captain folks are here, but also it's already has a great community so feel free to also make use of the Google Groups to reach out to the captain community.

A

With this I'm already handing over to Aaron he's the first on the list to present and discuss his PR. That he's like this. Let me stop.

B

My sharing, thank you very much yeah today, I am the first one to start presenting my pull request. It's about the assisted delivery use case in the bridge. So now we displace service out of thinking for and allowed to approve or decline a deployment in the bridge. That means, as described here in this pull request and I. Think I already this present. It's part of this feature, so this plane that the service is out of sync, but now we also yeah.

B

Of course we took care of some empty state views or if there is no service deployed yet or no no service on board it yet and yeah. The last thing, when there are some open approvals that haven't been approved or declined, yet we also show them now on the right side in a bridge and you can approve or decline that one from the bridge, of course, I have a short life demo.

B

So when, as you can see here, I have a project sock shop with two services, the cars to be and the cart service and I have from a production stage. I have manual approval configure, so the card service was not deployed to the production stage. Yet, as you can see, in death, I have cards in version, 0, 11, 3 and also in staging, but production is 2 versions behind and I can see those two deployments and, of course, I can decline.

B

A deployment, as you can see and I will show it in the services tab again. It will then show that this one was declined and I think this is a bug because I got the same event twice and I can of course, also approve a deployment, as in this case, for example, this 0 11 3 version, and now the deployment will run. So this is this second configuration change which will trigger a configuration change and deploy the new version, and a few seconds later it should show that there is 0 11 3 deployed on production.

B

A

Know that demo gods are not on your site when you have the applicative at this, you.

A

Timing is you actually.

B

I haven't been able to reproduce this issue on my machine yet, but now I obviously found a use case where I can reproduce it and.

A

B

Think that there is either something broken still because you should a deployment. It's not does not take that long, but yeah oh yeah, there it is, you saw it. It jumped to 0 11 3. So there is that was deployed yeah, as you can see right now today, ad so yeah, and here we can see, of course, in the production that was very fine.

B

Besides those dublicate events, how.

A

Do you find out that something is deployed you're still carrying the event stream, or is that happening? Why are the materialist view that we have no.

B

Currently, I just check for the deployment finished event, which also provides me to link to the deployment so.

B

Yeah, that's not yet using materialized view, information, ok,.

A

But do you know if the information is there or is it because I think the deployment finished event I can.

B

A

Is quite early, that is just the cannery stage and it's not deployed on primary. Yet.

B

Yeah I'm not sure if I have that information in the materialist yet I.

A

B

For example, I can fetch all the latest events by type, so I could fetch the latest deployment finished event, but the latest the very latest deployment finished event could also be deployment finished on a death stage. Ok,.

C

Yeah, that's a great point of you bringing up so that interrupts, but should we think about introducing very dated deployment like meaning something was deployed, and then we run at least a set of health checks to make sure that it's really deployed? Do we assume that deployment finish is really only st., but when it's evaluated that is up and running yeah.

A

I'm, not sure I think it's out of scope for what are they developing right now, but it's certainly for the next version when we work on a ship yet control our workflow engine etc. So for that, that's certainly something to think about anyway. Let's follow up on this issue, maybe on github or I'll, create a follow up issue. In fact where we can discuss the what are the API? Has the information or needs it, and maybe then Andy?

A

You can also bring your point into it, because then we also need to consider the difference between a version that's being tested in the version that has been evaluated successfully. Yeah I think both is needed.

B

Yeah yeah, thank you very much and actually I'm handing over to Christian. Thank.

A

You so let me share my screen. There we go so today. Actually last week already have shown you a little bit about my enhancement proposal number 18, that I'm working on this is about simplifying our installation method, to not install sto and to not install nginx ingress, basically get rid of third-party dependencies.

A

The reasoning behind this I'm just gonna go into the nicer view here. The reasoning behind this, basically, is that many users have already told us that they have East you installed and they have been using the experimental flag for reusing the ingress anyway, that we've provided here in the docs. So like the ingress, install option, reuse has been used quite often apparently hodo with marketed as experimental. Most people have treated it as production-ready.

