keptn Developer, 25 Mar 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keptn Community & Developer Meeting - March 25th, 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 uh welcome to this week's uh captain community and development meeting it's march 25th, and uh we can have a couple of themes to show and uh to discuss, um as always we're following the cncf code of conduct. uh Please be nice to each other, um but uh you are as always, I guess, and uh so for um today I just want to start with the action items that actually came out from the last meeting and I believe, they're still in progress.

A

um We decided to also include the spell check for um the spell check bot that josh has presented in the last minutes, uh also for the captain, documentation and the capital tutorials. I think this is not yet done, so we probably need to keep it as an action item also for next week, or is there any update on this.

B

There is no update on that one, but on this week um we noticed that at some point this spell checker is a little bit. I don't want to say annoying, but it's brings up too many advices so to say, and we decided to change it and to just use it for the documentation and to really written documents and not to decode, and this is what we, what we want to uh change.

A

Okay, but basically this holds still true uh since.

C

A

Enable this and the other part is basically to um uh it's a bit too verbose.

A

To fit our needs, okay, yeah, that makes sense.

D

It depends a little bit. um I think it makes much more sense if you only inspect so to say the strings which eventually end up in the output to the user in the source chord to only take this part into consideration, but we need to we need to clarify with uh josh if this is doable. If we can tweak spell check like to do such things, so I would not say let's turn it completely off, but let's see if we can relax it a little bit only considering strings and output statements, maybe it works.

A

Should we put it here as an action item or is there an issue.

D

A

Creating an issue I will send it to you later on: it's not cool actually.

D

I think uh we should also.

A

um Once we or if we put this into an issue, we can remove it from from this agenda and don't have to to carry it on. uh Probably we can. We can do it until next week and then I'll present, how the uh I think, especially for these two parts for the documentation, tutorials.

A

um Yeah thanks everyone uh that is uh also joining from the community. Please just add your names, hi raj uh great too, to also see you here um then I want to to give a shout out um to gabriel. He just released a blog post on this, and let me just close my slides here, um a great blog post on how to to do modern, continuous delivery on kubernetes. So it's really developer focused um it's. It's actually really kind of a hands-on blog post.

A

So if you're interested in I put the link in here- and the next step is probably also to have this kind of tutorial, where you can use the potato ed application and then learn how to to deploy the potato head to the multi-stage deployment with captain uh with.

C

Quality gates, um edit.

A

So that's pretty cool and with that I'm already handing over um since I uh I'm also excited about the captain uh about the.

B

Latest captain.

A

Release uh and uh yeah, please take it away, I'm not sure who is uh who.

B

A

B

I added this link, but it's just a summary of what have been done. The last two um last four weeks, and I don't want to talk too much because it's the stage for the people that did the work. I was just writing the release, notes and not doing the implementation, but thanks for having it open right now, yeah it provides you a high level summary of what you get with 0.8.1.

B

Just let me pick the key announcements which are support for using for creating a secret using the captain, api or cli, and this is built on top of kubernetes. This means that we leverage kubernetes there to store the secrets, then then really cool enhancements for in the bridge.

B

As you can see here, the environment screen got the capability to show you which sequence sequences got triggered, and you can also now add a display name for your slos so that you have then a more human human readable as lone amps, which makes it uh more friendlier to look at, and this is what I want. The last announcement I want to point out is that, with this new bridge and new version, we also provide a mock-up um for how we envision the captain. Uniform should look like as when you follow captain.

B

You may have heard this term for um yeah the one or other time, and the captain uniform is the concept of the services that are executing tasks and doing their job as part of their sequence and in the future. We need a way to show you in the ui, where the services live, whether they are healthy or not, and which subscriptions do they have, and with this release, we want to provide you a mock that just gives you an idea how this how this can be implemented and we are well.

B

uh We would like to get your feedback to also then um make enhancements in this regard and to open this mock. You just have to add or to go your you go to a project in your bridge and then you just add ff minus uniform, and this brings you then, to the to the mockup all right, and I think I will stop now, because I see that others want to show their features and implementations that they did and we'll hand over to the next in the in the list.

