Deploying RASA bot on ec2

I am aware of the pros and cons of deploying rasa assistant in production using Docker compose, k3s and k8s thanks to this detailed blog post.

Now, I am curious how deploying in an EC2 instance by imitating the methodology used in local installation would fair.

What I am suggesting is:

  1. Create a python environment inside the instance
  2. Pip install rasa and rasa X
  3. Move the folder containing code there from local and train a model.
  4. Open 5005 port of the instance and access it from API gateway.
  5. Similarly, access rasa X in the browser
  6. Attach a load balancer for traffic distribution across multiple identical instances.

To my understanding, it will make the deployment much easier. I am specifically interested in knowing cons of using this strategy. Thank you

I tried this approach but it has been quite a while ago and it was a much bigger effort than either the standard docker-compose or k3s setup. Maybe you’ll have a different experience. Let us know.

1 Like

Can you please enlighten me on how this deployment will impact the availability and scalability of the bot.

If your concerned about availability and scalability it’s hard to ignore the advantages of a Kubernetes based deployment. k8s is designed for this purpose.

docker-compose has some features around scalability but nothing like k8s.