keptn Community, 27 Aug 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keptn Community Meeting - August 27th, 2020

Description

August 27, 2020
CNCF Slack Channel: Join here
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:
https://github.com/keptn/keptn/pull/2230
#2139 Improve color-coding for problem event
https://github.com/keptn/keptn/pull/2238
#2086 Show failed quality gates in environment screen
https://github.com/keptn/keptn/pull/2243
#2085 Show remediation workflow in environment screen
Andreas:
Upgrade from Keptn 0.7 to 0.7.1: https://github.com/keptn/keptn/issues/2234
Johannes:
Allow to set the namespace which is used for Keptn: https://github.com/keptn/keptn/issues/2195
Security requirements for CII: https://github.com/keptn/keptn/issues/2205
Christian:

Roadmap → KEP / SPEC / Issues
KEP 06: The next generation of Shipyard: https://github.com/keptn/enhancement-proposals/pull/6
Keptn CLI: keptn delete service xyz --project=abc
Work in progress:
https://github.com/keptn/keptn/projects/29
Backlog Refinement / Next Sprint:
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

B

All right, yeah, we are live uh hi. Everyone welcome to the to this community meeting um yeah. Unfortunately, we cannot use the the link that is provided online as we were zoomba bombed a couple of minutes ago, and this is not a reason why we jumped over to this zoom session.

B

All right, then, let's share my screen to see the agenda for today uh this time we are not that many folks, um because yeah it's holiday season and a couple of of of guys are enjoying theirs their time, uh yeah outside and they're, not in the office, but still, let's do this community meeting.

B

First of all, when you want to get started with with implementing an issue for captain, then feel free to check out the good first issues list.

B

Here you find a couple of issues that we classify as good first issues that are yeah to some extent very nicely framed and a good starting point for getting first hands-on experience with implementing captain issues.

B

Then there is also the captain mailing list. This is a google group. When you want to get notifications and and emails about captain and upcoming announcements, then please feel free to subscribe there and it's one email per week that will be sent out every monday, all right yeah.

B

Then I don't want to lose more time and we jump directly to the pr's and the issues that were implemented on this week and I will directly hand over to airmen who has to who wants to show us cool enhancements in the ui and yeah I mean the stages. Is yours.

C

Thank you very much, johannes. I'm going to present you three pull requests today, so the first one is titled. I improve color coding for problem event. So let's see what was that? What was the issue there?

C

We had the case that a problem resolved event was reported by dana trace service, but on the left side, where you see the root events, it was uh labeled as red and like a problem event, uh and that uh brought us into some discussion. So uh we decided to even um yeah problem open events, so, as it should happen, usually you get it first, the problem open and then the problem resolved event and in case that the problem is resolved.

C

We yeah decided to mark this event, so the root event as green again and the result is um that now, when you have a problem event as long as it's open, you see it as red as a red root event, and the stage where it occurred is red, but in case that it's resolved or you have an evaluation that is successful. So after, for example, remediation actions have been executed, then you see that the problem open event is green.

C

Again, because of this successful result, the second issue that I'm going, the second pull request, I'm going to show you is the show, filled quality gates in environment screen.

C

So we have the environment screen where you have the overview of the stages, and the idea is that with those two icons or actually with this icon here we um indicate that there was a quality gate broken on that stage. So when you click on that stage, then you see on the right side that yeah here one quality gate failed next to the service that failed. You see this same icon again. So when you expand it, you see the details.

C

In this case we can see quality gate of cards. 0 11 3 failed with a score of zero. So a very bad score, and in this case, because blue green deployment is used, a roll back to 0 11 1 was performed in case of direct deployment.

C

You would just see the first message here and you would see that the version 0 11 3 is deployed, but still you get the information, the quality gate failed and when you hover over this bubble here, you got also a pop-up with some details about this evaluation and one more poor request show remediation workflow environment screen. So this is again in the environment screen.

C

So when you have a problem event that occurred, we also want to display that in the environment screen, because it's stage specific and here you see, for example, the card service in production. There was a problem and in this case a failure rate increase on web service items controller. So this is the problem title that is reported in this case by dyna trace and then captain executed. Some remediation actions like toggle feature flag, enable promotion to off, but the evaluation here was not um yeah successful, so actually there should, or all remediation actions will be.

B

C

And after each remediation action we have some evaluation and if the evaluation is successful, then the problem disappears or in case that we get the problem resolved from dynatrace, for example. Again then also the problem here is resolved and just to show you a short demo of this one. I already explained the problem resolved in this case is marked as green, and we have here the problem that is currently open.