E

All right, thank you. Let me just share my screen all right today. I want to show you two pro requests. uh Both are related to two services that live in the captain contrib organization, so the first one is the dynatrace service, and this pull request was about establishing feature parity with previous versions that were based on the previous captain version, so 0.7.

E

So the natural service provided a couple of features that were only compatible with that previous cabin version, but not with 0.8 yet, and that has been addressed with that pull request. So, in essence, what has been added in the dyna tray service is now a better support for the auto remediation use case. So basically, what the dynadress service is now able to do is it will subscribe to action.triggered, action.started and action.finished events and uh will then post comments to the correlating problem in diameters that actually kicked off the remediation use case.

E

So if we- and that allows the user to get a better insight in what is happening in the scope of the problem that was detected by the energy. So if you look at that screenshot, we have the the problem page that basically highlights the impacted service and down below. We see that captain or the diner tray service pushes comments to the problem in order to provide an easy way to keep track of, what's happening actually during the remediation.

E

So, basically here we see that a certain remediation action has been started or triggered.

E

Then if the remediation action has been executed successfully and also the result and afterwards the service will also uh please subscribe to the evaluation, finished events and we'll detect if an evaluation finished event has been related to a problem and if that's the case, it will also comment the result of the evaluation to the comment.

E

All right. Are there any questions about that new feature.

E

All right, if not, let's continue with the second service, so this is about the dynatrace sli service. Also, here we basically released three versions that are have been compatible only with uh captain 0.7, and here I also worked on establishing feature parity with uh the version that is compatible with canon 0.8 and this maybe to get a better overview of what has been introduced uh here is yeah. Let's have a look at the release notes of those three intermediate versions.

E

So, let's start here so with version 0.71. The dynatrace sli service also introduced support for dynatrace service level objectives and the correlating api for that. So now, you're also able to retrieve the values for service level objectives that you have created in your dynatrace tenant.

E

Also, what the service is now able to do if it fetches or certain metrics in the context of an auto remediation, it will not only fetch the metrics that are referenced in the slo jamal, but also in addition, if, if it runs in the auto remediation context, it will also fetch the number of open problems that are currently open for the for the correlating service that the auto remediation is performed for, and this will also then be returned as a smi value, all right other than that.

D

E

Have if you have used the service in the past, you may have noticed that uh it also has the feature of generating your sli values and snos uh via the by a dyno trace dashboard.

E

So what that allows you to do if you don't want to use the slo that yaml and sli dot the amplifiers for defining the service level objectives and the clear is for the indicators, maybe, for example, if you don't like yaml that much you could also use the dinatory's dashboard feature and create a set of tiles that follow a certain naming scheme in order to to define which service level indicators should be fetched and how they should be evaluated, and with this intermediate version, additional support for additional sl or dashboard title types have been introduced so now, you're also able to reference a dashboard, a slo tile.

E

So that's a new type of tile in the dashboard and also the data explorer type can now be used in order to generate sli.

E

So using this kind of title you can generate the custom, chart, time series and then use the information from that to uh include it in your evaluations.

E

All right, I don't know that a couple of fixes and better error logging have also been introduced. So that's now also part of this new upcoming version that will be released in the next two days. All right. Any questions about that.

E

F

E

Hand over to yemen.

F

Thank you very much. um I want to present a rather small um pull request or improvement in the bridge uh since yeah, mainly bug fixes have been done, but the one nice enhancement here is. We now also show the evaluation result on the root event. Next to the stage information, and let's have a look at the issue that we had here or how it was requested.

F

uh Previously we showed the stages. This sequence went through and also indicated the result.

F

So you saw here uh red or green or orange uh based on the evaluation or if there are some problems, for example, if something happened, it was indicated in red, but you had not the result immediately here, but you have to click inside this stage, this sequence and to see the details, and now, with this pull request in this overview tile, we show the stages this sequence went through and also include the evaluation result um that was provided in this stage uh next to this stage uh info.

