在AWS中部署机器人实例

Question

I have no experience with AWS or bot deployment for production, so I'm looking for some suggestions on best practices.

The project is a simple Twitter automation bot written as a node.js application. Currently I am using Cloud9 in AWS to host it, but I feel this is likely not the most effective means.

What I need:

• Ability to easily deploy the bot/codebase.
• Multiple instances so I can deploy a new instance for each user.
• Ease of access to logs and updates.
• Usage reporting.
• Ability to tie into a front end for users.

I'd like to use AWS if possible to familiarize myself with the platform, but open to any suggestion that I can incorporate an easy workflow.

Current workflow to deploy new bot:

1. Create Cloud9 EC2 instance
2. Install dependencies
3. Git clone from repository
4. Edit configuration with users' access keys
5. Run bot from console
6. Leave running in background

This has been very easy thus far, but I just don't know if its practical. Appreciate any advice!

Answer 1

Given that the bot needs to be constantly running (ie it can't just be spun up on-demand for a couple minutes, which rules out AWS Lambda) and that each user needs their own, I'd give AWS ECS a try.

Your initial setup will look something like this:

1. First, create a Docker image to run your bot, and load it into ECR or Docker Hub .
2. Set up ECS. I recommend using AWS Fargate so you don't have to manage a VPC and EC2 instances just to run your containers. You'll want to create your task definition using your bot Docker image.
3. Run new tasks as needed using your task definition. This could be done via the AWS API, AWS SDK, in the AWS console, etc.

Updating the bots would just involve updating your Docker image and task definition, then restarting the tasks so they use the new image.

You should be able to set up logging and monitoring/alarming with CloudWatch for your ECS tasks too.

Usage reporting depends on what exactly you want to report. You may get all you need from CloudWatch events/metrics, or you may want to send data from your containers to some storage solution (RDS, DynamoDB, S3, etc.).

Tying a front end to the bots depends on how the bots are set up. If they have REST servers listening to a particular port, for example, you'd be able to hit that if they're running on ECS.