C

So when I go to my environment screen, I see my production stage with this cards service in red and when I what I just wanted to show you- and I think in this example now it's broken, because this is my local environment.

C

But here would you would see it in a pop-up, but I will show you with the other use case when we have a configuration change event that was broken. So in this case the evaluation on the staging stage is bad. Then I can see this same quality gate that is broken on my environment screen and when I expand the card service- and I get this information here, I can hover over this bubble and see also the result of this evaluation. So I see here uh the score and also the sli breakdown.

C

So I can see that here the response time p95 yeah, was very bad and that's why this evaluation failed um yeah, that's from my side. So if there are no questions, I'm handing over to andreas.

C

D

Thank you, yemen really nice features.

D

Then I would like to share my screen. So this week I've worked on upgrading captain 0.7 to 0.71. So you probably know 0.71 is around the corner. We will start releasing it next week and here, of course, we need some way to automatically upgrade from 0.7 0 to 0.71 and therefore I've implemented.

D

This issue here, which introduces a new cli command, which call is called captain upgrade so as soon as you have the new cli in version 0.7 one you can then write, captain upgrade and captain will then automatically upgrade all images on your cluster.

D

Of course, you can still also use um helm directly and please use therefore this image, which will be available next week and but the cli here allows you to to simply um make it more convenient to do this uh hidden upgrade.

D

In the background. Of course, there is so. This is the corresponding pull request here, and this is a little bit bigger, because I need to reflect especially the install command, because I was able to reuse lots of the install functionality for the upgrade tool, and here I would need a review.

D

Please and finally, I would like to show you our version json file, because here I also added now a section for captain directly, so we already had the cli we already had version info for the bridge, and now we have version info for the captain cluster and what you can see here is. um We are also listing all stable versions and, for example, a new, stable version will be 0.71, and then I additionally provide information which sila, which installations can be upgraded.

D

So there is an upgrade path from 0.70 to 0.71, and this, of course, is used in the cli for doing the upgrade mechanism.

D

So that's from my side any.

B

Questions um I haven't, I have a question to airmen in this regard. uh Actually, could we use it was just an idea? Could we use the version.json file that andreas showed us that we can inform the user that there is a possibility to upgrade? Let's assume uh user is running capped, node.7 and we release 0.71.

B

Can we you then check the version file to tell the user you are able to upgrade to 0.71.

C

The question is: how do we retrieve the cluster version and in case that we can provide the cluster version with an api call, then we could display this information also in the bridge yeah.

C

Currently, I only have the bridge version.

D

So the cluster version is already available.

C

Oh no yeah, of course. That's that's the version.

D

Let's double check this because I'm probably we also need some exactly this information, also for the bridge, upgradable versions.

C

Yeah, I can check that so I can. If I have, I have the api version provided. So I can that's the cluster version, let's say and in case there is an upgrade table version for that current version.

C

I can show the information that would be possible.

B

Cool all right yeah. uh It was just an idea, but yeah.

A

Of course, let's frame.

B

An issue around that and let's first.

A

Create an issue.

B

C

It maybe there again, but.

B

C

It would be possible, yeah, sounds cool.

B

B

Then I will take over and talk about the issues I was working on. The first one was actually an issue that I did not implement kudos to gabriel and and andreas who did the work on that issue, and I was just a guy who verified everything, and it is about the possibility that you can install captain in a specific namespace.

B

Let me go there to show you this in a real example. Well, here we explained what this new feature is all about, and what I did is I used the newest captain cli, which will be available with 0.71. Then I run the command captain install and then you have here the possibility to add the flag.

B

Namespace equals the namespace. You want to install captain, and in this case I used captain tiger yeah. I did that and then I executed all the tutorials for onboarding quality gates and the remediation actions, and they worked pretty well.

B

Everything works together, and this was just a verification step that we did not miss any piece in any part and so that we can are also sure that there is nothing breaking when we go um when we go and when we release the new version- and there was also an implication on the captain- generate support archive command, I verified and then I finally also uninstalled captain, and here you also have to specify the namespace again.

B

There were just a couple of tweaks necessary but yeah. Now it's everything working together.

B

This was the first one and the second one- and the second issue was working on was security related because jurgen is is working on increasing our page.

B

This cia patch has currently already a good value of about 95, but he wants to get it to a 100 and for the last five percent uh he needs a statement or clarification for three security questions, and I did a research or I scrolled through the source code and took a look at all the tokens and keys that captain is using and what I came up with uh with a conclusion that shows that captain is using four kind of tokens.

