A simple example of a micro-service written using Python Asyncio and hosted on Kubernetes (Google Container Engine)
This repository has two goals:
- Provide a complete example, and later short tutorial, on how to run a Python Asyncio based application on Kubernetes / Google Container Engine.
- Document my current understanding of how best to structure the basics of a Python Asyncio application for use with Kubernetes. Specifically this refers to things like handling the shut down (SIGTERM) signal. It took me a while to figure this out so hopefully this example will save others time.
If you know of a better way to accomplish any of this please get in touch or submit a PR.
All of this can be done in the Google Developers Console. Google has good docs on this.
For later steps you'll need to remember the project name, region/zone name and cluster name.
This provides the gcloud command and Kubernetes (kubectl).
gcloud components list shows which sub-components are installed. Just install all of them.
####Fedora
dnf install docker
systemctl start docker.service
Note that the above only starts the docker service once. You need to enable the service if you want it to start on every boot.
systemctl enable docker.service
This example uses a Fedora container image as the base and installs third party Python 3.5 packages because Python 3.5 is not yet available in Fedora 23. If you want to use another distribution (why would you?) just edit the Dockerfile.
While inside the repository directory:
docker build .
When complete you will be able to see the image when listing the available Docker images.
docker images
Now we need to tag the image to upload it to the Google Container Registry (GCR). Each Google Cloud project has it's own GCR name space. Replace the image ID and Google cloud project name in the example below.
docker tag -f IMAGE_ID gcr.io/PROJECT_ID/python-asyncio-kubernetes-template:latest
You'll need to authenticate with Google and set the project and region/zone before this will work.
gcloud docker push gcr.io/PROJECT_ID/python-asyncio-kubernetes-template:latest
cloud container clusters get-credentials CLUSTER_NAME
kubectl create -f controller.json
When a replication controller and its pods are created, none of the pods are reachable. If you want to reach them you need to create a service. The example service.json file creates a load balancer service which makes the pods reachable from the Internet. If you just need them reachable within the cluster, set the type to ClusterIP.
See http://kubernetes.io/v1.0/docs/user-guide/services.html for more information.
Once the service is up you can get its IP, including the external IP Google allocates by running:
kubectl get services
Note that it seems to take a little while for the external IP address to be assigned.
kubectl get rc
kubectl get pods
kubectl logs PODNAME
kubectl scale --replicas=1 replicationcontrollers python-asyncio-kubernetes-template
Also, your Docker, Kubernetes and container logs should now be available in Google Cloud Logging. The output of the containers doesn't seem to integrate very well into Cloud Logging well though (as of 2015-11).