Cloud Native Computing Foundation KubeCon + CloudNativeCon Europe 2020 - Virtual, 4 Sep 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Build an Automatic Canary Release Pipeline in a Kubernetes-native Way - Ying Chun Guo, IBM

Description

Don’t miss out! Join us at our upcoming events: EnvoyCon Virtual on October 15 and KubeCon + CloudNativeCon North America 2020 Virtual from November 17-20. Learn more at https://kubecon.io. The conferences feature presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.

Build an Automatic Canary Release Pipeline in a Kubernetes-native Way - Ying Chun Guo, IBM

Tekton and Knative are two Kubernetes-native open source projects to extend Kubernetes capabilities. They have defined a series of powerful and flexible Kubernetes objects, in order to build CI/CD pipeline, to create event driven flows, and to manage serverless applications. This presentation will introduce and demonstrate how Tekton and Knative collaborate seamlessly to build an automatic canary release pipeline in a Kubernetes-native way. In this session, you will learn how to monitor CI/CD events with Knative Eventing, how to extract information from events payloads and feed these information into a CI/CD pipeline with Tekton Trigger, how to define a k8s-style pipeline to build and deploy serverless applications with Tekton Pipeline, how to manage traffics with Knative Serving, and more importantly, how these components work together to build the automatic canary release pipeline.

https://sched.co/Zeie

A

Hello, welcome to this session, I'm going chun from ibm china. Today I will share a kubernetes native way to build a canary release pipeline.

A

My chinese name is in chun. I also have an english nickname as daisy. I have long history in open source projects from open office to openstack, and now I focus on serverless open source projects. I have been working on serverless for quite a bit of time, I'm a committer of apathy opponents and an acting contributor of kennedy.

A

In this session. I will share a way to use k-native and tecton to build a canary release pipeline on top of kubernetes. I will show a live demo and touch some part of the technical details.

A

Canary release is a technique used to reduce the risk associated with rolling out new code or functionality to everyone by making the new release only available to a small group of end users due to the small size of the end user group. The impact to the new release is relatively small if it's determined that the bug exists or the new functionality isn't well received, then it's easy to roll back.

A

This release pattern can bring some benefits. It allows deployment to production easier and more often reduces deployment life cycle size lens. It allows the test in production, which can make sure the testing environment is exactly same as the production.

A

It allows phased roll out of new applications which reduce user exposed to negative issues. It can ease drawback implementation in case of problems, of course. Besides the benefits, it also adds complexity to the deployment management.

A

You are able to implement a canary release just using kubernetes. You will need to build your app into a container image. Push the image into a ripple, create deployments configure the replica set. You may also need to configure endpoints ingress or virtual services for the traffic routing. You may need to use yaml or helm to do the deployment.

A

If you want to use is2 or some other gateways, you may need to master more knowledges just in order to build a canary release pipeline. All these resources have to be managed by yourself. I don't think it's a delighted. Experience then it comes. Canadian kinetico is designed to provide a service platform for developers which will ease the developer workload to manage their cloud-native applications on kubernetes kinetico defines a set of primitives to build this kind of modern, source-centric and container-based applications.

A

Since kinetico is kubernetes native, it can leverage all the tools in kennedy ecosystem, for example, the auto scaling and the telemetry it leverage service mesh, for example, is to to define routing and network programming and to enable security access kinetic. You can use zipkin grafana promises to get better observability.

A

It has two major components: one is kinetic serving which defines a series of crds to control. Serverless workload behaves, for example, the rapid deployment of serverless container, auto scaling up and down to zero. The other is kinetic eventing, which provides resources to enable the late, binding event, sources and even consumers.

A

Let's spend a little bit of time on the kinetic major components, because we use these concepts in the following content: cognitive serving builds on kubernetes and is2 to support, deploying and serving of serverless applications and functions.

A

It defines crds to describe a developer view of cloud native application for developer service acts as a top level container for function. Configuration acts as a stream of environments. Different configuration will create different revisions. Revisions are immutable, snapshots, snapshots of codes and configures a route provides a named endpoint and a mechanism for routing traffic relations.

A

So with canadian a developer can easily deploy a function as a service while connecting will create the corresponding configurations, revisions and route and including all the below kinetic objects.

A

It also provides a capability to automatically scale up and down to zero traffic can can be splitted based on percentage. A route can submit traffic to several revisions. This is a key requirement of canary release.

A

Cognitive eventing is a system that is designed to address a common need for cloud-native deployment and provides composable primitives to enable late, binding events, sources and event, consumers. It defines primitives to manage event, consume producers and consumers. It allows provide a persistent data channel for the event state flow. It has defined some event. Cells among them, I want to highlight, is the api server source api server source provides a capability to consume kubernetes cluster events by a function that has been implemented as a canadian service.

A

You can monitor the quit update, delete of cognitive serving objects, for example the creation of a cognitive service, configuration revision or the update of a route. By doing this, you are able to capture the events happened in a canary release pipeline and you are able to consume. Such events implement a function as a kinetic service to handle such events.

