Atlantis in ECS Fargate fails to get custom aws provider - ERROR: failed to get shared config profile

[gatemken]

Terraform Version

Error running plan operation: running \"/home/atlantis/.atlantis/bin/terraform1.8.1 init -input=false -upgrade\" in \"/home/atlantis/.atlantis/repos/gsen/tftest/3/default/myrepotest/us-east-1/ecr/test\": exit status 1\n\nInitializing the backend...\n╷\n│ Error: failed to get shared config profile,

Terraform Configuration Files

provider.tf
terraform {
required_providers {
aws = {
source = "hashicorp/aws"
}
}
}
provider "aws" {
profile = "my-test"
region = "us-east-1"
}

backend.tf
terraform {
required_version = ">= 0.12.2"

backend "s3" {
region = "us-east-1"
bucket = ""
key = ""
dynamodb_table = ""
profile = "my-test"
encrypt = "true"
}
}

Create any test resource in your main.tf

Debug Output

Works fine when ran locally

Expected Behavior

ECS fargate should be able to find the profiles listed in ~/.aws/config

Actual Behavior

ECS on FARGATE fails to read profiles in ~/.aws/config

Steps to Reproduce

Deploy atlantis in ecs fargate using the terraform-atlantis-ecs module
setup gitlab or github webhooks
create your backend.tf and provider.tf and make sure profile are set.
i.e
provider "aws" {
profile = "my-test"
region = "us-east-1"
}

Removing profile = "my-test" works. keeping it leads to an error

Additional Context

Note: everything works if i don't specific a profile. For our use case, we need to be able to run terraform code locally and with atlantis. Specifying a provider is a MUST.

References

No response

[jbardin]

Hi @gatemken,

You mention that removing the profile entry in the provider allows the configuration to work, but the error you presented indicates there was a problem with the backend configuration. Can you confirm which (or both) of the components is failing here?

Thanks!

[gatemken]

@jbardin it is both. Specifying the profile keyword in providers.tf or backend.tf causes that error. not specifying the profile flag leads to successful plan/apply.

This is my Dockerfile:

FROM ghcr.io/runatlantis/atlantis:latest
COPY aws_config /home/atlantis/.aws/config

cat aws_config
[profile my-test]
role_arn = arn:aws:iam::account-id:role/my-aws-admin-role
credential_source = EcsContainer

terraform version 1.8.1 runs on the ecs container by atlantis.

I have try setting ENV in the ecs container and in the docker container but it still fails. you should be able to reproduce the issue.

FYI: This works perfectly fine, if i deploy atlantis on EC2 instance.

[gatemken]

I need this to work on EC2 Fargate not EC2 Instance. I would appreciate it if someone can reproduce the issue and work out a possible fix or workaround

[gatemken]

This issue can be closed