F

So you see that the color- and you also see the result so to say that the score of this evaluation that was executed in this stage- um yeah- that's actually what I wanted to present here. Of course, I can also. I always have a live demo here present.

F

So now, when you see your uh sequences and also in the service screen, uh when you'll have a look at the executions here, you see already in the overview what happened here and, for example, uh this is a good example here I have a deployment that failed in stage and yeah. It didn't make it to production in this case, or, for example, here we have a good version of my card service and because of the good result in staging it's deployed in production.

F

um Yeah, are there any questions from the community or, if not, if there are no questions regarding this feature in the bridge, I'm handing over to ant.

D

G

So just one moment.

D

So we'll make it a little bit bigger.

D

A

um It's still very small.

A

Yeah, thank you.

D

Good um yeah, in the upcoming version of captain, we will have a new service, it's called a secret service and this one will actually provide an api to create or delete the secrets in the in the system. um Why do we need this? um In the end, we want an abstraction for creating secret uh secrets, which shall work for different kind of back ends.

D

For example, you want to create sequence in a void system, or you want to create secrets or secrets natively on kubernetes and potentially also other secret uh systems which are are out there. So there was a discussion going on on github and basically, if you're interested, you can of course read it, or um there was a final proposal, then, in the end, uh how such a secret api can look like as a first version. um So it is not really that interesting from an api point of view. Basically, it's just um yeah.

D

They are just crowd operations where you can create update or delete uh secrets in the system. The interesting part is that the service which is taking over this responsibility, the secret service, um shall be implemented to support different types of back ends. In the end, it should be configurable via environment variables, where you just put in the type of backend you want to use. For example, if you have an already existing void and there's a back-end implementation for this one, you can choose it where this environment variable back-end type. In the first version, only the.

D

Back kubernetes type is supported, but they're eventually coming more than if needed. So that's the default kubernetes. If you just start the service, it will use that back end.

D

As usual, a secret consists of key value pairs and is bound to a scope. We will take a look at that a little bit later. What that means, but here are the quad operations, because you can create a secret update the secrets or delete the secrets.

D

For now getting a sequence where this api is not implemented, I will come to that also later how you can actually make use of the secret that you are creating via this api but yeah. As you can see here, um the endpoint will be secret. Slash version number, then secret, and if you post there a specific payload, for example, this one is a little secret, which has a name, a scope and a data block which consists of key value pairs. So nothing nothing special here you can create it.

D

This way update it that way and also delete it, then, via query strings, so you say: please delete the secret with that name in scope. Captain, for example, um maybe jump. Let's jump to how to use that one, um I will make it a little bit bigger.

D

So we have an api endpoint and it's also already created in the captain cli. So there will be a captain, create secret command or update or delete secret command where you can yeah access this api. Let's take a look at captaincreatesecret, for example,.

D

Just looks like that, um you give it a secret name, and this might be familiar to everyone using kubernetes. You pass in the payload with from literal flag and there you can put in the key and value pairs. For example, I have here a secret which has a key use username and the value beyond.

D

If I execute this command, we have here the message that it was successfully created and I can talk a little bit how it is implemented, as I talked as I mentioned before, only the kubernetes backend is supported right now. So how does this look like on the kubernetes side? Here I have, uh let's make it a little bit bigger.

D

G

No new zooming something.

D

Yes, so is it okay.

D

A

um Yeah, my my screen is a little bit.

D

A

But I guess it's fine for the others.

D

Okay, so we've created a secret called my secret and we should also find the secret here on the kubernetes side.

D

I think it's on the configuration secrets- and here we have the my secret secret um and it has the key value pair username band. Okay, nothing special! So far you would have expected. There is no secret on a community set. Okay, what's also happening behind the scenes is that specific roles gets get created.

D

Let's take a look at the rows, so we should see here a row in the captain namespace, which is called captain secrets default read, so this has something to do with the with the configuration of the secret service.

D

As I told you before, secrets can be bound to a scope and we can configure the secret service via a config map and take a look at these scopes which are currently supported by the secret service.

D

So it will read a yaml file which looks like this.

