aws-okta-processor

This package provides a command for fetching AWS credentials through Okta.

Installation

The recommended way to install aws-okta-processor is using pipx. This has the benefit that the command is available in your shell without needing to activate a virtualenv while still keeping its dependencies isolated from site-packages:

$ pipx install aws-okta-processor

and, to upgrade to a new version:

$ pipx upgrade aws-okta-processor

You can also install with pip in a virtualenv:

$ pip install aws-okta-processor

or, if you are not installing in a virtualenv, to install globally:

$ sudo pip install aws-okta-processor

or for your user:

$ pip install --user aws-okta-processor

If you have aws-okta-processor installed with pip and want to upgrade to the latest version you can run:

$ pip install --upgrade aws-okta-processor

$ sudo pip install aws-okta-processor --ignore-installed six

This will install the aws-okta-processor package as well as all dependencies. You can also just download the tarball. Once you have the aws-okta-processor directory structure on your workstation, you can just run:

$ cd 
$ python setup.py install

Getting Started

This package is best used in AWS Named Profiles with tools and libraries that recognize credential_process.

To setup aws-okta-processor in a profile create an INI formatted file like this:

[default]
credential_process=aws-okta-processor authenticate --user  --organization .okta.com

and place it in ~/.aws/credentials (or in %UserProfile%\.aws/credentials on Windows). Then run:

$ pip install awscli
$ aws sts get-caller-identity

Supply a password then select your AWS Okta application and account role if prompted. The AWS CLI command will return a result showing the assumed account role. If you run the AWS CLI command again you will get the same role back without any prompts due to caching.

For tools and libraries that do not recognize credential_process aws-okta-processor can be ran to export the following as environment variables:

AWS_ACCESS_KEY_ID
AWS_SECRET_ACCESS_KEY
AWS_SESSION_TOKEN

For Linux or OSX run:

$ eval $(aws-okta-processor authenticate --environment --user  --organization .okta.com)

On Unix systems pass a --target-shell in order to change the export command output. Bash is the default target shell. We also allow [fish shell]() as a valid target:

$ eval (aws-okta-processor authenticate --environment --user  --organization .okta.com --target-shell fish)

For Windows run:

$ Invoke-Expression (aws-okta-processor authenticate --environment --user  --organization .okta.com)

Other Configurable Variables

Additional variables can also be passed to aws-okta-processors authenticate command as options or environment variables as outlined in the table below.

Examples

If you do not want aws-okta-processor to prompt for any selection input you can export the following:

$ export AWS_OKTA_APPLICATION= AWS_OKTA_ROLE= AWS_OKTA_FACTOR=

Or pass additional options to the command:

$ aws-okta-processor authenticate --user  --organization .okta.com --application  --role  --factor 

Caching

This package leverages caching of both the Okta session and AWS sessions. It's helpful to understand how this caching works to avoid confusion when attempting to switch between AWS roles.

Okta

When aws-okta-processor attempts authentication it will check ~/.aws-okta-processor/cache/ for a file named --session.json based on the user and organization option values passed. If the file is not found or the session contents are stale then aws-okta-processor will create a new session and write it to ~/.aws-okta-processor/cache/. If the file exists and the session is not stale then the existing session gets refreshed.

AWS

After aws-okta-processor has a session with Okta and an AWS role has been selected it will fetch the role's keys and session token. This session information from the AWS role gets cached as a json file under ~/.aws/boto/cache. The file name is a SHA1 hash based on a combination the user, organization and key option values passed to the command.

If you want to store a seperate AWS role session cache for each role assumed using the same user and organization option values then pass a unique value to key. Named profiles for different roles can then be defined in ~/.aws/credentials with content like this:

[role_one]
credential_process=aws-okta-processor authenticate --user  --organization .okta.com --application  --role  --factor  --key role_one

[role_two]
credential_process=aws-okta-processor authenticate --user  --organization .okta.com --application  --role  --factor  --key role_two

To clear all AWS session caches run:

$ rm ~/.aws/boto/cache/*

Assuming a Secondary Role

If you can only assume a role from another role, you can assume both roles using --role and --secondary-role. Use --role to specify the first role ARN, then --secondary-role to specify the role ARN assumed from --role.

Example:

aws-okta-processor authenticate --user jdoe ... --role arn:aws:iam::111111111:role/OpsUser --secondary-role arn:aws:iam::111111111:role/SecretsAdmin

Project or User Configuration

aws-okta-processor can inherit arguments from a .awsoktaprocessor file located in the user's home directory or the current working directory.

.awsoktaprocessor

[defaults]
user=jdoe

[authenticate]
user=ssmith

In this example...

• authenticate > user overrides defaults > user
• {workingDir}/.awsoktaprocessor overrides ~/.awsoktaprocessor
• aws-okta-processor arguments override any options from dotfiles

Get Roles

To get roles, use the get-roles command. This command supports outputing the roles as AWS profiles, JSON, or custom formatted text.

# write all the roles as AWS profiles
aws-okta-processor get-roles -u jdoe -o mycompany.okta.com --output=profiles > ~/.aws/credentials

# get account and role
aws-okta-processor get-roles -u jdoe -o mycompany.okta.com --output=text --output-format="{account},{role}"

# get JSON
aws-okta-processor get-roles -u jdoe -o mycompany.okta.com --output=json

Output Types

• json (default): output as JSON
• profiles: output AWS profiles to be stored in ~/.aws/credentials
• text: custom formatted text using --output-format and tokens

Output Format Tokens

• {account}: name of the account
• {account_id}: account Id
• {account_raw}: account information as seen on Okta site (Account: blah-blah (id))
• {application_url}: full Okta application url
• {organization}: organization as provided
• {role}: role ARN
• {role_suffix}: last element of the role (delimited using AWS_OKTA_ROLE_SUFFIX_DELIMITER or -)
• {user}: user as provided

Getting Help

• If it turns out that you may have found a bug, please open an issue

Acknowledgments

This package was influenced by AlainODea's work on okta-aws-cli-assume-role.