Kubernetes Quick Dive
In this lab you are going to deploy the instavote application stack as described here in a kubernetes environment using kubectl commands. Later, you would learn how to do the same by writing declarive yaml syntax.
Purpose of this lab is to quickly get your app up and running and demonstrate kubernetes key features such as scheduling, high availability, scalability, load balancing, service discovery etc.
Deploying app with kubernetes
Before launching the app, create a new namespace and switch to it
kubectl get ns
kubectl create namespace instavote
kubectl get ns
kubectl config get-contexts
kubectl config set-context --current --namespace=instavote
kubectl config get-contexts
Launch vote application with kubernetes in the newly created namespace.
kubectl create deployment vote --image=schoolofdevops/vote:v4
You could now validate that the instance of vote app is running by using the following commands,
kubectl get pods
kubectl get deploy
kubectl get all
Scalability
Scale the vote app to run 4 instances.
kubectl scale deployment vote --replicas=4
kubectl get deployments,pods
High Availability
kubectl get pods
The above command will list pods. Try to delete a few pods and observe how it affects the availability of your application.
kubectl delete pods vote-xxxx vote-yyyy
kubectl get deploy,rs,pods
Load Balancing with Services
Publish the application (similar to using -P for port mapping)
kubectl create service nodeport vote --tcp=80:80 --node-port=30300
kubectl get service
Connect to the app, refresh the page to see it load balancing. Also try to vote and observe what happens.
Roll Out a New Version
kubectl scale deployment vote --replicas=12
kubectl set image deployment vote vote=schoolofdevops/vote:v5
watch the rolling update in action
kubectl rollout status deploy/vote
Launching and connecting to Redis
If you try to connect to the frontend vote app (available on port 30300) created above, andtry to submit the vote, it thrown an error.
Thats because there is no backend service redis available.
You couldcreate a redis service with clusterip and the deployment for it using the following commands,
kubectl create service clusterip redis --tcp=6379
kubectl create deployment redis --image=redis:alpine
validate pods, deployment and service for redis is available, and service has discovered the pods as endpoints with,
kubectl get all,ep
[sample output]
NAME READY STATUS RESTARTS AGE
pod/redis-59bc74f5b5-fgv9m 1/1 Running 0 8m45s
pod/vote-5cbbf546d8-9c624 1/1 Running 0 41m
pod/vote-5cbbf546d8-zwg89 1/1 Running 0 39m
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/redis ClusterIP 10.96.179.131 <none> 6379/TCP 9m11s
service/vote NodePort 10.96.233.140 <none> 80:30300/TCP 51m
NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/redis 1/1 1 1 8m45s
deployment.apps/vote 2/2 2 2 66m
NAME DESIRED CURRENT READY AGE
replicaset.apps/redis-59bc74f5b5 1 1 1 8m45s
replicaset.apps/vote-5cbbf546d8 2 2 2 41m
replicaset.apps/vote-75cfd886bc 0 0 0 43m
replicaset.apps/vote-9c55988f4 0 0 0 66m
NAME ENDPOINTS AGE
endpoints/redis 10.244.2.33:6379 9m11s
endpoints/vote 10.244.1.30:80,10.244.2.27:80 51m
If you see output similar to above, redis is now setup along with vote and at this time, if you submit the vote, it should go through and show you a check mark.
Exercise - Deploy Complete Instavote App
Deploy the additional services with the following spec to complete this application stack.
| Service Name | Image | Service Type | Service Port | Node Port |
|---|---|---|---|---|
| worker | schoolofdevops/worker:latest | No Service Needed | N/A | N/A |
| db | postgres:9.4 | ClusterIP | 5432 | N/A |
| result | schoolofdevops/vote-result | NodePort | 80 | 30400 |
If you see db deployment failing, fix it by adding the environment var as,
kubectl set env deployment db POSTGRES_HOST_AUTH_METHOD=trust
After deploying all services to validate,
- Browse to vote and result services exposed outside to see the UI
- When you submit a vote, it should be reflected on result
- To submit multiple votes, use either a different browser, or use incognito window.
Cleaning up
Once you are done observing, you could delete it with the following commands,
kubectl delete deploy vote redis worker db result
kubectl delete service vote redis db result
Summary
When you deploy an application in kubernetes, you submit it to the api server/ cluster manager. Kubernetes automatically schedules it on a cluster, networks the pods, provides service discovery. In addition as you observed, your application is scalable, high available and is already running behind a load balancer.