image-cloner

A k8s controller which watches the applications and caches the images by re-uploading to our own registry and re-configures the applications to use these copies.

Prerequisits

1. A cluster is required, it can be a managed cluster like GKE etc. or a local kind or minikube cluster.
   The makefile contains the way to create the kind cluster with the following command. If you don't have kind installed:

$ make install-kind

$ make cluster

2. The operator need a secret which holds the registry creds to the private docker registry where all the image will be cached.
   This has to be created upfront before deploying the operator to your cluster. Below are the steps

apiVersion: v1
kind: Secret
metadata:
  name: private-registry-creds
  namespace: system
type: Opaque
stringData:
  DOCKER_REGISTRY_URI: <registry_name/repo_name>  # example "index.docker.io/ishankhare07"
  DOCKER_REGISTRY_USERNAME: <username>
  DOCKER_REGISTRY_PASSWORD: <password>

And then create this secret in the cluster:

$ kubectl apply -f secret.yaml

Use your registry to host operator

The controller image is currently available as a public image on dockerhub at ishankhare07/image-cloner, in case you want to host your own image
build and push to your registry in the following way:

$ docker login <your_registry>
...
...
...

$ IMG="<your_registry/username>" make build
$ IMG="<your_registry/username>" make push
$ IMG="<your_registry/username>" make deploy

Use existing registry image

If you are okay with using the current image on the public dockerhub, just go ahead and deploy:

$ make deploy

This will apply the following to your cluster:

1. Create namespace system
2. Create clusterrole for the controllers
3. Create clusterrolebindings
4. Create deployment which will actually run the controller.

Running locally

When running locally, create a file called config/config.env which contains the private registry cred as follows:

DOCKER_REGISTRY_URI="<registry_name/repo_name>"
DOCKER_REGISTRY_USERNAME="<username>"
DOCKER_REGISTRY_PASSWORD="<password>"

This allows easily testing the whole flow, next just run:

$ make cluster
$ make run