D

This is the default configuration so to save the secret service. Is it has just a single scope which is called captain default? So if you don't provide a scope here from the cli via minus minor scope, this default scope, captain default will be used and in the background, so each scope can have different capabilities.

D

First, the captain scope as it is right here only has one capability with which is called cap secrets default, read and yeah, as the name suggests. It only permits read access to secrets in that scope, so this one here actually matches them with the roles created by the secret service and the kubernetes side. So here we see.

D

Rows, and here it is, if we take a look at what's in.

C

D

We see okay on this role is for managing secrets. It only provides a read access to get to those secret and there's one resource bound to this role, which is my secret, of course, because that's the secret we've created.

D

So if I, for example, create another secret, um I don't know a secret too. I should see here that the row gets updated with an additional resource, name secret too. So now also, this secret can be read via this role um because it was configured by the secret service and we can of course delete the secret again and then it should disappear from that list.

D

Again um now, what's now important for services who want to actually read the secrets, secrets or make use of it, they need, of course be deployed with the service name, which is bound to these roles.

D

So this way you can right now make sure your service can access the secrets.

D

I have no uh example service available right now, but this is just a matter of configuration how you deploy your service. um If you want to edit your scopes, it's currently only possible, via this.

D

Config map here, so, if you want other capabilities, for example, you've created a secret yeah, my secret, for example, but you have services which might be in the need of reading and updating the secret. You just created an additional capability here, and this will then also be created as a role on the community side and the service who needs to update or read. The secrets needs to have a role binding.

D

Road binding to this role, which might have the name captain secrets default, update, read or, however, you want to name it.

D

um Yeah so currently there's no api for managing scopes. Maybe this will come later you need to for now. You need to edit this here by hand if we only need to provide an api. Also for that there will be an end point where you can say: captain create secret scope and it it will then take care of creating the data entries in this file or, however, it will be implemented implemented. I don't know right now, so any questions to this aspect.

A

um Yeah I I would have um two questions, um so this is basically a convenience function to not have.

C

A

Kind of to to um automate the generation of the the config map and also the um what was it, the service role or something whenever you create the secret.

D

um Well, that's just the default configuration the city secret service needs some sort of configuration to know which roles to create and how they map to the back end.

D

By default, you only have that this one, but you can of course extend it like.

D

However, you need it, so it's not like a convenience function or how you call it, but it's just the concept how the secret management you know on an abstract level works like if you have void, for example, you you or you have name spaces. If you have something from data bricks or something it's called also uh other scopes there. I think, and also it's abstracted a way away via scopes and capabilities. Here you can think of capabilities like um acls, for example, ac. So, basically, basically they are acls for roles.

D

So the captain default role has one acl, which only provides you with read access.

A

Okay, but if I add a secret via cube ctl, it won't be picked up by this song, so you have to create the secret with the concealer. Yes,.

G

Won't be picked up.

D

So um the nice thing is, from a user point of view, you.

D

Don't you create secrets always in the same way, depending on which.

D

Secret back end is implemented in the secret service. It's doing its things in the background, but from the user point of view it will always be the same. You.

D

Don't need to bother with that um there's also, of course, a swagger ui for that which unsurprisingly looks like this here, so you can play a little bit with that to create update or delete secrets using that api.

D

So now the question may arose arise. um How do I, for example, if we have now avoid like.

D

An existing secret management system, for example vault. How can I write or extend the secret service to also support this secret management system? Well, as I or as I mentioned right now, only kubernetes is supported.

D

um Can we make this bigger.

D

D

What's the shortcut of making it bigger control shift plus now, but I'm not sure which operating system you are.

F

Try maybe control and what your mouse will.

E

No think you have to uh activate that in the settings of goland. I've recently stumbled upon that. So the settings- okay.

E

E

E

E

And then the item at the very top so change. Thank you. Gold plus mouse wheel.

D

D

Okay um right here you see the default implementation of the secret backend, the of the secret service, which only deals with the kubernetes backend. If you to extend the secret service, the only thing you need to do is here: the back-end package implement another.

D

Another interface, which is called.

