keptn User Groups, 4 Apr 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Developer Experience: GitOps for Keptn

Description

Keptn is moving from Git-based to true GitOps allowing developers to focus on pushing new features without having to learn the ins & outs of Keptn. Watch this webinar to learn about the developer experience Keptn gives you.

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
Sign up to our newsletter: https://bit.ly/KeptnNews

A

Welcome everyone to today's special captain, webinar developer experience github with captain. I have asked thomas who is driving all of our githubs initiatives with captain to come back and do a webinar specifically focusing how github's with captain really feels for the developer. Thomas I'll leave it over to you, because I know you have a lot of great cool live demos to show.

B

Okay, so uh hello, andy thanks for having me, my name is thomas, I'm a principal and planetary maintainer and today I'll show you some thing about youtube subscribing. So um what were our initial challenges? So we got the feedback that users want to deliver software and not to learn captions, so they don't want to put in many imperative commands. So they only want to check in some code and want um to use captain with this.

B

And um yes, so we come directly to the demonstration today. So, um let's switch over to the shell um after uh git repository is set up. We can see we can initially trigger all of that with a simple cubesat command. So let's say cubesatl apply means from demo and on the other hand, we see that our captain instance is pretty empty at the moment and in a few seconds we should get a new project.

A

And also maybe for a reference. This is a special hosted version of captain from banner trace. It's called cloud automation, but essentially what you are showing here is the same: whether you're hosting captain open source yourself or whether you're using uh the dynatrace enterprise version, which we call cloud automation.

B

So the background our reconciliation loop is running and it might be to take some seconds until this project comes here and we see that we got the potato head project in there. So what happened at the moment? um We prepared a repository which includes all of the configuration we need for captain, so we specified some kind of an instant. So where is our captain instance? And how can you connect to this?

B

We also specified our project like how is our project called, and how can we connect to the upstream repository?

B

Also, we define certain sequences, so we defined everything which what should happen when we trigger new deployment such we configured, our monitoring, we deployed it, we tested everything and afterwards we evaluate everything and we specified our series so in this case the potato head and we specified that we want to have a captain service which is called potato head which belongs to the project potato head and we need the event artifact delivery retriggered to deploy this and at the same step we also defined the captain service deployment itself, which tells captain that the version 002 of the potato head should be deployed.

B

A

Could you zoom in just a little bit on the on this on the browser? I think it will just help a little bit for the next file. You show.

B

If it works, if it works,.

A

Control plus, just to give you an hint at least that's how it works on my machine. Okay, is.

B

A

It's better. Thank you, perfect,.

B

Okay, so this was the service deployment itself and we defined three stages here, so we have what they have for hardening and the production stage, and if we take a look on the cloud, automation or captain again, we should see that in our in our instance, we got our service potato head and our sequence for delivering the potato head is triggered now, so we got a promotion event, we got a monitoring week uh will be configured and so on, and this might be the way how it feels like for developers and further.

B

In further consequence, everything will be deployed, everything will be tested and so on and after some time everything will be finished.

B

Okay, some questions from you, andy yeah,.

A

I got a question, so can you go back to the git repository please.

B

um Yes, just a second, let's take this one.

A

Okay, so if you could go to the uh what you showed us earlier um exactly so just that I understand the only thing I really need as a developer. Obviously somebody will, you know potentially help me in in the initial setup of all of this, uh but it's all cid based. That means the captain project, the sequences, the stage the services and then the individual files, like my helm, charts my test files, my slis and slos. This is also all here in the git repo.

A

I think under the base and under the stage is directory, correct.

B

Okay, yes, yes, yes, there are two directories here, so there is a base directory which, uh which includes all of the files and configurations you need in every stage for a for a particular service. So for the service potato head, we have specified our dyna trace configuration, we have specified our helm, chart job configuration, k6 configuration and so on, and this might be the same for every stage and we have a second directory, which is the stations directory which consists which includes all of the configuration which should override um the base for the particular stage.

B

So if you have to change some values in the values file, you can change this in the stages and then it will be merged to the to the base directly.

A

And if I now, I know that, currently the sequence is running to deploy uh this particular version of potato head. I think it was version 0.2 or something like this. If I would now want to deploy the next version, the only thing I need to do is I just need to update my crd to a different version and that's it.

