A python microservice framework


Quick start
Project files
OpenAPI specification
API objects
Server code
Deployment pipeline
Docker packaging
JWT authentication
Error handling
Asynchronous execution
Database serialisation

Deployment to Google Kubernetes Engine

PyMacaron microservices can be deployed to GKE clusters using the pymacaron-gcp addon.


You will need to have:

Creating the staging cluster

PyMacaron does not create the staging cluster. You should do it yourself and name it ‘-staging'. The recommended setup for the staging cluster is a 1 zone cluster, with autoscaling from 1 to 2 instances (to avoid resource exhaustion during new deploy), using a cheap low mem/low cpu instance type.

PyMacaron does however configure the staging cluster, by creating a deployment, adding a service and an ingress on the first deploy, then updating the deployment on all further deploys.

Creating and configuring the live cluster

PyMacaron does not create nor configure the live cluster. It only does a rolling deploy towards it by applying a deployment named ‘-live'.

You should therefore create the cluster yourself using whatever tool you want, but make sure that all resources you deploy share the app name ‘-live'.

Deployment pipeline

When deploying to GKE, ‘pymdeploy’ goes through the following specific deployment steps:

  1. Deploy the docker image as a new deployment to a google GKE cluster named after your app’s name suffixed with ‘-staging’. Optionally add a service and an ingress to expose the staging environment to internet.

  2. Run the acceptance tests against the staging environment. Stop if tests fail.

  3. Re-deploy the docker image as a new deployment to the live GKE cluster.

Configuration in pym-config.yaml

To be able to deploy against GKE, the following key-values must be present in your project’s pym-config.yaml:

docker_repo: <PROJECT_ID>      # The ID of your gcp project
deploy_target: gke

See here for details on ‘pym-config.yaml’.