Table of Contents
It is still very common to develop an application locally on a laptop/desktop before pushing it to a production-like environment. The local development environment is kept as close as possible to production using technology such as Docker or AWS SAM when working with AWS Lambda. However, when working with AWS resources through Identity and Access Management (IAM) policies, local IAM permissions are typically different from the permissions the application will have in AWS. This inconsistency can cause issues later in the development workflow: an application that fully worked locally can run into errors when ran in AWS if the IAM permission there are different.
In this blog post I first zoom in into the issue a bit more and then explain how I solved this issue for myself using a simple Python utility.
IAM permissions: the challenge
It is not uncommon to work with admin-like AWS IAM permissions (for a development, testing or staging AWS account, hopefully not production!) in a local development environment. You can login to the AWS account and see and change pretty much every resource. You create STS tokens for local use, using the AWS CLI or the SDK in your applications. If these applications use other AWS resources such as an SQS queue or a DynamoDB table, they have no problem connecting to these resources because the application is using your admin-like permissions. Everything works and is easy, fine and happy.
You then push your application to AWS where it runs as a Lambda function or within an EC2 instance. Of course, you follow security’s best practices and apply the least-privilege principle to all your AWS resources. Your resources only have exactly the permissions it needs to connect to other resources.
This is an annoying inconsistency between your local development environment and the first stage of pushing your application to an actual AWS account. Applying the principle of early feedback, the sooner you learn your IAM permissions are off, the better.
The solution is simple, really: assume the role that your application (Lambda / EC2) is going to assume in AWS and use it while running the application in your local development environment. All that is needed is a simple utility that makes it easy to switch between different IAM roles. As I searched around for possible solutions, my requirements were;
- Make it easy to switch between different IAM roles. When working on a number of Lambda functions, each of these functions might have their own IAM role. It should therefore be easy to switch between the different roles used by the Lambda function.
- An intuitive, easy-to-remember command line interface. I don’t want to copy/paste my role ARN each time or look in my bash history for the correct role when I need to assume a different role. Instead, I prefer to use an alias to easily switch to a previously-configured role.
- Less is more. Just a single utility with a single purpose.
As I wasn’t able to find a tool fulfilling these requirements, I put one together myself. With two simple steps you can use it to easily switch between roles.
Configure your AssumeRolePolicyDocument
First, you will need to edit the
AssumeRolePolicyDocument for the role you are going to assume from your local development environment. The following JSON is a default Lambda
AssumeRolePolicyDocument including an additional line that gives my development role permissions to assume this role.
Important: be sure never to give permissions such as these to a resource in a production account. This is a huge risk and opens up a simple way to mistakenly change resources in your production account, and opens up the potentials for abusing these permissions to retrieve sensitive data.
Install and use the
Next, install my assume utility with pip:
pip install assume
Now, say you have two different Lambda functions that both use a different IAM role:
publisherfunction using IAM role
subscriberfunction using IAM role
First, make sure to edit the
AssumeRolePolicyDocument for these roles as described above. You need to give the role or user that you typically login with to have permissions to assume this role.
Next, let’s add these roles to our
assume add publisher --role-arn arn:aws:iam::012345678912:role/publisher
assume add subscriber --role-arn arn:aws:iam::012345678912:role/subscriber
You can optionally use the
--profile flag to use a profile other than the default profile for assuming the role.
Now you can easily switch between the two different roles. For example, run the following command to assume the
assume switch subscriber
To clear any role and switch back to your default role, run the
Check out the readme in the GitHub repository for the other commands.
Increase your development cycle and receive early feedback regarding IAM permissions. This solution should make it easier to spot IAM permission errors earlier in the development workflow, and will make it easier to implement strict, least-privilege IAM permissions for your AWS resources.
Have you ran into this issue before? Do you use a similar or a completely different solution? I’m interested in other approaches so definitely let me know!