s0nar

Anomaly detection service 🔎👻

Deployment notes

First of all, download the .env vars for production deployment at following link:

• s3://s0nar/env/master/.env

Create s0nar API and workers images

docker-compose build --no-cache .   # In project's root folder

Then start all containers using docker-compose, except LSTM-GPU worker:

docker-compose up -d api mongodb rabbitmq flower arima-worker lstm-cpu-worker

We don't start the gpu worker because docker-compose has no support for nvidia-docker flags, then:

docker run --name lstm-gpu-worker --gpus all s0nar_lstm-gpu-worker:latest

Embedded postman collection:

First, establish the env vars and deploy mongodb and rabbitmq (and flower if you want to track the celery task)

cp .env.sample .env # maybe you need to update these variables
docker-compose up -d mongodb rabbitmq flower

Then start s0nar service

python app.py

Finally, start the workers:

• auto-arima worker

celery --app=project.ml_task.celery_app worker -l INFO -n worker.arima -Q arima -c 1

• LSTM-CPU worker

celery --app=project.ml_task.celery_app worker -l INFO -n worker.lstm-cpu -Q lstm-cpu -c 1

• LSTM-CPU worker

celery --app=project.ml_task.celery_app worker -l INFO -n worker.lstm-gpu -Q lstm-gpu -c 1

Once both services are running you can test it using postman or cURL.

Development requirements

• python >=3.5 (using conda, virtenv, etc, it doesn't matter )
• pip3
• docker
• docker-compose
• cURL or Postman