A

To be honest, it did work so good, but that basically tells us people have east you installed anyway, and they don't want us to install history and in fact it goes as far as some people really have a policy that they're not being able to install SEO and they need to work with whatever they get yeah and then this also creates some problems, and another problem is that: is you complicated to use so when you don't need a steal, for instance, when you just want to access API and bridge, and everything else is not something you want to be exposed, then you don't need yest, you and you can just use the kubernetes load, balancer or node port, and that's something that most people that know cube anita's can work with issue instead.

A

Well, if you haven't touched this, you then don't do it just use load balancer at North Port. You don't need a service mesh for simple tasks. So, there's a couple of more reasons that I don't want to go into, but basically I've written, that down in the cap and what I want to show you today and what I would ask you for comments either now or afterwards? Why I slack email or on the enhancement proposal itself is the installation what it would look like for the user.

A

So we don't need to go through technical details, but I think this is a good starting point.

A

Let's talk about what does it feel like for the user to install captain, assuming that this enhancement proposal gets implemented, so the full installation, I'm saying Google communities engine here, because I know it works with what we're doing might work on different communities providers like Amazon and asure, but it might work slightly different, but the full installation is something that is a little bit more involved, obviously, because our full installation, if you want to use Bluegreen deployments, requires SEO and we're linking and that's what we will do in the docs we're linking to the docs of easy.

A

Oh there's a nice getting started guide on how to install SEO in a demo profile. So it's really not something I would recommend for production, but this is what you can use. If you don't have issue installed and you don't want to read like through all the docs, then just use this to install SEO, and after that you would install captain. Why are the typical captain install command, but you can select a captain API service type right now here and say: I want to automatically expose the captain API why a load balancer.

A

So this is one of the options that we will introduce with this cap. What that does is obviously it installs captain all the dependencies or all the services that you need to run it, but there is actually no third party dependencies that come, except for the ones that are really core or central dependencies like the nuts messaging queue and mom. Would it be for storing things but other than that everything is just pure captain code and once that is installed, it exposes the API.

A

Why a load balancer- and this is where Chiqui comes into play- Jiki. He provides you a very nice and convenient load, balancing service similar to I'd, say AWS and assure but it's important to know that it that the cloud provider needs to provide a load balancer and not captain anyway question. You know if someone forgets the option.

A

Would it be possible to expose it afterwards, like we do with the bridge? Yes, but that's also something I've written it down above, but I have an example below here as well. The idea here is that we don't want to mess around with that too much so on, captain install the Installer has higher privileges. That was something of the some of the things that have been explained the last week, I think so.

A

We've we've been playing around with privileges and we've stripped as many privileges as possible, so modifying a service of kubernetes and setting it to load. Balancer requires some privileges like right privileges on the cluster that we don't want to give away. Just like that, so we would like to dedicate that to the user so for the Installer. Ok, the Installer needs to do some things.

A

Obviously, but if you want to mess around with this afterwards, we would like to point you to the cubes ETL patch command, where you can just patch a service like in this case bridge and adapt us back to type load, balancer or node port or cluster IP. I have examples in the dock that you can look at anyway.

A

Once the installation is complete, it is still possible for the CLI to automatically detect where the captain API is running and to automatically configure the captain CLI locally, with the IP address of the captain API so, for instance, HTTP 1, 2, 3 4. Please note this is important here, HTTP not secure.

A

So this example here is really just a demo environment, and if you want SSL support, you need to deal with it by yourself, but that's the recommended way anyway, because why would you trust captain to generate an SSL certificate when there has been plenty of software written? That? Does that much better than captain, for instance, Stewart manager for p1 it is. There are tutorials out there to get let's encrypt running with any ingress gateway, also or alike, and it just works so yeah. Let's assume the authentication has worked. You don't need to deal with bridge.

A

The next step for you in the full installation is to take a look at east you again so east you will have a public or will have an IP address and the port that you need to gather why this command by the way. This was something that is in our installation method in our installation script. It has done exactly the same.

