Django Celery Example on Kubernetes

Example used in the blog post How to Use Celery and RabbitMQ with Django

Build and publish the docker images needed [ django & celery ]

https://hub.docker.com/u/bseenu/

Ensure to create the secrets in EC2 parameter store and give the kubernetes node role

read access to the secrets

Create the kubernetes cluster using kops

Install kops, my environment is mac

brew install kops

Create the S3 bucket to store the kops state

aws s3 mb <<bucketname>>

export KOPS_STATE_STORE=s3://<<bucketname>>

kops create -f k8s-templates/cluster.yaml
kops create secret --name srini.cluster.k8s.local sshpublickey admin -i ~/.ssh/id_rsa.pub
kops update cluster srini.cluster.k8s.local --yes

The above command will create kubernetes cluster with one master and two nodes [ Master: t2.medium, Nodes: t2.micro ] and will generally take around 7-8 mins

Create an EFS Fileshare to be used as data volume

aws efs create-file-system --creation-token <<clusterfs-name>>

Make sure the efs is in the same vpc as the kubernetes cluster and whitelist the node security groups so that they can mount the efs file system

Kubernetes deployment

Set up the persistent volume claim

kubectl create -f k8s-templates/sqlite-configmap.yaml
kubectl create -f k8s-templates/sqlite-storage.yaml
kubectl create -f k8s-templates/sqlite-pv.yaml
kubectl create -f k8s-templates/sqlite-pvc.yaml

The above sets up the EFS filesystem as persistent volume which can be mounted by containers

Django deployment

kubectl create -f k8s-templates/django-deployment.yaml 

The above sets up the django containers, you can verify by checking the command kubectl get pods you should be seeing them in the running state

kubectl create -f k8s-templates/django-service.yaml

The above exposes the django deployment via ingress load balancer, you can get the ingress load balancer name by running the command kubectl describe svc/django-service, you should be able to point to the name on port 8000 and see the django page

Celery deployment

kubectl create -f k8s-templates/celery-deployment.yaml

The above sets up celery workers which refers to the same broker url as django and also uses the same data volume for writing to sqlite3 as django app reads from

Doing an upgrade

kubectl set image deployments/django-deployment django-web=<<docker-image>>

The above will bring up the new containers with the new images and terminates the old ones, and the rollout will be done in a controlled way ensuring your app is not down, In this example lets say if you have 4 containers it will do 2 at a time

Rollback an upgrade

kubectl rollout undo deployments/django-deployment

The above will rollback the recent django deployment, if you want to rollback to a specific revision you need to use the revision at the end of the command, you can use the kubectl rollout history deployments/django-deployment to see the revisions, Please not you need to use --record in your deployments for this

Deleting the cluster

kops delete cluster <<cluster-name>> --yes