AWS_terraform.md

Ml Flow on AWS setup

AWS CLI

• Install AWS CLI - link to instruction
• Generate key for the user that you will be using to manage terraform resources - link to the instruction
• Login using aws configure command

Terraform setup

• Terraform version used here : v1.5.0
• Install terraform - link to the official documentation
• Install terraform - docs for automatic documentation (optional) terraform-docs

Create bucket for terraform state

• For terraform to work optimally we should store our terraform state in cloud. You can store it locally but it is not recommended. Details about terraform state.
• Go to Amazon S3 and click Create Bucket
• Choose a name and region for your bucket
• You can leave the rest of the settings default

Initial terraform setup

• change bucket value in main.tf file to name of the bucket you just created
• adjust variables.tf and local.tf so the data matches your project, if you are not using vpn use 0.0.0.0/0 instead (it will make your application available from any IP)
• run terraform init

Documentation for terraform elements HERE

Create ECR and push image

• create ECR using command terraform apply -target=aws_ecr_repository.mlflow_ecr
• build image based on Dockerfile from /mlflow-gcp/AWS
• upload image to ECR with tag latest - official instruction

Commands for pushing to ECR

Change region, aws_account_id, repository_name and imgine_id so it matches your project.

• authenticate docker for aws

aws ecr get-login-password --region region | docker login --username AWS --password-stdin aws_account_id.dkr.ecr.region.amazonaws.com

• tag your imagine

docker tag imgine_id aws_account_id.dkr.ecr.region.amazonaws.com/repository_name:latest

• push your imagine to AWS

docker push aws_account_id.dkr.ecr.region.amazonaws.com/repository_name:latest

Set up rest of infrastructure

• use terraform plan to review what elements will be created
• use terraform apply to set up the rest of the infrastructure (it will take a while)

Testing

• in folder AWS/sample_client create .env file base on template
• check_tracking.py - uploads an artifact to the bucket, it will be visible in MlFlow GUI
• upload_model.py - uploads more advanced model