A

Techton is also a kubernetes network system. It provides a powerful and flexible flex, flexible, open source framework for creating ci cd systems. It lets you build test and deploy applications to public cloud or on-premise systems. Your high, it hides hides the implementation and provides a series of coordinated style resources to define your tasks and your pipelines.

A

There are two major projects in taxon which help to trigger a process to build and deploy. Your application. Tecton pipeline provides several quantity, still resources for declares for declaring ci cd pipelines. For example, you can define tasks which are a collection of steps you would want to run as part of your ci flow.

A

A pipeline is a sequence of tasks and parameters and resources use that input and output of the tasks tactile trigger, allows you to extract information from events payloads and feed this information into a pipeline pipeline and run the pipeline so that you can trigger your pipeline by a crown job event or by a github event.

A

We will use these two features to build. The canary release pipeline you will use evoc is automatic and much easier. Then you manage the kubernetes resources by yourself.

A

I will do a live demo and show you step step by step. How can native and tactile work closely to implement a canary pipeline?

A

First, I will introduce some background of the demo. My application source code is stored in github.com. My container image is stored in ibm container registry on ibm cloud.

A

I have a kubernetes cluster running on ibm cloud and enable the kinetic add-on so that the managed, kinetic and tactile has have been installed in my cluster just by a single click.

A

Let's assume I have a hello service running in my production environment. I have a lot defined to pass 100 traffic to my hello production revision.

A

Let's show this in our terminal.

A

A

You will see there is one revision there named as hello.

A

A

I have one load in my system used as used for product named as product route. It has an endpoint. Let's try the endpoint.

A

It's a hello service. It will return a hello statement: okay, so hello, world. This is kubecon europe 2020 and it will return uh kubecon, europe, 2020, stably.

A

Okay, all these resources are in my production, environment,.

A

I will start the demo I plan to use canadia and attack tong to create a canary release pipeline. I'm going to do a git, commit and trigger attack time pipeline to do the tasks, since it need time to complete the task and it's a live demo. I will make a commit first and explain later.

A

I will go to my git ripple. Okay, this is my git repo. I will change the source code of my application and make a make a commit.

A

You see there is a kukkong europe 2020, I'm going to change it to 2021.

A

And the make up committee's commitment message.

A

We have done the commit.

A

I will start a watch window.

A

To monitor the running pawn point in the.

A

A

In the back end, I have a tactile event listener defined in my cluster. I have an ingress and a webhook is defined in github to send the github event to this event listener.

A

A

A

I have a I have a webhook defined.

A

So when a git commit happened, github will send an event message to the event listener and the event listener will pass the event message to a trigger.

A

The trigger will extract the useful data and pass them to a tactile pipeline. The tactile pipeline has two tasks. One task is to build the cells into a container image and push the push it to the image registry.

A

The other one is to deploy objects to kinetic.

A

Let's see the detail.

A

You see, I have two tasks just to image and deploy use. Kubekato.

A

I have a pipeline and in my watch window you will see two tasks have already been completed. Why is sauce to image? The other one is deployed to cluster.

A

After these two tasks completed a configuration and revision and route will be created, the route will generate an endpoint.

A

We can invoke that endpoint for testing.

A

A

You see, I have two revision now, this hello production is the old one and the hello internal movie 2021 is the it's a it's the new one we just created.

A

And we have two routes: one is the production route and the other one is named as hello internal internal. It's an is a new one, new new route for the new revision. So we can test.

A

A

You will see the output message is kubcon 2021 now. So this is a new version of.

A

A

We can do a fully test for the new version.

A

After testing, we ensure that the new endpoint of the new version of the app has has no failures or exceptions in the tester case test cases.

A

We want to integrate the new revision, the new endpoint to the real production and split a small part of traffic to it and try this new new version.

A

We will update the route.

A

So you see currently the traffic ascent is, since it sent 100 to a hollow production revision, and now we are going to split.

A

A

A

To the new revision.

A

We will apply.

A

A

Okay, it has been updated now, let's test the production.

A

A

You see mostly it will return 2020. for fewer cases. It will return 2021.

A

As a summary in this presentation, I showed you how to use kinetic and tactile to create a canary release with tacton and the kinetic. You don't need to understand the lower level objects of kubernetes. You only view tasks, pipelines, kinetic services, revisions and routes, one with tactile trigger and cognitive inventing. You can make this pipeline more automatic.

A

The pipeline is triggered by a github committee event with api server source. We mentioned the infinitive event team. You can monitor the creation of a kinetic revision and trigger automatic testing. I don't include this part in my demo, but surely you can implement it.

A

The last part of the canary release is too low out to the new revision graduately.

A

Maybe we don't want, maybe we don't want to have the that part automatic, because we want to do the roll out carefully and manually. We can use kinetic route to define traffic. It's very easy!

A

So that's the end of my presentation. Thank you very much for your time.