Configuring K8s

Owned by Eugene Fedorenko (Unlicensed)
Last updated: Nov 17, 2020 by Emily Simon

When a project is being deployed to a K8s Cluster, FlexDeploy offers the ability to configure a Helm Chart and/or deployment strategies for the deployment.  Both configurations are done on the following tab:

Field Summary

NameDescriptionNotes
Deployment Name

This is the K8s deployment name.  The field is a groovy script by default

so be sure to enclose literal strings in quotes.

Result after a deployment:

kubectl get deployments

NAME
springboot1.0.5

Manifests to deployActual K8s manifest files for your deployment.  Files will run from top to bottom.

Usually consists of a deployment.yaml and service.yaml. File paths can be absolute or relative to the temp directory.

To make use of the Deployment Name field above, your deployment.yaml should use FD_KN_DEPLOYMENT_NAME flexdeploy property and can

also use the FD_PROJECT_DOCKER_IMAGE_NAME property configured on the docker tab.

deployment.yaml 
apiVersion: apps/v1beta1
kind: Deployment
metadata:
  name: ${{FD_KN_DEPLOYMENT_NAME}}
spec:
  replicas: 1
  template:
    metadata:
       labels:
         app: superapp
         fd_deployment_name: ${{FD_KN_DEPLOYMENT_NAME}}
         fd_app_name: SpringBoot
    spec:
      containers:
        - name: superapp
          image: ${{FD_PROJECT_DOCKER_IMAGE_NAME}}
          ports:
            - containerPort: 8080
service.yaml 
apiVersion: v1
kind: Service
metadata:
  name: superapp
spec:
  selector:
    app: superapp   
  ports:
    - port: 8080
      targetPort: 8080
Blue/Green Istio Manifest

Istio manifest file to apply when running BlueGreen deployment strategies.  If using a different strategy you can leave this blank.

The BlueGreen strategy is effectively an 'On/Off Switch' for your pods.  When using this strategy all of your pods will deploy with a BLUE label, meanwhile 

traffic is always routed to GREEN labels.  Testers and QA will test BLUE pods through back doors and other means.  When ready you can 'flip' the switch to

label all BLUE pods as GREEN resulting in them opening up for regular traffic.

As you can see in the below manifest file, the GREEN route will always receive 100 percent of the traffic through the Istio Mesh.

istio-bg.yaml 
apiVersion: networking.istio.io/v1alpha3
kind: Gateway
metadata:
  name: superapp
spec:
  selector:
    istio: ingressgateway
  servers:
  - port:
      number: 80
      name: http
      protocol: HTTP
    hosts:
    - "*"
---
apiVersion: networking.istio.io/v1alpha3
kind: DestinationRule
metadata:
  name: superapp
spec:
  host: superapp
  subsets:
  - name: green
    labels:
      fd_version: green
  - name: blue
    labels:
      fd_version: blue
---      
apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: superapp
spec:
  gateways:
    - superapp 
  hosts:
    - superapp    
  http:
  - match:
    - uri:
        prefix: /version
    route:
    - destination:
        port:
          number: 8080
        host: superapp
        subset: green
      weight: 100 
    - destination:
        port:
          number: 8080
        host: superapp     
        subset: blue   
      weight: 0
Canary Istio Manifest

Istio Manifest file to apply when running the Canary deployment strategy.  If using a

different deployment strategy you can leave this blank.

Canary is similar to the BlueGreen strategy but instead of being an 'On/Off Switch' think of it as a 'Dimmer Switch'.  Instead of cutting all traffic over in one moment

you can adjust the weights as you go.  For example, you can start at routing 90 percent of traffic to GREEN and 10 percent to BLUE, then 70/30, then 50/50, then full cut-over at 0/100. 

You will notice the the manifest file looking similar to the BlueGreen manifest with the exception that we can see FlexDeploy variables used in the weight fields

(Check out the pipeline documentation to see where these weights are set):

istio-canary.yaml 
apiVersion: networking.istio.io/v1alpha3
kind: Gateway
metadata:
  name: superapp
spec:
  selector:
    istio: ingressgateway
  servers:
  - port:
      number: 80
      name: http
      protocol: HTTP
    hosts:
    - "*"
---
apiVersion: networking.istio.io/v1alpha3
kind: DestinationRule
metadata:
  name: superapp
spec:
  host: superapp
  subsets:
  - name: green
    labels:
      fd_version: green
  - name: blue
    labels:
      fd_version: blue
---      
apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: superapp
spec:
  gateways:
    - superapp 
  hosts:
    - superapp    
  http:
  - match:
    - uri:
        prefix: /version
    route:
    - destination:
        port:
          number: 8080
        host: superapp
        subset: green
      weight: ${{FD_KN_OLD_WEIGHT}} 
    - destination:
        port:
          number: 8080
        host: superapp     
        subset: blue   
      weight: ${{FD_KN_NEW_WEIGHT}}
AB Istio Manifest

Istio Manifest file to apply when running the AB deployment strategy.  If using a

different deployment strategy you can leave this blank.

