A multi container app deployed on a Kubernetes cluster

As a non-German speaker living in Switzerland, I often need to quickly translate large texts, but I get annoyed by character limits on Google Translate or DeepL. Learning German may have been a way better call, but instead I decided to deploy a translation application. It's made of 3 containerized microservices:

• A Flask frontend to get inputs and display translations
• A FastAPI API backend to translate English text, using open-source models (SpaCy, Hugging Face)
• A MySQL database to store previous translations

In this repo, we deploy the app on Kubernetes cluster, see also this blog post.

For the deployment on a single node with Docker Compose, see this repo and this blog post.

Deployment on a Kubernetes cluster

Run the following commands in your Kubernetes cluster:

1. Create the secrets for the database's credentials:

kubectl create secret generic mysql-db-secret \
  --from-literal=MYSQL_ROOT_PASSWORD='<root_password>' \
  --from-literal=MYSQL_USER='<user>' \
  --from-literal=MYSQL_PASSWORD='<password>'

2. Deploy the app

kubectl apply -f unlimited-translation-k8s-dev.yaml

3. Connect to the app through the IP of the frontend-flask-service Service:

kubectl get svc -o wide

Note 1: the manifest used in the above command requests a LoadBalancer service and an automatically provisioned PersistentVolume, so it is expected to be applied on a cluster managed by a cloud provider (unless you implemented these features yourself... I used GKE).

Note 2: the unlimited-translation-k8s.yaml contains an additional Ingress to publish the app on my personal domain, translate.vlgdata.io