A

It uses this command to get the IP address and port of the East you English gateway, and you need to note it down and after that you need to create a config map where you store your hostname. So this is the equivalent of our previous captain configure domain. Basically, you store a hostname in a config map and in this case it's like five, six, seven, eight dot me piyo, but again, that's the demo setup for production traffic.

A

You would have your own domain or, if you're, using AWS, you would have your own domain anyway for like elastic load, balancing and stuff.

A

I'm aware this is a lot of information that's being processed, but in the end it's like one, two three, maybe so in the best case, it's one two, three four commands you need to execute. Instead of just having captain install and in the worst case, you need to do some troubleshooting and you learn some things about Cuban leaders.

A

That is the full installation. Obviously, now I want to jump ahead and also show you a little bit about the control, plane, installation or the quality gates, installation what it used to be. I'm. Sorry can I just jump in here, because we have questions from the community and thanks for using the Q&A feature with the questions, and the question is: if we also support a redhead service mesh which is which uses is do under the covers.

A

Is it something that we kind of also extract from? Or what would your.

A

It's it's hard to say out of the box. One thing that I know is that we have a hard dependency on the namespace East your system at the moment, which is something we can work on. So if it is using East EO under the hood, then it should be possible, but it might require some additional work from our side.

A

Nevertheless, that would be the full installation for Bluegreen deployments if you don't need the full installation with Bluegreen deployments, but the control, plane, installation or a quality gates. Installation is enough, then. The answer for me is a clear.

A

Yes, because the control plane, installation we've been able to trim it down to not have any hard requirements on the ingress gateway or or service measures, so yeah, essentially the control plane, installation I'm, showing you variant two here we have variant one above variant: two is not using a load balancer to expose every single piece of the of captain, but it's using ingress objects, which is something that has been introduced, I think with communities 114.

A

So in this variant we install captain, we don't have a pre requirement right now and unless you want to use like a specific English controller, obviously have to install it yourself, but for captain itself you can just install it like this and after that installation you're, basically done with the control plane. The only caveat here is: we've exposed the captain API using service type cluster IP, so you cannot reach it from the outside and to reach it right now.

A

You would have to do with cube City Airport forward, which is definitely not very convenient, and this is where the ingress controller, like east, your nginx or traffic and I, know I, have a typo here. I have that fixed in a local committee.

A

I still need to push it, but the local sorry that a ingress controller is where you can work with an ingress object right now, and this is what an ingress object basically would look like the most important part is you need to specify a host and you need to tell the ingress controller where this host is routing traffic to, in this case, it's running it to API gateway, port 80 and after you have applied that you can do basically anything that you could have done before with captain 0.6, you can authenticate your captain.

A

Seal I just provide the endpoint with the host name that you've specified here and provide you API token, either directly, while cube cut' will get secret or if you have it stored in a variable or somewhere else. Obviously, just replace it here. Only one thing to note: if you use HTTP here, the seal I forces you to select it here as well. Why the scheme, flag, I, think it's a security measure I'm, not sure if it will keep it or not, but for now this is what it requires.

A

All right so for a control, plane, installation, in my opinion, a lot of things get easier. Obviously it could have used load balancer here as well and skip the ingress gateway completely, but nevertheless, either way you choose I think it gives you a lot more flexibility and, most importantly, it reduces a lot of dependency that are now being shipped with captain, so I think that's a huge plus and I would hope that this captain enhancement proposal gets merged and implemented soon.

A

I know that we got some very good feedback from the community, so Sergio he's not joining the meeting today as far as to see, but he is very much in favor of getting like extracting the actual ingress, more and also from the community. There's the feedback that the redhead services match is something we should definitely take a look at because it's also being pushed very hard from redheads or something probably, we have to take a look thanks for the link document.

A

C

Quick question the parameter platform: does this parameter still do anything? Would this be an option to say platform ets chicky aks is maybe changing the default, so it doesn't just make sense at all anymore.