The AB deployment strategy still involves BLUE and GREEN pods like BlueGreen and Canary.  The difference here is that BLUE traffic is only ever open based on a

'Routing Rule' which could be an http header of your choosing.  For example, below we can anyone with the 'end-user' header set to a FlexDeploy property configured on the

pipeline will get sent to a BLUE pod, and everyone else is routed to GREEN.  After the end-user has validated the BLUE pods, we flip them all to GREEN similar to the

BlueGreen strategy.

istio-ab.yaml 
apiVersion: networking.istio.io/v1alpha3
kind: Gateway
metadata:
  name: superapp
spec:
  selector:
    istio: ingressgateway
  servers:
  - port:
      number: 80
      name: http
      protocol: HTTP
    hosts:
    - "*"
---
apiVersion: networking.istio.io/v1alpha3
kind: DestinationRule
metadata:
  name: superapp
spec:
  host: superapp
  subsets:
  - name: green
    labels:
      fd_version: green
  - name: blue
    labels:
      fd_version: blue
---      
apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: superapp
spec:
  gateways:
    - superapp 
  hosts:
    - superapp    
  http:
  - match:
    - headers:
         end-user:
           exact: ${{FD_KN_ROUTE_RULE}}
    route:
    - destination:
        port:
          number: 8080
        host: superapp
        subset: blue
  - route:
    - destination:
        port:
          number: 8080
        host: superapp
        subset: green
Helm ChartThe Helm chart to apply during deployment.Can be a chart in the repository configured on the deploying endpoint, a folder with a chart, a chart package or URL.
Helm Release Name

Release Name in Helm to easily track all deployments on the endpoint.

Has nothing to do with FlexDeploy Release*

If it is unspecified, Helm will generate a random name.
Parameters MapParameters to pass to the Helm Chart
 Example

In this example we will walk you through deploying your Kubernetes project through FlexDeploy. In this example we are assuming you have Kubernetes/Docker installed  on a specific endpoint. 

SCM Setup

For our first step we will add our Kubernetes manifest files to some sort of SCM. For this example we will be adding a deployment manifest to a git repository. Below you can see our .yaml file we added.

 deployment.yaml

Workflows

We will now create a workflow to build and deploy our manifest files. For our build we will use an existing workflow so we will not need to create a new one. For our deploy workflow we will add the Kubernetes deploy operation. For the inputs on this operation you will just need to add the context which in our case is minikube.

Once we have added this we will then activate the workflow.

Topology

We will now create our topology. First we can create our endpoint if not already created. After creating an environment and instance and associating our workflow and endpoint with our new environment instance we can set all of the properties. 

We will also need to create our SCM instance in topology. This integration instance will point to our repository containing our manifest files.

Project

We will now set up our project by configuring our workflows, instance and SCM instance.


We will then go to our kubernetes config tab and set some information. First we will need the deployment name. This will be the name the deployment is given. This will be returned by kubectl get deployments once the deployment is finished. We will also set our manifest files. This will be the path to the manifest files relative to our artifacts directory. In this example we will only need to give them the name of the file.

Build and deploy

We will now build and deploy the project. The build will pull our manifest files from git and put them in our artifacts directory. Our deploy will use those manifest files in the kubectl deploy command. If everything worked you should be able to see your new deployment on kubectl get deployments and see your new deployment.

Deployment Strategies

After defining the K8s information and Deployment Strategy Manifest files in the project configuration as laid out above, the majority of the leg work is done by FlexDeploy Pipelines, though you can accomplish the same approach through traditional Projects and Workflows.

Pipeline

Pipelines now consist of a new 'Step' called 'Apply Deployment Strategy', which we can see used below throughout the stages with the K8s icon.  On the step you can set the Deployment Strategy you want to use (BlueGreen, Canary, or AB) and what weights or properties you want to apply.

For more specific information on the Apply Deployment Strategy Step see here.

The below pipeline is using the Canary Strategy updating the weights as it progresses through the stages, starting at 10-90 and ultimately finishing with a full cutover at 100 - 0. 


Workflows

To use deployment strategies within FlexDeploy you need to use the Kubernetes Deploy Plugin operation within your workflow.  If you have your Deployment Strategy information configured in the Pipeline as outlined above, you can defer the setting of those inputs in the workflow.

Helm Configuration

In order to deploy a Helm chart to a K8s cluster, FlexDeploy uses the deploy operation of Helm plugin, which invokes helm upgrade --install command under the hood. So, the deployment workflow should contain a step as in this example:


If Chart Name and Release Name are unspecified, FlexDeploy will take the values from the Helm Configuration tab.


FlexDeploy overrides values of the chart with the values configured in the Chart Values list. These values are specified with Groovy scripts evaluated on-the-fly during the deployment. The scripts may reference any Project or Environment-Instance (available for the project) properties.


There is a smart suggester of available properties. Just start typing something from a property name and hit Ctrl-Space.


The following macros are not currently supported in the footer:
  • style