Google Translate Clone (DevOps Project)

1-Application

• This project is part of GL5 DevOps Lab. We are meant to create a distributed application, then use it with other tools in order to understand different DevOps concepts.

• Google Translate Clone is what its name implies. It consists of a google-like frontend made with react and 2 microservices made with flask, one for text translation and language detection and the other for text to speech.

• The front was based on TheScientist01's work.

• Text translation api uses the googletrans package.

• Text to speech api uses the gtts package (google text to speech)

You can access the app here

ToDo DevOps

• Observability
  • Metrics (Prometheus + Grafana)
    • General Metrics
      • Total number of requests metric
      • Success labels
      • Endpoint label
    • Business logic metric
      • Number of english translations
  • Logs (Datadog)
    • Request_id
    • Client_ip
  • Trace (Signoz)
    • Generate traces for handlers
    • Request_id and client_ip
  • Helm
    • Helm chart
    • Values.yaml
    • Labels addtion
    • Env variables for open telemetry
• Deployment (Kubernetes)
  • Microservices
  • Deployment strategy
  • Configurations and secrets
  • Network architecture / Network policies
  • 1 Endpoint exposed publicly
• Automation
  • Infrastructure provisioning (Terraform)
  • Initialization for workloads
  • Multi-environment setup
  • Microstacks
  • Deployment

2-Observability

1-Logging

• I used python's logging package to add labels to the logs such as the client ip and a unique generated request id.
• The logs are streamed to stdout and datadog agent will take care of collecting every log.
• We can then filter through the logs based on the labels we defined.

2-Metrics

• I used prometheuse client to save multiple metrics in both of the apis.
• I used generic metrics to keep track of total number of requests (successful and failed) with a Counter, the cpu and memory usage with Gauges and the response time of requests with a Histogram.
• I used counter to count the number of english translations. This can be automized to include all the languages and generate good data insights on people needs.
• The metrics are sent to the /metrics endpoint which prometheus collects periodically.
• Grafana uses it as a datasource to create Dashboards.

3-Traces

• I used signoz to track the traces of the requests through the apis endpoints.

3-Deployment

• I used Kubernetes to deploy the 3 image containers of my app (frontend, text translation api and text to speech api) on Azure Kubernetes Service.
• I used the default namespace and multiple replicas for the microservices.
• The app is exposed externally with ingress.
• I also deployed the needed monitoring services (Datadog, Prometheus, Grafana, Signoz).

1-Architecture

2-Deployment Strategy

I will use the A/B testing deployment strategy since it fits my needs:

• Zero downtime when deploying.
• Ability to test on a real traffic.
• Targetted Users: the app will be used by users around the globe. As we improve language translation models, we want to test them per zone to make sure it is translating properly.

4-Automation

1-CI pipeline

• I configured github actions workflows to automate the build step of CI.
• Each Workflow build a docker image and push it to my DockerHub account when it detects pushes on each app.

2-Microstacks

I used Terraform to automate the infrastructure provisioning. The terraform backend is stored on a container in Azure Storage. The goal of microstacks is to create a well isolated and maintainable infrastructure layers.

1- AKS Stack: This stack takes care of creating a resource group and AKS on Azure.

2- Infrastucture Stack: This stack takes care of provisioning the kubernetes cluster for the entire project.

3- Ingress Stack: This stack takes care of deploying ingress to the cluster.

4- Observability Stack: This stack takes care of installing our monitoring tools using the kubernetes provider and helm provider for these tools (prometheus, grafana, datadog , signoz).

3-Deployment

• To automate the deployment, I used helm to define the clusters and Kubernetes Deployments and to bring the external observability services.

• You can find the Helm chart and all the configurations here

• Values.helm:

replicaCountFront: 1
replicaCountApi: 2

image:
  tag: latest
  ttapi: melekelloumi/text_translation_api:latest
  ttsapi: melekelloumi/text_to_speech_api:latest
  frontend: melekelloumi/gtc_frontend:latest

http:
  ttport: 8081
  ttsport: 8082
  frontendport: 80
  frontendporttarget: 3000

labels:
  app: gtc-frontend
  env: production