B

One is the api token. Then we have credentials for the bridge credentials for the mongodb, and there is also a session token for the websocket communication and then further down. I explained how we are generating those tokens and how we are storing them inside captain and based on this information, I then came to a conclusion that we can um or that that there are some tasks necessary to um fulfill the first requirement, and there are also some tasks required to fulfill the second and third requirement.

B

But this I want to discuss with yurin if we should yeah do really an action item or an implementation item on that or if this statement is already good enough to to get the five percent of the cii patch yeah. But this is something I want to um discuss with during going forward.

B

All right yep, this was my final task. I was working on this week, then christian is not practicing today. This means that we can jump over uh to the working items so.

D

I have a quick note to your first issue so because this now allows us to install captain multiple times. So, for example, you can have captain one captain, two cabin three and you can also use exactly this feature in order to split your data. For example, if you have a pre-broad environment and a product environment, and you would like to only see data in the bridge of your pre-prod environment, but not from the product environment, then you can simply install captain two times and you have a really nice split this.

D

This is an advanced use case, which can be really useful totally.

B

Agree- and this is now very, very convenient, very neat, because you can also think about not separating those uh for different environments, but you can also get a different get a team or a team can have its own captain and then just this team sees the data from this captain and another one sees the data from another captain and you, but but still you can install both on one kubernetes cluster.

B

This means that you just operate one cluster, but you can have multiple captains for multiple teams. I think this opens a lot of uh very cool use cases and also uh opens up possibilities for for supporting different organizational level uh um challenges.

B

That's true yeah and it's coming with uh with the next release and yeah we're excited about that.

B

Then I wanna skip the discussion about cap 0.6 as we did it the last couple of meetings, just a quick reminder when you want to get more informations about the new shipyard format. Please jump to this captain hartman proposal and check out what you can expect from captain from the shipyard version o2 and the next time we will see the implementation of the cli feature that allows us to delete the service yeah.

B

We want to ask florian to show us this feature, and this is the reason why we postponed it to next week and now, let's go over to the working items, and let me just move over to the uh to the ready for review column and, as you showed us, the upgrader, the captain cli command for captain deleting the service will be shown next time.

B

The same is true for that and that one here we have an issue for from florian, and here again then airmen you showed us that the bridge is is changing the color when the problem resolved comes in and you also showed us the quality gates in an environment screen.

B

I think this pull request is not merged yet not accepted yet, and there's still some work to do. Yeah.

C

Let's leave it in the review section. uh Actually I saw that there is some. There are some parts that are not working yet correctly. Okay, then, I will.

B

C

B

In progress, yeah, okay, all right perfect, then, what's in progress, um yeah the specification of the event, but that's uh yeah! That's something! That's something, and that's not um just that. We should not put pressure on that issue um as long as we have time.

B

Then would be nice to get feedback from christian on this issue, as he is working on getting the feature out of the promesa service to actually install promises. We don't want to do that anymore because it should be done manually and by the person that is operating promises.

B

And I'm still working on that one uh yeah, because our travis ci build is currently broken and I need to figure out the reason why it's broken and then I need to provide a a fix for that and airmen. Are you yeah? I'm.

C

Working on that.

B

All right and last time uh we assigned this issue to you, airmen yep, that's also actually working progress cool. Then, let's have a look uh into the backlog. Here we have the first uh issue that is also bridge related because, with the cool feature of getting the icons there that allow us to focus on the problems and the fail quality gates, it would be also very convenient to get these icons working as a filter.

B

This means, when you click on this, I I can you filter the the list on the right side, and the same is true for these two. These labels up there when you click on one of these, then the list of services will be shrinked and filtered.

B

Right yeah, because when I think about having a couple of services there, then it's really difficult to find that one that you're looking for and these filters will help a lot I mean. Can I sign yeah.

C

Of course, let's just I will take that one just wanted to say that I heard from environments where they have like 50 to 100 services.

B

Yeah, that's true, and then it's it's a challenge to look for the for the right one.

B

Then we have another issue that is about uh also convenience feature for working with captain, because what we would implement with this feature is that in the user menu that you have on the top right corner, when you click on the small icon on the small person icon, that is showing a person, then you should get the possibility to copy and paste and just to copy the caption api token and then also the captain auth command.

B

This should help to to get started, and then you don't have to use cube cdl for retrieving this information, and you can directly start to work when you have this token available.

C

Yeah, since it's in the bridge again, you can assign that one to me right.