A

It's no longer a requirement in the sense of you wouldn't need it for Google. You wouldn't need it for assur I! Think for AWS I! Don't know you will definitely need it for openshift, because there are some differences with openshift, but parameter is getting less and less important. So I can think that the installation on Google cloud could also work by just saying captain install platform, pew Benitez- it's not part of my cap, although but I- think it's a good follow up issue that we should consider to maybe simplify that as well.

A

Does that answer your question? Yeah.

C

A

C

Think also, the all the changes I mean, as you know, that Sergio he created the captain in a box and then it almost created the keys. So that basically just means that obviously, these scripts, that they are providing. They just need to take care of installing on a fresh kubernetes cluster in ingress and maybe.

A

C

Over but it's it's the much better option because they are now under control on whatever install so that they can even extend these demo packages or these installation packages with any type of ingress and it had put load balancer. Absolutely.

A

I tried it actually. Actually, let me put this pull request up. First, so captain github io repositories or a documentation I have a pull request and I'm jumping right into the preview.

A

What I've done is this is the captain quality gates, installation using manifests 40.6 and with what I just told you like getting rid of SEO and nginx I took the time to also check, if my assumption, if my enhancement proposal would work with captain 0.6 and the answer is yes, it would work and it does work and what I did is I have enhanced the access they kept an API section here by what would it look like to use the kubernetes service type note port?

A

Most of the commands here are obviously directly written by cube CTL, so they have nothing to do with captain. It's really just exposing captain. Why are built-in community services and also using a Q&A? This ingress object? I've described it here as an example that that's work with captain 0.6, so basically that should also help people like search, oh, that are building great pieces of software around captain 0.6 right now and the the funny part. What an interesting part I want to mention here with the community's ingress object.

A

I have followed a digital ocean tutorial about nginx, ingress and let's encrypt, and it took me like 20 minutes to get the captain API exposed wire. Let's encrypt SSL certificate with that tutorial, so it shows us the great benefit of just using built-in kubernetes options. Instead of doing it ourselves.

A

Right any more questions or any more feedback on that cap or on exposing things and services using completists objects.

A

Also, nothing from the community now, okay, right now, no I know that a couple of folks they are excused for today they have other duties, could not join. I was just talking to two Christian Haeckel money's, not a call. Unfortunately, he cannot join so my question would be they kept a teen, it's about two getting like implemented quite soon, or is there still option for for feedback in from the community or how should we basically?

A

So we would like to get some feedback, but we would also like to get started on implementing next week. So I'd say by the next community meeting. There should be issues written down and will be presented in a community meeting, and maybe some of them have already been started implementing. So we would like to have feedback what we would like to have it within, let's say by Tuesday, if there is any, if it's positive feedback. Obviously there is no implications, I guess, but if you want something change, then that should be quite quickly.

A

Nevertheless, I believe the feedback that we will get will not have a huge impact on the actual implementation, because the implementation is is very, very. It reduces it to a very, very basic set of things that captain provides, and it requires us to document it better. Basically, but like I would say everyone that is watching the recording. Probably please don't wait for the next community meeting to discuss kept 18 again, but please comment on the pull request that Christian already opened and have the discussion offline there, because we won't.

A

Probably we won't wait for the next community meeting for for a final discussion, but we will discuss it. Offline yeah also feel free to get in touch with us like all right, so I guess, Christina come on and swipe are not presenting today. So.

B

A

Unfortunately, they can't join. Maybe we will get swipes issue quickly discussed because it has, it has been done and it should just be moved. So, basically, we have asked the community on increasing the MongoDB data, store volume and swap has thankfully done that for us in the Installer manifest and in the MongoDB datastore manifest. So now, we've kept in 0.7, that's being released soon, there's a little bit more storage available and that's a good thing, because we are also using more storage now, because we have a lot of improvements in our mom.

A

Would it be data store that will influence that, let's say like this, so that one is done kept 18 I have just presented and I'll jump over to our planning port or actually to a work-in-progress port.