D

Secret back end, unsurprisingly, and.

D

Here you put your logic, for example, you have a void back backend put your logic which is dealing with with voltnip in the in the background background, so that's clear.

D

And you need to provide the initialize method, which will be called for registering your backend. So in this case you, this kubernetes backend here called just kubernetes- will be registered in the registry of the secret service and you. What do you register? You tell the secret service how to create your secret backend, so you basically give it a factory function to create your backend for the kubernetes backend. It's of course uh creating the kubernetes go client or reading the configuration from the cluster and providing the api here in the implementation of the kubernetes secret backend.

D

So it's really easy to extend. The service only needs to implement the secret back-end interface and provide initialize function here, and then you can basically based on environment variable choose which back-end the secret service shall use.

D

Based on the string here, which you use for registering the backend.

D

D

I think that's from my side.

D

If there's no question, I will stop my sharing and thank.

D

A

Cool thanks thanks grant. um I saw that george also joined for today's meeting and uh I think there are no more items on the. uh uh I think I messed up and uh I have to close my window.

C

A

C

A

Is also here I will promote yours to the panelists. Maybe we can uh briefly um what's his back.

A

um Maybe we can ask him how we could actually deal with uh the um the spellbot in in the best way to to make it work uh kind of to make it only work for strings, and not only and not also for other parts like variables, variable names and these kind of things josh.

B

uh Can you hear.

A

Us are you with us here, yeah, I'm here, okay, cool uh awesome, so um I want to share again my document, but I have so many browser texts and uh I kind of lost it.

A

um So our our um idea, what we've discussed- and uh we also discussed the briefly yesterday on slack- was that um the the spell bot is very the spell checker is very verbose and very uh kind of aggressive when it comes to checking for uh misspelled words, and we were thinking about if we can kind of limit it and drink it to only take strings that are actually um yeah, not variable names, not other names of packages and these kind of things. If.

C

There is any way otherwise.

A

We would probably first um kind of use it for documentation and the tutorials markdown files and all these kind of files that, if there's any way, maybe you can help us out here.

G

C

Right, sorry, um the normally you it it. The easiest thing you can do is either exclude files entirely or exclude patterns.

C

So one pattern that you could, for instance, exclude, would be the beginning of the line to the first quote or the beginning of the line or the or the last quote, to the end of the line, um and it would it might work, assuming that your average line only has one string, um I'm not sure. Well, you could. We could definitely try that and see what it looks like. uh I don't know how. Often you have like more than one string on a line.

D

Yes, we have that. Actually, I think uh there will not be a hundred percent uh solution for that, uh especially if we think of different programming languages, we don't uh just have golang, we also have typescript and so on right. We should also work for bash and so on. I think it would be a more uh should be a more flexible solution, which is maybe not detecting 100 of all errors, but nonetheless pretty a bunch of them.

D

um I think we need to try it out how it behaves.

C

I can offer you the the ability to actually talk to the bot now, and so, instead of you being verbose and being annoying each time, you make a commit it's for both first time and you talk to the bot and it updates the commit and as long as you fetch in the that extra commit.

C

Maybe that's good enough.

A

And this would be only once and then like for, for the next committed would know that this word is also now excluded.

A

Yeah, uh actually, this has to be decided by uh by more development.

D

For example, can I share my screen for a moment sure um what I mean so I'm writing this.

E

D

Where we could work on this, ideally with the help of charles church, for example, in our source code, we have, for example, this import statement, which is, I think, from a development point of view, totally fine. um You have import k, kads error abbreviation for that. uh I think that's fine, however. However, the spell check is complaining now that, via some logic, I don't know- I don't know which algorithm was used here, but it's kind of confused about this part of that line that and you need to add nowadays to the ignored file.

D

So I do some some things to to fix that and we came across several of these situations in a couple of days. Maybe maybe this aspect should really be not that strict.

C

So the easiest thing for this example, because you're going to run to a lot would be to add k8s as a pattern and then the bot, instead of seeing s-e-r-r we'll just see, er, and I think that's actually a word and therefore there wouldn't actually be anything for it to do.

