Excercise - 2
In this lab, you'll learn how to update the number of instances in a deployment has and how to roll out an update of your application on Kubernetes with zero downtime. A key requirement of a container orchestration system is to automate the management steps of applications. Scaling up/down and handling application updates with support for rollback are a couple of essential use cases (but not the only ones).
For this lab, you need a running deployment of the guestbook application from the previous lab.
Scale apps with replicas
A replica is a copy of a pod that contains a running service. By having multiple replicas of a pod, you can ensure your deployment has the available resources to handle increasing load on your application.
kubectlprovides ascalesubcommand to change the size of an existing deployment. Let's increase our capacity from a single running instance ofguestbookup to 10 instances using:If you remember back to the architecture overview in the previous exercise, the
kubectl scalecommand updates the desired state in the etcd database in Kubernetes. The watchers and controllers will now try to make reality match the desired state of 10 replicas by starting 9 new pods with the same configuration as the first.To see your changes being rolled out, you can run:
The rollout might occur so quickly that you may only see
deployment "guestbook" successfully rolled outfor the output.Once the rollout has finished, ensure your pods are running by using:
You should see output listing 10 replicas of your deployment:
Update and roll back apps
Kubernetes allows you to do a rolling upgrade of your application to a new container image. Kubernetes allows you to easily update the running image but also allows you to easily undo a rollout if a problem is discovered during or after deployment.
In the previous lab, we used an image with a v1 tag. For our upgrade, we'll use the image with the v2 tag.
To update and roll back:
Using
kubectl, you can now update your deployment to use thev2image.kubectlallows you to change details about existing resources with thesetsubcommand. We can use it to change the image being used.Note that a pod could have multiple containers, each with its own name. Each image can be changed individually or all at once by referring to the name. In the case of our
guestbookDeployment, the container name is alsoguestbook. Multiple containers can be updated at the same time. (More information.)Check the status of the rollout. The rollout might occur so quickly that you may only see
deployment "guestbook" successfully rolled outfor the output.Test the application as before, by accessing
<public-IP>:<nodeport>(use the same as the previous lab) in the browser to confirm your new code is active.Remember, to get the "nodeport" and "public-ip" use:
and (replace mycluster below if you used a different name for your cluster):
To verify that you're running "v2" of guestbook, press down shift while clicking on reload to discard the cached content and look at the title of the page, it should now be
Guestbook - v2If you want to undo your latest rollout, use:
You can then use
kubectl rollout status deployment/guestbookto see the status.When doing a rollout, you see references to old replicas and new replicas. The old replicas are the original 10 pods deployed when we scaled the application. The new replicas come from the newly created pods with the different image. All of these pods are owned by the Deployment. The deployment manages these two sets of pods with a resource called a ReplicaSet. We can see the guestbook ReplicaSets with:
Before we continue, let's delete the application so we can learn about a different way to achieve the same results by using resource files instead of providing command line options.
To remove the deployment, use:
To remove the service,use:
Deeper dive on configuring resources
Although it was convenient to create the guestbook deployment using the cli, in practice most applications and other kubernetes objects are created using configuration files in .yaml format. For example, the guestbook-deployment.yaml file is an example configuration file that would deploy the guestbook image with a total of 3 instances. This file shows the key parts needed for each kubenetes object. After the API version and resource type, there is a metadata section which specfies the name of the resource and a set of labels. Then, there is a spec section which defines the desired state. First, there's the definition of the replica set object for the deployment. Within the replica set, there is a template for the pod controlled by the set. Within this template, you can find labels applied at the pod level and the spec for the container(s) that will be deployed with each pod.
In the last section of this lab, you will again deploy the guestbook application, along with other objects for a multi-tier application using resource files.
After these are deleted, proceed to the next lab of this course
Last updated