keptn GSoC in Keptn, 4 Oct 2022

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: Keptn Integrations for GitOps-based Deployment Tools: ArgoCD and Flux

Description

In this session Meha Bhalodiya, a GSoC contributors in Keptn, will present her project: "Integrations for GitOps-based Deployment Tools: ArgoCD and Flux".

To be able to use GitOps tools in combination with Keptn. The main goal of this project is to find a proper way how the tools could interact with Keptn and to build an integration for such tools. Deployment tools can be integrated via the helm-service or via the job-executor service. Currently, there’s no way to integrate GitOps Tools in Keptn. This project aims to provide a way such that such tools can be integrated into Keptn return deployment states after the state has been synced.

More about Keptn in GSoC and the projects: https://github.com/keptn/community/blob/main/mentorship/gsoc/2022/projects/gitops-deployment-tools-integration/README.md

A

Hello, everyone, my name, is Miha I. Am the captain face of contributor of integration for github's, based deployment tools focused on flux and Argo City? So, yes, I am honored to give the final presentation today about this project. uh I would like to have a brief introduction about myself. First, so I am an undergraduate student from gujar Technological University, pursuing a Bachelors and computer engineering, so I am also a newbie in the captain. Open source community and I am grateful.

A

That Captain can give me a opportunity to realize the idea and this project so first understand what Captain is so essentially it's quite hard to understand at First Sight. But yes, we can say that it's an even driven orchestration engine for cloud native apps, so it uses a declarative approach to automate the delivery and operations, so they can be scared up to a large number of services.

A

So here, as you can see in the picture when a change is detected, github's tool plays a role and notifying the captain about that and Captain can trigger a deployment after approval or automatically. So initially there might be some configuration of Argo, City or say flux somewhere, so the first step would be like Argo, CD or flux is connected to a get repo which contains a deployment, manifest hand, chart and things with this repo.

A

When there's something new, then I'll go our flux. It triggers the deployment and notifies Captain about this, like deployment, dot, triggered or started event, and then the Manifest also hand charts get applied to the Target cluster and at last, Argo, CD or flux notifies captain that everything is finished and it makes the pr to update the production.

A

Now before creating a project. We need to understand what is the shipyard file, because Shipyard file is the main thing which is used while you are playing with the captain. So let's understand it from the documentation.

A

So yes, uh Shipyard is defined at a level of project and it means that all services in a project share this same Shipyard definition now Shipyard defines the stages each deployment has to go through until it is released in the final stage, that is, production, State and a shipyard can consist of any number of states, but at least one uh that is like a stage must have at least the name. Property and same applies with the number of sequences it needs to be have at least one.

A

So let's have a look at the shipyard file of our project. So here you can see it has uh stages like say, test production and inside the stage it have. It has sequences here. I have one named kickoff and it has various tasks to do so. We can see this thing in our captain dashboard.

A

Here you can see, I have one sequence running kickoff and it has various tasks running inside it.

A

And it says that it started and it finished it gives us various notifications.

A

And yes, if it fails foreign.

A

A

So yes, this is uh all about Shipyard Phi.

A

Now, uh let's come to the notification controller, which is the main component and uh Captain flux integration. So, yes, let's understand this. We can get input or Source changes from hand, release updates, terraform, git repositories, customize or say all of them at once and then notification controller is the component of the flux essentially would get the notifications and then flux will send the Json payload, which can change anytime.

A

So the controller handles the events coming from the external systems like say: GitHub, gitlab, bitbucket Etc and it notifies the getoff's toolkit controllers about the source changes. So the controller handles events imitate by the github's toolkit controllers uh and those Badgers thumb to the external systems based on the event severity and the involved objects.

A

So, yes, it is all about handling events. uh There might be inbound or outbound. Now, let's understand uh the two main Concepts in the notification controller, that is provider and alert which we need to use while we are doing such Integrations, like especially in the flux, if you are playing with around the clocks. So yes, let's have an overview of the just to get UPS toolkit components. So what provider is the provider is a type of a API which is used to define how events are encoded in the web hook address?

A

Well, they are dispatched, so the provided type can be slack. Ms teams just called generic Google Chat and many search. So when type generic is specified. Why I am talking about generic, because here you can see our project is using the type generic right. So yes, when type generic is specified, the notification controller will post the incoming event in Json format to the webbook address.

A

Now this way you can create custom handlers that can store the events in elasticsearch Cloud watch stackdriver Etc. So here you can see the web hook address right so to receive the event as notifications, a provider needs to be set up first and once you have set up a provider, we need to create a new alert resource in the flux system to start receiving notifications about the hand release right. So, let's come over to the alert.

A

So the alert API uh says that how events are filtered by severity and involved object and what provider to use for dispatching right here you can see uh we have given the provider reference now to verify that the alert has been acknowledged by the notification controller. We need to execute one command, that is Cube, CTL, hyphen and flux system get alerts, and you will see the alert name. Multiple alerts can be used to send the notifications to different Tenors or say slack.

A

Organizations now here with the one important specification is event severity in the alert, so the event severality can be sent to info or error. Now, when it is said to invoke like in our case, we have set it to the info, what it will do, like the controller, will alert if a kubernetes object was created, updated or deleted, and also it sees that help checks are passing.

A

Another important thing is: uh it gives you a dependency if the dependency is delaying the execution or an error occurs now, when the severity is set to the error, the customized controller will alert on any error encountered during the reconciliation process. Now this includes customized, build and validation errors, apply errors and health check failures. If you don't specify an Event, Source namespace, the alert namespace will be used.

A

So here you can see how a provider and the payload play uh play place, a role together when uh event have an error.

A

And if the status is all right, so yes, we can also trigger a new sequence in the Captain dashboard.

A

Here we are selecting Potato Head.

A

And let's say the status test.

A

A

A

A

A

And it says that test evaluation get SLI is passing and get promotion is failing. So, while it is failing so that is also given here like what the fills are missing. What validation error is there, so this is. This is what we have triggered a sequence.

A

Also, uh we can see the load testing I have my grafana dashboard running is so here. I have selected Captain Ingress as an Ingress object, so you can see how many requests are being made per second and yes, you can see all at a time.

A

So yes, this is all about the captain flux integration. We can extend up this to the other things like integrating the web hook like we are working with the slack web hook thing, so you can have a look at it. Artifact.

A

A

Can be done with this, like webhog integration, we can add this to our integration so that it sends us the notification and the slack, and we get to know like what uh changes have been made. What uh failure is being there and yes, we can subscribe to our captain event and to push the notifications to the slack and, yes, we can. uh We can receive the notifications of each and every activity has been made to our get repo.

A

So, yes, we are working on this and, yes, there are many future Scopes uh with this Captain flux integration. Thank you.