Deploying in AWS using Terraform and S3FS as a Mountpoint

[ricardoteix]

I utilized this project as a component of an AWS course. It required the creation of a media player portal, and after conducting some research, I came across MediaCMS. Although it wasn't subjected to extensive testing with a large volume of videos or simultaneous users, it served its purpose for a study project.

I currently don't have it up and running since it's no longer needed, but I have the project documentation in a GitHub repository (here), and you can find an image illustrating the AWS services used in this repository or at the end of this post.

Since it was built with Terraform, it's relatively easy to replicate the exact infrastructure I created following the instructions in README.

Certain decisions were made due to the necessity of utilizing AWS infrastructure:

• Server: I deployed it with an AutoScaling Group comprising 2 active EC2 instances and others in a stopped state for processing uploaded media. This approach ensured that the two active instances wouldn't be overwhelmed with encoding tasks. A Load Balancer was also employed.
• Each EC2 instance has an S3FS installation mounting an S3 bucket in the "/media_files/" folder.
• A CloudFront distribution was established to serve all media playback directly from the bucket.
• I used the SSM Parameter Store to store credentials and some other configurations, which can be retrieved by each EC2 instance.
• Database: I transitioned to an RDS instance.
• Redis: I switched to using AWS Memcached with a Redis node.

All changes in the repository were made using a shell script (here) from the tag 3.0.0 in git repo (https://github.com/mediacms-io/mediacms).

Since I'm utilizing the Parameter Store for all database credentials and other settings, I needed to adapt the settings.py file to retrieve this information. The Python code changes were as follows:

import boto3
import json

DEBUG = False

ssm_client = boto3.client("ssm", region_name='us-east-1')

response = ssm_client.get_parameter(Name="mediacms", WithDecryption=False)
parameter = json.loads(response["Parameter"]["Value"])"

For every instance where information or credentials from the Parameter Store needed to be utilized, they were replaced in the shell script with something like parameter["rds_addr"], as demonstrated below:

DATABASES = {
    "default": {
        "ENGINE": "django.db.backends.postgresql",
        "NAME": "mediacms",
        "HOST": parameter["rds_addr"],
        "PORT": "5432",
        "USER": "mediacms",
        "PASSWORD": "mediacms",
    }
}