Kubernetes Deployment Update Strategy

Kubernetes creates a second ReplicaSet, where it deploys the new pods and removes the old one from the original ReplicaSet

2 Types of Update Strategies in Kubernetes.  

Recreate: Removes the previous version and loads the new, good for development(interrupting the service). 

RollingUpdate (default): it updates to a new version gradually based on configured parameters.

The parameters are: 

maxUnavailable: The number/percentage of pods that can be unavailable during the update. 

maxSurge: The number/percentage of pods that can exceed the replicas requested. 

minReadySeconds: The number(only) of seconds to wait before the next pod’s creation. 

Kubernetes ReplicaSet selector with matchExpressions

ReplicaSet is a tool to manage the pod replicas and update strategies, and it is more flexible than the replication controller. 

It provides set-based labels selection support instead of the equality-based labels (ReplicationController). 

With equality-based (a replication controller) labels support, you can match labels only with an "is equal" or "is not equal" assertion. 

Match a group of pods that has a label zone with the value prod or test, but not both of them

With a set-based labels selection, you can express more powerful assertions, like in, not in, or exists. 

Match all the pods with the prod label in prod and test, using the matchExpressions property of the selector 

deployment_replicaset.yaml

--- 
apiVersion: apps/v1
kind: Deployment
metadata: 
  name: myapp-deployment_replicatset
spec: 
   replicas: 3  
   selector:    
     matchLabels:
       app: myapp
     matchExpressions:
       - {key: zone, operator: In, values: [prod, test]} 
   template:    
     metadata:         
       labels:           
         app: myapp           
         zone: prod           
         version: v1        
     spec:           
       containers:             
         - name: myapp-frontendmvc              
           image: makader/myapp:frontendmvc-v1              
           ports:               
             - containerPort: 80

Deploy ASP.NET CORE application with the Kubernetes Deployments

Update the application with a rolling update strategy that releases new pods more gradually, one pod at a time. 

Roll back the application to a specific version of the deploy history using the deployment. 

Run the command without --restart=Never (Deploy ASP.NET Core Application images in Kubernetes (K8s) & POD). Command to create a simple deployment 

 kubectl run frontendmvcv1 --image=makader/myapp:frontendmvc-v1 --port 80 

kubectl get deployment frontendmvcv1

Deploy ASP.NET Core Application different image verison in Kubernetes

Deploying the pods (individually or with a replication controller), using the same labels, except for the version, for which you use the value v2. 

Modified the code (frontendmvc).

Build the image, re name the image tag version 2 & publish the image to Docker Hub

docker build -t frontendmvc .
docker tag frontendmvc makader/myapp:frontendmvc-v2
docker push makader/myapp:frontendmvc-v2

Kubernetes Lessons Learned

ASP.NET CORE and Docker

ASP.NET Core Application with Docker

Add SQL Server 2017 Docker Container to ASP.NET Core MVC/Web API Project

Docker Compose

Deploy ASP.NET Core Application images in Kubernetes (K8s) & POD

Kubernetes ReplicationController

Access ASP.NET Core MVC Application with Kubernetes and NodePort

Create a Kubernetes service using a YAML

Deploy ASP.NET CORE application with the Kubernetes Deployments

Kubernetes ReplicaSet selector with matchExpressions

Kubernetes Deployment Update Strategy

Kubernetes kubectl Commands

kubectl cluster-info - To show information about the Kubernetes cluster. It's a command line tool installed with K8s that allows us to interact with a cluster

Create a Kubernetes service using a YAML

Instead of exposing directly (see the blog), create a service in a declarative way using a YAML file

forntendmvc_service .yaml

--- 
apiVersion: v1
kind: Service
metadata: 
 name: myapp-svc
spec: 
  selector:
    app:myapp
  type: NodePort
  ports:
    - port:80
      nodePort: 30001

Access ASP.NET Core MVC Application with Kubernetes and NodePort

Execute the command "kubectl describe pod myapp-rc-wl284",  each of them is running on a specific IP, internal to the node.  

The Node IP (192.168.65.3), but you can’t access it directly because the forwarding rule to send the requests from the node IP to the pod IP is missing.

Kubernetes Service Type using -type options

The command to expose the replication controller 

kubectl expose rc myapp-rc --name=myapp-rc-service --type=NodePort

The --name option sets a name for the service. 
The --type option sets the service type, and can be one of the following four values:

• ClusterIP: The default value that exposes the pods internal to the cluster. 
• NodePort: Exposes the pods as a static port on each node that contains the pods
• LoadBalancer: Exposes the pods externally using a load balancer of a cloud provider. 
• ExternalName: Available from version 1.7 of kube-dns to expose the pods through the contents of the externalName field.

Kubernetes ReplicationController

Create multiple instances of the pod using a ReplicationController, which is a master controller that replicates the specified pods, balances them on all the nodes, and reacts if some pod dies to restore the desired state.

forntendmvc_rc.yaml 

--- 
apiVersion: v1
kind: ReplicationController
metadata: 
 name: myapp-rc
spec: 
   replicas: 5  
   selector:    
     app: myapp    
   template:    
     metadata:         
       labels:           
         app: myapp           
         zone: prod           
         version: v1        
     spec:           
       containers:             
         - name:myapp-frontendmvc              
           image: makader/myapp:frontendmvc-v1              
           ports:               
             - containerPort:80

Deploy ASP.NET Core Application images in Kubernetes (K8s) & POD

Publish the image in Docker Hub

Rename the image tag from frontendmvc:v1, to <account_username>/<repository_name>:<tag> (in our case, makdockerhub/myapp:frontendmvc-v1). 

docker tag frontendmvc makader/myapp:frontendmvc-v1

Publish the image with the docker push command. 

  docker push makader/myapp:frontendmvc-v1

 

 

Now we can deploy a frontendmvc container everywhere (aws, azure, pcf, Kubernetes,...).