D

Okay and the next time it's called k7 and that's the same thing then right.

C

Well, you can adjust the pattern so that, instead of looking for k8s, it works for case slash, ds um and because you now have a pattern, you can see how to adjust it. uh It. I specifically have a pattern that I often use, which is.

H

C

And the numbers, because, especially in kubernetes glands, there are a lot of things that have like alpha v3 or alpha v5 and alpha's a word and the vc. Suffix thing is also a thing and by virtue of just adding its own little pattern, it causes things to just sort of work. Quite well.

C

D

C

D

It could be that the longer we are using this thing, all those all these problems will kind of fade out, because we now have a pattern for everything.

D

um But I'm still not a 100 convinced that checking those stuff is even helpful. I don't know.

A

But might it be a uh a good approach to let's say only, do this the check for uh user-facing components and, let's say, uh exclude the event broker and exclude uh the I don't know like the health service or whatever.

C

And only have it for.

A

The cli before some parts that that include user facing code. Yes, I also.

D

D

I also thought about such a thing. Maybe this is. This would be good, uh for example, if you know okay, this source code file now contains strings. We will which will be presented to the user. Then you put just a command tag or an annotation to that file at the top or somewhere, which is uh known to the spell checker, and if it's present it will check this specific file for spelling actions for all other files for all other source code files.

D

It will do basically nothing but those files which have this special word in it as a annotation command, whatever I'll consider for spell checking, would.

C

D

C

That you can do today um by not by internals of the file, but there's a file called only and that takes patterns. So just.

H

As there's an excludes.

C

File, if you have as soon as you have an only file, only things that match those files, those lines will be checked.

C

So if you had a in the only file, you had say slash ui in it, then it will only look for files that have that are in a ui directory, for instance, and it will ignore everything else, and if you have a second one that offers slash cli slash, then it will look for either things that are in the ui directories or in cli directories.

C

So that's that is achievable today with a line or two of text. Now it is external to the file, but um that means that the the spell checker doesn't have to read all the files to decide. Oh, I shouldn't be looking at this file at all and it's centralized.

D

Yeah one way we could try.

D

Well, I would put this this thing things, for example, in the cli code, where a lot of strings are presented to the user, but I would not uh put it like for internal packages which will never get to the end user, and if there are special abbreviations or something like that, it doesn't matter so much.

D

We could try to add this.

C

I mean I can I can give you like half an hour on slack right after this call to try and set you up with a set that works for somebody.

A

Do you want to take this.

G

Yes, I can take this.

A

D

Yeah then, I will nevertheless uh submit this issue. Then we can at least we have a issue to work on to make a branch and so on, and we can also use this for discussions on even other solutions we might which might come up sure. Okay, thank you.

A

And then I see uh we got another item on the chandeliers and I think we still have like 10 minutes left so um actually this. uh I think I added this as the backlog for next week and I actually also want to keep. Actually I want to keep it for.

G

Next week, raj anyway,.

A

I will just um promote you also to panelists or um because what we actually.

A

Yeah, I think, he's training, but what we actually want to have is, um I think, last week or two weeks ago, we kind of uh distributed the the developer community meetings.

I

I thought the fourth uh thursday is the only meeting.

I

So I thought this is the so I thought that I can show this in today's meeting.

A

Is it also possible to show it in next week's meeting.

I

A

uh That would actually be appreciated, since we are kind of focusing on the core parts of captain in one meeting and then more on the the captain ecosystem in the other meeting, and I think this contribution fits perfectly into the next meeting.

C

So I will keep this.

A

I will edit here, but I would just move it to the backlog for for next meeting.

A

Okay, so we have it, uh but thanks for for being uh proactive and adding this to the to the document, um but I think with this we can uh close this um and uh vernon and joshua thanks for for uh taking up this uh or picking up this. This item um yeah see you all next week in the meeting and then we will take a look at the rush contribution and also another community contribution uh from to the to the cli there. Any more questions or talking points that we should take before closing.

A

Okay, then thanks everyone have a nice evening and see you all soon take care bye.