A

There's plenty of issues on ready for ready for review. Florian cannot join us today, so we'll discuss them next time. I mean I. Think yours I have not been merged yet so we can leave it here as well. Thank you. I'm interrupts yeah.

A

Then you harness is also on vacation. He will present or work on this issue when he's back. The same is probably true for this one. That's assigned to him.

A

Do we know if Christian has any progress on the retrieve metadata issue? It's the 1843 is.

B

A

So I guess it stays within progress. That's what we know and for the issue 1781. He has created a PR.

A

Was that in a different repository, maybe well, let me check t that's.

C

A

He told me: oh it's already closed, so oh, that's this one! Okay, it has been merged, it has not been presented. Did we not have it on the port or did I overlook it? No, it is I did overlook it, okay, yeah! So if you can't join and present it next week, that would be nice Christian. If you, if you listening to this afterwards, please join us again next week.

A

All right introducing API version for a mediation file. Ondrea's is working on that, but has or hasn't started, but we'll work on that we've presented it last week, I think, and god we are, is on vacation right now, as well as plenty of people on vacation now, and he will be working on the version. Checker interrupts user when there is an error issue. So basically, when you are behind the firewall or you don't have an internet connection, the captain, CLI version check seems to be causing some trouble.

A

Then the next one in the backlog. So let's jump actually to the planning and backlog, refinement or backlog. Yeah planning that atcha meter merge where we will merge the captain. Kahn trip to meet extended service of Andy karna is in progress. I believe Florian said today that he is doing that so I'm moving that here.

A

So then the next one is an issue for captain's bridge I'm, not sure. If your harness has presented this last week, I, don't think so. He's presented it to me I think anyway, the idea is I jump into the issue. We have our heat map for our evaluation done event, and this heat map itself is in Captain bridge in in the screen with all the traces and the wish now is that we get a separate view that just displays the same heat map.

A

So the same he'd met component that we already have and just displace this and nothing else, and we get a deep link into that. It's also the definition of done so provide a separate view, provide a deep link, and this should help people that just want to display the heat map, but not full captain's bridge. So, for instance, if you want to use an iframe, you basically could do that. However, we've we've written it here out of scope for this issue. Right now is authentication because captain bridge will have basic authentication enabled when it's public.

A

So that's something you need to deal with when you use an iframe also, if there is any course related issues, cross-origin requests, that's something that the end user needs deal with: it, not captain and anything ready to iframe security issues, I've written it down here twice and purpose, because I frames generate a lot of issues. I can't stress this out is not part of this issue, so it's really just providing a separate view and providing a deep link of that view.

A

Oh, and to make it easier to to guess the deep link, we will provide a button here in the normal evaluation done event screen that will just open this view in a new browser. Tab.

A

So I guess we don't have many participants, I think I mean would have taken this issue if he's here, unless anyone else wants it.

A

All right, um I'm, not sure what happened to I mean he's, got an unstable connection, but I know that he will be doing that. One.

A

All right, the next one is my cap actually I'm working on the cap, so I'll assign it to me, but I will need some feedback and an approval. I can't assign it to me. I will need some feedback and I will need an approval, but I'll contact the people that can give me that approval all right. Then the next one is an issue in captain examples.

A

This issue is about introducing the app community's I/o version, metadata label into our cards and cards, to be example, and if someone wants to do the extra work, also in the Unleashed server and the simple note service and the idea behind this I'll open the Q&A this documentation. For this is it's a recommended label and we've introduced some of them already, because it was easy to introduce the name and the part of, and the managed by label, because managed by would be captain in this case, and the part of would be the project.

A

Name and component would be database for the MongoDB and API or website for our cards micro service, but for version we didn't have a semantic version at that point, or actually we still don't have a semantic question, but we can reuse the version of the docker image. I think I've written that down here as well. Yeah use the docker image and with that we should be fine at the point. Why do we need this is actually for monitoring software? That is monitoring the kubernetes deployments at the community spots.