B

You have to change two things in at first, you have to stay. You have to change the meter data file in the artifacts directory, so in the base directory to the version zero zero. Three. This tells the captain upstream repository that there is a new version of the of the artifacts and if you only change this version, nothing happens so uh its own. It gets only committed to the upstream and when you change the version in the service file. So when you change the version here, then it gets updated and so.

A

B

Simply change, you can simply stage your changes and afterwards change them, apply them.

A

Cool, that's awesome, perfect and I think, can we have a final look back at your bridge at the captain, ui or in our case cloud automation, but just see what's happening there.

A

B

um We saw that our monarch configuration was running so with in our case we configured our monitoring system in this case, dyna trace to use the configuration which is in there. Then afterwards we deployed everything now afterwards test for running. In my case, I used k6 for running running twists. After of this service and afterwards we we are running some revolution against our trees, so we will get some some sli spec if every slice against the slow speed. If everything works- and I think in a few seconds or minutes, everything will be finished.

A

Cool and also just to recap again, the k6s test scripts are in the repo that you've shown earlier. uh Also, the sli and the slo definition is there that is currently being evaluated, and if you have a new service, a new micro service. The only thing you need to really do is add the additional service crd and then the deployment cd and from that point on captain will take care of running the sequences again to then run through the same thing. Configuring monitoring, deploying testing and then evaluating, and hopefully we can.

B

And one funny one funny side effect at this point is we are using metrics, which we defined in the monitoring configuration before for our slis. So.

A

B

Configured our monitoring and using that using them to evaluate the state, the falsehoods.

A

Perfect all right, then, let's hope that we can still wait as long as until the evaluation is actually complete. I know the screen only refreshes every 30 seconds, I believe so.

B

A

Let's do it one more time almost there sometimes it takes a bit longer yeah. um Is there anything else from a development perspective that you have worked with a lot of teams that are kind of starting with this? Are there any other questions that people have.

B

What's pretty, what's pretty funny is um sometimes I get the question if this should be if this can be split in different repos, so it might also be the case that you want to specify the project and service configuration one repo and the service, um the service deployment configuration other reasons, and yes, this is also possible, so you can split whatever you want in different repositories and the operators will take care of that. Everything is runs in the correct order.

B

What happened in the meanwhile, so our evaluation got finished. We can also take a look on our sli, so we see that we got no errors. We saw we see that our response time was, as we expected in our slos, and everything is running fine and now the same process runs in the next stage. So now we are on the second stage in the hardening stage, and this will work uh go through until production in this case, and yes, that's the way it should.

A

Be last question that I have here the promotion step. This is really the step where the version information, the configuration information is kind of applied from depth into hardening. I would assume is that the promotion step.

B

Yes, that's the that's the step which brings the captain upstream repository in a state that everything is consistent and that exactly this, this version we want to deploy is present in the is present for captain. So one thing our approach allows is also that you can roll back roll back and so on. So you can also deploy the last version again and so on so, and the promotion step ensures that captain has always the information.

A

Cool all right, I think we let this run, but we can see it's really multi-stage sequences, it's currently deploying in hardening, and then it continues depending on how you specified your sequences. I assume it will eventually make it into production. Yes, it looks good. It looks good. It's awesome, anything else that we need to wrap this up.

B

Yes, so um if you need some more background to this to all of this, we we had a captain music group around two weeks ago, where I also talked about some some background regarding this and gave some more information.

B

And yes, if you want to try it out, feel free to do so, so that everything is running on it.

B

You can try it out post issues, if you, um if you find some, if you want to make something better, I think I will approve also almost every pr- and yes, there are also could be some of code projects for that. So if you want to step in open source development and so on, um make a proposal, and yes, I would also be glad to work with you and last but not least, just share your thoughts.

B

So you can ping me on the captain's leg and there is also a channel for that in the capital to operate. The working group channel awesome.

A

Thomas, thank you so much for doing this, because I think it was really nice, concise and really showing how this feels for a developer.

A

If everything is set up and as you said, if people want to know more there's a lot of ways that they can help you and come back to you and ask questions or give feedback.

B

Yes, thanks for having me.

A

Thank you, bye-bye.