python-k8s-controller

A python tool that allows you to manage deployments and services in you kubernetes cluster.

Installation

Set up Kubernetes

You need to configure you kubernetes RBAC to run this script. Examples are given below. Make sure to change names and namespaces to suit your needs:

File: cluster-role.yaml

kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
  name: python-k8s-controller
rules:
- apiGroups: [""]
  resources: ["pods"]
  verbs: ["list", "watch"]
- apiGroups: ["extensions"]
  resources: ["deployments"]
  verbs: ["list", "watch", "patch", "create", "delete"]
- apiGroups: [""]
  resources: ["services"]
  verbs: ["list", "patch"]

File: cluster-role-binding.yaml

kind: ClusterRoleBinding
apiVersion: rbac.authorization.k8s.io/v1
metadata:
  name: python-k8s-controller
subjects:
- kind: ServiceAccount
  name: default
  namespace: <your namespace, for production don't use 'default', set up a custom namespace>
roleRef:
  kind: ClusterRole
  name: python-k8s-controller
  apiGroup: rbac.authorization.k8s.io

Configure your cluster with the above role and binding by saving the configuration to a yaml file, adapt the namespace of the binding to your needs and then run:

kubectl apply -f cluster-role.yaml
kubectl apply -f cluster-role-binding.yaml

Run using docker

If you want to run this application interactively using docker:

# display help
docker run --rm -i -t eduardrosert/python-k8s-controller python controller.py --help

# open interactive shell
docker run --rm -i -t eduardrosert/python-k8s-controller sh

Run using kubernetes

If you want to run this application interactively in your kubernetes cluster/ or your minikube installation:

# display help
kubectl run -i --tty python-k8s-controller --image=eduardrosert/python-k8s-controller --restart=Never -- python controller.py --help

# open interactive shell
kubectl run -i --tty python-k8s-controller --image=eduardrosert/python-k8s-controller --restart=Never -- sh 

Usage

usage: controller.py [-h] [--namespace NAMESPACE] [--deployment-name NAME]
                     [--trigger-rollout] [--duplicate]
                     [--label-selector LABEL_NAME=LABEL_VALUE]
                     [--trigger-smart-rollout] [--no-cleanup] [-v]

A tool to perform certain reoccuring tasks in your kubernetes cluster.

optional arguments:
  -h, --help            show this help message and exit
  --namespace NAMESPACE
                        the kubernetes namespace to work with (default:
                        'default')
  -v, --verbose         increase output verbosity

Deployments:
  Restart or duplicate deployments by name.

  --deployment-name NAME
                        the metadata.name of the deployment
  --trigger-rollout     Triggers a fresh rollout of all Pods by updating the
                        creationTimestamp of the Deployment's Pod template
  --duplicate           Duplicates the deployment. Creates a copy of the
                        deployment under a new name. The current timestamp
                        will be added to the deployment's name to create a
                        unique name, for example 'my-
                        deployment-20191011171423'. If the deployment already
                        has such a suffix, the deployment's copy will get an
                        updated timestamp.

Services:
  Rollout deployments and update associated services by label.

  --label-selector LABEL_NAME=LABEL_VALUE
                        the common label for deployment, pods, and service
                        selector, e.g. app=myapp
  --trigger-smart-rollout
                        Triggers a smart rollout with zero downtime to the
                        service: Duplicates the Deployment(s) with new name(s)
                        and a new 'patch' label, points the service(s) to the
                        new deployment as soon as all pods of the new
                        deployment are running
  --no-cleanup          Deletes the old deployments after patching the
                        service. Use in combination with --trigger-smart-
                        rollout.

Example usage:
----------------------------------------------------------------------------
  Redeploy all pods associated with 'my-deployment' using using the 
  deployment's strategy (e.g. RollingUpdate):

        python controller.py --namespace my-namespace \
                             --deployment-name my-deployment \
                             --trigger-rollout

----------------------------------------------------------------------------
  Create a copy of 'my-deployment':

        python controller.py --namespace my-namespace \
                             --deployment-name my-deployment \
                             --duplicate

  or

        python controller.py --namespace my-namespace \
                             --deployment-name my-deployment-20191011171423 \
                             --duplicate

----------------------------------------------------------------------------
  Trigger a fresh rollout of all deployments with label app=my-app and
  patch the services, then remove the old deployments:

        python controller.py --namespace my-namespace \
                             --label-selector app=my-app \
                             --trigger-smart-rollout

----------------------------------------------------------------------------
  Trigger a fresh rollout of all deployments with label app=my-app and
  patch the services, but leave old deployments in place:

        python controller.py --namespace my-namespace \
                             --label-selector app=my-app \
                             --trigger-smart-rollout
                             --no-cleanup