A

It can read the version field, not just the name, but also the version field and maybe introduce some additional intelligence e for monitoring data. So let's say you have, for instance, dynaTrace monitoring or also Prometheus monitoring, and you want to see why a certain service has a high CPU usage and you can investigate on a service directly. Why are Prometheus or dynaTrace? It will tell you, ok, it's version 1.3 and with version 1.2, you can look in the history.

A

You can see that the problem did not exist, so there is some more intelligence that you can work with. If you have that flag a question from my side arm, that's so the issue is not that captain will automatically create these tags when you're onboard a service, but actually you have to provide it in the help chart. Oh yes, yeah.

A

They have to be provided in a ham chart right now, there's no way for captain to add these automatically, but here in the help chart deployment yama, for instance, we already have some of them and the idea is you need to introduce the version flag here, but it needs to be obviously variable, so that might create some troubles. I'm, not sure. Basically, you have the image name is here and with the image name, you get the tag, I'm, not sure if you can actually extract it from here.

A

So maybe this is more of a research and tried out issue rather than make it perfect issue, but yeah.

A

Okay, are we planning to kind of give a warning when the user on board the service with the help chart that does not have these recommended labels, not at the moment.

A

It's a good point, although I'll get that on my list actually, can you actually keep that sweet the other way around? Can you comment on the issue.

A

Any examples repository anyway is anyone, that's here up to do that issue.

C

I'll, try it sandy. Okay,.

A

That's nice I think it.

C

Also helps me to test one a day. I mean you know in dynaTrace were coming more version aware, and it also allows me to then test at new capability. Oh that's.

A

Nice yeah cool, so then the next issue is the documentation issue. Basically, in the last couple of weeks, we've changed the behavior of captain configure bridge to force the user to use the user and password flag. Actually, we've introduced two flags, and then we force the user to use them, because, right now, with the latest changes, I mean dip.

A

Captains which uses the public API of captain, which requires authentication and therefore captain bridge requires authentication, and we need to work around this a little bit by forcing the user to have the user and password flag. Otherwise, captain bridge would just expose the API and that's not what we want. So it's basically just the documentation issue for Captain zero-point-seven, look at any place where captain configure bridges showing up and and edits the documentation to use the user and password flag.

A

Anyone up for that one.

A

Because if not florian has told me that he wants to do this, one.

A

Alright, the next one is also something actually quite nice, in my opinion. So right now, when you do captain install the issue, is that it's not quite clear what is being installed unless you specify the use case, quality gates option.

A

However, this option is kind of misleading, with captain 0.7 that we're working on right now, because you cannot just install quality gates, you will also install maybe the self-healing use case or any other use case that we are providing and therefore we want to introduce the notion of captain control plane.

A

So some of you might have heard that already, but we've never really used that, and we also will will try to not use it too much in the future until we have a clear definition of what the control plane actually is, but think about the control plane as the minimum viable product. That captain offers you process events and to process certain things like a quality gate or remediation scenario. So a remediation scenario might involve external services.

A

They are not part of the control plane, but the fact that we can process remediation events or self-healing events or whatever they're called, is something that is part of the control plane and at the same time there might then be an execution plane later that connects to the control plane externally. This is not part of the issue right now. This is just to give it a bigger picture of the issue.

A

This is something we want to achieve and for that we should revert the install options, so we should no longer say captain install use case quality gates. Instead, we should say captain install and this installs as the minimum viable product that we can use and when we do a captain install use case, for instance, continuous delivery. This is when we do the full installation where we also install the helm service. We install what else. Do we install now other than helm service I? Think it's not much I, think it's helm service and Jumeirah service.

A

They will come with this full installation and obviously, with this full installation comes the requirement of SEO, which we've talked a lot team kept 18. So I'll just don't talk about it anymore, but so, basically the idea is revert.

A

The way captain install works, captain install just installs the control plane, the minimum viable product kept me install use, cos, quality gates will stay available as an option, but basically isn't Elias for captain install and we will introduce captain install use case, continuous delivery, which will install our continuous delivery components like helm, service and chimney the service.

A

Are there any questions regarding this issue, or is anyone already saying that they want to do it? What they want to do it.