B

Thanks a lot and we'll move it over and here one last issue for the bridge: it's about the situation that info slis are here shown in in red, but this should not be the case.

B

um Actually, we did not define a definition of stun and not how it should look.

A

B

But there is, I don't want to say a bug, but some some problem that that indicate this issue is as red uh this this this sli thread.

C

Yeah, it seems like there is either something some uh something misconfigured. Let's say it like that or yeah a little back in the in the bridge uh because for the process memory in this case we have no um breakpoints defined. So no no.

C

Targets but it's marked as red, that means that a target is not met actually, so I have to investigate that one since it's in the bridge, so you can assign it to me.

B

Okay, um I mean: should we discuss that one, or uh should we at least define a definition of done.

C

Or is it clear for the definition is that sli without criteria should not be read, never true yeah, because there are no criteria. Actually, so that's okay, but, as I said it could be that it's actually a misconfiguration.

C

Of course the bridge can handle that as well.

D

C

uh But we will see.

B

All right, yeah and then um the last issue that I want to pull in for the next week is yeah the cap. No, that's our old.71 release. um Let me just quickly open that one.

B

It is our checklist for releasing a captain version and yeah still or there's always a lot of work to do like testing, then getting the documentation up to date, then also releasing all the services that are living in the captain, contrib organization and so on, and so on. Yep, and I hope that everybody in the team can help to work on that and that's a jointly.

B

um Let us jointly work on on releasing 0.71.

B

I just will move it over and when we come to the point that we work on this checklist, then yeah feel free to assign yourself.

B

Okay, then, let me just uh talk about that one. um Actually, this is, would be nice to have, but it's not necessary for the 0.71 release. It is related to the captain, enhancement proposal, odet 8 o.6, and it's the idea of extending the distributor to make the distributor possible to receive cloud events on a post event and then just also to do a check of the x token and then forward the event to the to the service itself.

B

Yeah, it's more or less nice to have, and when someone has or finds time before we start working on the releasing process feel free to work on that one.

B

Otherwise, next week we will assign it to someone and to these issues I heard from suri and from imred that they can't or did they don't have time to work on this good first issues.

B

This is the reason why I moved it back to the backlog and when someone wants to get started with an issue with a good first issue then also feel free to grab one of these and to dig into uh coding on captain.

B

All right yeah, since we are a smaller round today, we're also coming to the end of the official meeting and I'm done from from my side. But now, as we have some minutes left, maybe a question to gabriel gabriel.

B

What's do you have any feedback for us or what are you currently uh working on.

A

Yeah yeah thanks johannes, what I wanted to bring um to the to the discussion. Maybe we will not have an answer today, but maybe make a note to discuss it further. Is that um what I'm working is one is the ansible tower service action provider and there is a question from from like the development that I got is when you define the remediation file, you have to specify the values and uh what what it's not clear to me yet is how captain will handle the data type or, if it's gonna, do.

A

I'm not sure if you remember the thread that we had um about this, so the let me just share the screen here: real quick! Is it okay? If I, of course absolutely.

A

All right here you go so basically here is um this one? Let me put it on the on the main on the main channel here.

A

So when I define the action and I specify a value, I can either put it as a as an integer, for example, if it's a number or I can also put it as a as a string in the ammo file, but that that makes it different for for for the services, not just my service, but any service, whether or not they're gonna have to convert the data type um or if captain is going to do any conversion before the event reaches the service or if it's just going to be like a documentation, for example, for the ansible tower service, the values must be specified as string or as numeric.

A

I mean that's, that's the that's the situation that I'm facing, of course, as of now I'll, just you know I can just like, since I have control of both the remediation action file for the development.

C

A

Service I'll, just it's not stopping my development, but I think it's a fair question to to decide.

B

Absolutely um and that's actually a good question whether we can make it a captain feature to convert it, but would you then look for a string or.

A

Yeah exactly so so, for example, um let me find it here if, if it's the same, it's the same line still yeah. So here is um it's on the ansible tower service already running the template, so I've already done the verification that the action is supported.

A

So I identified that it's a it's a launch for the job template and now I'm gonna have to fetch the values, so the values are defined in the data type, as as this data type a map of string and an empty interface.

A

um When I actually have to use it, I I I'll have to get the map and convert it either to string or to or to int. But that depends on the use that I'm gonna do right.

C

A

Like in this case, I'm I'm adding the job template value to the url, so it has to be a string so yeah. I can define a string in the values of the remediation file and then in the actual provider. I can also understand that this is a string, but if I think that this is a numeric data type, then it will break it will crash at runtime.