A

Right, if not I, think that's also one of the issues that Florian has communicated to me that he wants to do.

A

But I'm, looking forward to this issue being implemented that it's a it's actually a small issue. But it has such a huge impact and.

A

This one I can also move here because I've assigned it to Andy. Here there we go the next one is basically reported by one of our community members and he had an issue when using captain install and not getting enough. Debug output, a troubleshooting output and I was able to verify that welcome. Back I mean he was.

A

He was not able to verify what the cause of the problem is, but I was able to verify that there is a problem, and sometimes you just end up not being able to install captain, because you lose connection to your cluster or something else is wrong with your crew leaders cluster. So it's not necessarily a captain problem, but captain should report that there is a problem and right now you end up in a waiting for install or pod to be started, and this is just there for 10 minutes and you don't get any output.

A

In addition, if the connection to the communities cluster is not working or whatever, you should also get some debug output, and for that I've written the definition of done here. If the user specifies the verbose flag, we should lock if a connection to the community's API can be established or if it fails, it needs to be locked as well. We should lock every command we execute when talking to the community's API or using cubes ETL, because that also helps the user and troubleshooting.

A

What was the last command that was was executed, so in this case, waiting for installer pot to be started, I would like to see like okay, the CLI is try to detect if an install apart is running by using cube, CTL get ports or get chops, probably, but that's something that is very valuable to the user for troubleshooting and where appropriate, additional messages really should be locked for troubleshooting purposes. So when we touch this, we just look through all other parts of the installer in the CLI that we should adapt.

A

The logging Andy I think you wanted to say something.

C

Now, I'm, just making I'm just answering some questions from Chris who has been asking questions on the Q&A okay.

A

C

May have been unmuted and said something.

A

Look if all right anything for relevant for this issue for one of the past, ok, all right! So it's actually just an issue adapting to cui I'm, pretty sure if Christina Coleman would be here, he would be looking into that other than that is anyone willing to take it else. I think that's also one of the issues that Florian has told me that he would like to do.

A

Right I'll assign it to Florian for now I mean now that you're back I've assigned you that get heat map of evaluation done event, issue I. Hope! That's! Ok with you, because I think yeah.

B

That's cook yeah.

A

A

So, ok, we still have 10 minutes and we have three issues. I think that's a perfect match.

A

Actually that one 18:56 I can quickly talk about and then tell that florian is doing it anyway, because it works perfectly with the previous issue. It's actually linked together. So sometimes the captain installations just fails with this message, and it's not clear to the user why it has failed. That definitely is a reason or maybe there is no reason that it is failed, but there is something something is wrong.

A

It might be a timing issue or something like that that doesn't occur on normal clusters, but it occurs when you have a cluster with a very limited amount of CPU available. So, for instance, if you have a weak, laptop or a Raspberry, Pi or whatever any other micro computer, that has not enough power to run a communities cluster with many cores, we will run into that issue.

A

Sometimes in this case it's Travis CI, but the issue will exist in other platforms as well, and I've took some time with the initial analysis and found that there is like a line that says: if there is any error on SCD error or STD out, then we return, unkept and installation was unsuccessful. However, to look deeper into that, there is a method that parses the output and that sets the error occurred or install successful flag, and this is what we I believe.

A

We need to start and adding some more lot messages indicating what actually has happened and why it is happening. So it goes like I said hand-in-hand with the other issues, so I'll actually really just assign it to Floria, but, as you can see watching this video afterwards, probably with taking a lot of steps making sure the Installer is getting rock-solid now so we're hoping this is appreciated by the community and we will see the effect of it by getting less support issues. I think that's really very well investing time.

A

All right, then, if one more issue for Captain Bridge, we would also like to hear some comments. If there is people here that have an opinion basically right now, captain Bridge has the possibility to display a heat map or a chart for evaluation events, and if you click on chart, the representation right now is sometimes suboptimal, because it's split on the x-axis by date. Time, and if you have many revelation events happening like within the day, then this is fine if they're happening every day. Yeah.

A

But let's say you have a company, let's say 9:00 to 5:00 and maybe three deployments every day and you don't have deployments on Saturday and Sunday and maybe also not on Friday, because ya don't deploy on Friday. Then you have this gap, and this visual gap is really bad for a user and in our opinion it would look better if we can have the spaces removed and have like this distributed spacing equally distributed. Spacing between the cloud events for evolution done. Events.

A

Are there any comments on that which, which one is better or yeah.

C

The question is: what's going to be the label on the x-axis, where you see if ya, 47:11 and so on? That's.

A

That a perfect question, because we don't know we need it right now. We don't have increasing number right now and we don't have a label that tells us the number at least it's no forest label, so I think I mean what we discussed is just using the date and time and they are tilting the label. At the moment we.

B

Have two options we can provide the context or the time and the context has no information. So we.

A

B

To display the time the date at the time, and maybe maybe.

C

Maybe this would be a great because Cristiano mentioned, and maybe this would be a great way to introduce an optional label, something like a version number that you can pass in from the external tool like what I do over to champions. I pass in the Jenkins build number I was.

A

Just gonna say: I, don't think your version number is the right thing, but the build number real.

C

A

C

Then just say: if the build number is there, we use the build number. If it's not there, we use the time back.

B

To the time yeah.

C

B

Yeah, of course we can define a specific label and try to fetch that one yeah I think there will be.

C

A

Any objections calling it built and our or any better names.

A

Image: it's it's open for anyone to define.

B

A

A label so that the payload would be.

A

C

And then built number equals something you know exactly so yeah because most most of you know most of our users are triggering kept and from any external tool and they have some concept of a build or something like that. Exactly yeah yeah.

A

I mean I would leave it as a string. Obviously so, whatever.

B

You're. Writing there.

A

Is up to you, but we can define that as our best practice. We don't need to set anything in stone right now, but we can use that is build number okay, or is there any convention in the Jenkins community, maybe that they are using? Well, then Jenkins is to chop the.

C

Top ID but I think I was.

A

Actually, thinking about it, I'd, like ID number, is such a it.

B

A

It a chairman, touched a little bit because we.

B

A

Terry thing with numbers and ID is I, think more yeah. It's good.

C

Oh that's to build ID, let's to build ID, that's okay!.

A

All right anyone else for doing that issue then I.

B

Will take right.

A

Did you have any comment, I mean I, think I did interrupt. You know.

B

Know that it was just regarding to the built ID that nice.

A

Nice, okay, last but not least, there is a back reported by Thomas. He wanted to do to fix it, but he didn't have time to fix it. So we've moved the issue back into the backlog in captain zero point. Six point, two and probably also in our current master.

A

The helm service is not used using the environment variable configuration service, so we actually have that already now a manifest I think, but we are not using it within the helm service, which is not a problem at the moment, but it might create a problem afterwards when we at some point split the helm service away from the cluster. So this is more of a clean-up issue and should be, in my opinion, rather easy to do. But Thomas said he did find some issues. So I don't know.

A

Is anyone willing to take this.

A

Okay, if not I will leave it in a backlog. Maybe someone will contact or tell us next week that they are doing it, but for now I think we are done with our planning. Thank you, everybody all so Florian, if you're watching this afterwards. Thank you for telling me the issues you want to do and Kristin Eklund if you're watching this afterwards you're always welcome. If you want to take one of the issues that we've discussed today,.

A

Community meeting, if there are more questions from the community, just please type them or indicate that you have some more questions, I think a couple of them have been answered in the chat. It was around SMI providers if you can customize the sleep time in the dynaTrace SLI provider and basically the gist of the answer was no right. Now you cannot customize it, but there is no sleep time in general village.

A

Only sleep time if you're requesting metrics from the past now like time frame, which is exactly now, we have to sleep a little bit because planet race as a lot of provide a means to on these live before data processing and therefore captain waits to have all the right data.

A

So no more questions were coming in. So thanks everyone for joining this week- and uh yes, you all next week, bye, see ya, bye, bye,.

B