B

Can I quickly jump in just one one remark: as I see the problem now, I think it would would be a problem or it would be more problematic when captain does the conversion, because then the recipient of this um of this values of this con part will never know uh who did the conversion or what he can ex can receive, and as long as it's aligned with what is in the file and matches that values that will then be received by the action provider. I think then we are on a safer side.

B

D

Okay, I would like to make a few notes: can you please go back to slack.

D

So, first of all, um where do we see the values? Okay, I have the file here here.

A

D

First of all, um this was intentional in captain or when designing the remediation file, that we have chosen an interface or an object in java for the values property here. So this really allows the users of the remediation to define here any values. For example, you define the job. Template equals nine of type integer and so on.

D

So it's an object, so it can be anything.

A

D

I, of course, get the problem um if a user here enters some some properties which are not defined in your remediation action provider.

D

So what you you would now need is some kind of um specification, some json specification, which is first validated when before the remediation is triggered so and for now I would propose to simply make this via documentation.

D

So if you're using your ansible tower remediation action provider, you have to use exactly those values otherwise and those data types. Otherwise it won't work going forward. Of course, we can think of. How can we here ensure a real json specification that yeah and that is valid so, but for now.

A

D

um See this as documentation issue.

A

Previously I had, I think I can see the history previously I had created. Let's see if this one has.

A

I will find the previous version where I had the the actual, the verification uh on the on the value. So I could definitely. This would definitely be the the next the next step. So initially I I was doing a conversion from uh from numeric to string, because this is what I wanted to use uh and we can also catch if there are any errors here that don't comply to the specification.

A

This, of course, I'm not saying that uh if, if the value is not in the right type data type, I will let it crash, but we can also do this and and then just return an error, but in in either case the the end result is that the the action will not be executed so so for now should then I specify in the readme file of the ansible tower service provider.

A

Please specify values in x, data type and then I'll go use it and then I'll. Do this verification and return an error if it's, if it doesn't match.

D

Yeah, but um returning an error also should mean you should send an action from finished, but this action finished should contain as a result um so that we get this info message. Also in captain, then that the action simply failed, because there was a problem. Okay,.

A

D

Makes sense, I would like um to quickly show you the specification if I'm not too,.

D

Sure so um it's already merged cloud events, it's action, dot finished and let's have a look at the example. This is the best.

D

Then you should send an action.finished, but now as a as a status, you should choose here failed, not result. Also the status should be failed, so the result is really whether the action was successful or not, and the status is really whether and the action provider had internal errors or something like this.

D

Okay, so here you can see and the expected values which are recognized by captain succeeded error or none.

A

So if I do errored should I leave result new or should I put it failed.

D

Actually, it does not matter okay, it will be ignored yeah, but please don't make it uh a pass. So because this can lead to confusions in the bridge. Hey, I do have a status which is errored, but the result was passing so either make it failed or leave it empty. Okay,.

A

I then I have a follow-up because is the action uh descent action started and 10 action finished events already implemented? I don't, I know the event specification is done, but I think not not sure. If the function is already created. The template still says it's a as a to do so maybe I maybe.

D

A

B

Yeah, or I think gabriel is referring to the to the template for writing a an action provider.

A

No yeah, I mean the template says it is correct in the sense that it says. Let me get the correct uh line here.

A

If only on the event handlers, it says um my captain.send action started event, uh but then it has a commentary saying implement.

C

A

Action started event in captaingo, utils events.gov. I understand that this is. uh This is not something that every action provider will have to do. It's just something that still needs.

B

A

So every I still haven't uncommented it just to actually send the action started event, because when I looked here, I couldn't find the this function, or maybe it was a while.

A

D

So um can you please send us in the link, because then I can double check this tomorrow.

A

So that's where all the events, the types and maybe the function is not supposed to be here, or this is just the types. Oh no yeah here you go, send configuration change event, send deployment finished. They are all here, but the send action started is not. Let me place this in the chat.

D

Good point, so we will.

D

A

A

A

A

Yeah, but for me from my side, that's about it. I will then continue with the ansible service um and then I'll wait for the implementation of these functions to start sending the events and then send the error event as well. Whenever there's a data type error: okay, good! Thank you very much.

D

Thank you for joining the community.

B

And and sorry for uh for the uh inconvenience that we had to move over to the other, um some meetings no worries, but we will uh will hope to find a solution for the future.

A

B

All right dan yeah, when there are no further questions. um Let's close this meeting thanks for attending everyone and then see you next week, bye see you.

B

B

Video stop recording.