Chaos Toolkit Extension for AWS

This project is a collection of actions and probes, gathered as an extension to the Chaos Toolkit.

Install

This package requires Python 3.5+

To be used from your experiment, this package must be installed in the Python environment where chaostoolkit already lives.

$ pip install -U chaostoolkit-aws

Usage

To use the probes and actions from this package, add the following to your experiment file:

{
    "name": "stop-an-ec2-instance",
    "provider": {
        "type": "python",
        "module": "chaosaws.ec2.actions",
        "func": "stop_instance",
        "arguments": {
            "instance_id": "i-123456"
        }
    }
},
{
    "name": "create-a-new-policy",
    "provider": {
        "type": "python",
        "module": "chaosaws.iam.actions",
        "func": "create_policy",
        "arguments": {
            "name": "mypolicy",
            "path": "user/Jane",
            "policy": {
                "Version": "2012-10-17",
                "Statement": [
                    {
                        "Effect": "Allow",
                        "Action": [
                            "s3:ListAllMyBuckets",
                            "s3:GetBucketLocation"
                        ],
                        "Resource": "arn:aws:s3:::*"
                    }
                ]
            }
        }
    }
}

Or select one at random from an AZ:

{
    "name": "stop-an-ec2-instance-in-az-at-random",
    "provider": {
        "type": "python",
        "module": "chaosaws.ec2.actions",
        "func": "stop_instance",
        "arguments": {
            "az": "us-west-1"
        }
    }
}

That's it!

Please explore the code to see existing probes and actions.

Configuration

Credentials

This extension uses the boto3 library under the hood. This library expects that you have properly configured your environment to connect and authenticate with the AWS services.

Use default profile from ~/.aws/credentials or ~/.aws/config

This is the most basic case, assuming your default profile is properly configured in ~/.aws/credentials (or ~/.aws/config), then you do not need to pass any specific credentials to the experiment.

Use a non-default profile from ~/.aws/credentials or ~/.aws/config

Assuming you have configure a profile in your ~/.aws/credentials (or ~/.aws/config) file, you may declare it in your experiment as follows:

{
    "configuration": {
        "aws_profile_name": "dev"
    }
}

Your ~/.aws/credentials should look like this:

[dev]
aws_access_key_id = XYZ
aws_secret_access_key = UIOPIY

Or, your ~/.aws/config should look like this:

[profile dev]
output = json
aws_access_key_id = XYZ
aws_secret_access_key = UIOPIY

Assume an ARN role from a non-default profile

Assuming you have configure a profile in your ~/.aws/config file with a specific ARN role you want to assume during the run:

{
    "configuration": {
        "aws_profile_name": "dev"
    }
}

Your ~/.aws/config should look like this:

[default]
output = json

[profile dev]
role_arn = arn:aws:iam::XXXXXXX:role/role-name
source_profile = default

Assume an ARN role from within the experiment

You mays also assume a role by declaring the role ARN in the experiment directly. In that case, the profile has no impact if you also set it.

    "configuration": {
        "aws_assume_role_arn": "arn:aws:iam::XXXXXXX:role/role-name",
        "aws_assume_role_session_name": "my-chaos"
    }

The aws_assume_role_session_name key is optional and will be set to "ChaosToolkit" when not provided.

When this approach is used, the extension performs a assume role call against the AWS STS service to fetch credentials dynamically.

Pass credentials explicitely

You can pass the credentials as a secret to the experiment definition as follows:

{
    "secrets": {
        "aws": {
            "aws_access_key_id": "your key",
            "aws_secret_access_key": "access key",
            "aws_session_token": "token",
        }
    }
}

Note that the token is optional.

Then, use it as follows:

{
    "name": "stop-an-ec2-instance",
    "provider": {
        "type": "python",
        "module": "chaosaws.ec2.actions",
        "func": "stop_instance",
        "secrets": ["aws"],
        "arguments": {
            "instance_id": "i-123456"
        }
    }
}

Setting the region

In additon to the authentication credentials, you must configure the region against which you want to use.

You can either declare it at the top level of the experiment, add:

{
    "configuration": {
        "aws_region": "us-east-1"
    }
}

or

{
    "configuration": {
        "aws_region": {
            "env": "type",
            "key": "AWS_REGION"
        }
    }
}

But you can also simply set either AWS_REGION or AWS_DEFAULT_REGION in your terminal session without declaring anything in the experiment.

If none of these are set, your experiment will likely fail.

Contribute

If you wish to contribute more functions to this package, you are more than welcome to do so. Please, fork this project, make your changes following the usual PEP 8 code style, sprinkling with tests and submit a PR for review.

The Chaos Toolkit projects require all contributors must sign a Developer Certificate of Origin on each commit they would like to merge into the master branch of the repository. Please, make sure you can abide by the rules of the DCO before submitting a PR.

Develop

If you wish to develop on this project, make sure to install the development dependencies. But first, create a virtual environment and then install those dependencies.

$ pip install -r requirements-dev.txt -r requirements.txt 

Then, point your environment to this directory:

$ python setup.py develop

Now, you can edit the files and they will be automatically be seen by your environment, even when running from the chaos command locally.

Test

To run the tests for the project execute the following:

$ pytest

Add new AWS API Support

Once you have setup your environment, you can start adding new AWS API support by adding new actions, probes and entire sub-packages for those.

Services supported by boto

This package relies on boto3 to wrap the API calls into a fluent Python API. Some newer AWS services are not yet available in boto3, in that case, you should read the next section.

Let's say you want to support a new action in the EC2 sub-package.

Start by creating a new function in ec2/actions.py:

from chaoslib.types import Configuration, Secrets

from chaosaws import aws_client
from chaosaws.types import AWSResponse

def reboot_instance(instance_id: str, dry_run: bool=False,
                    configuration: Configuration=None,
                    secrets: Secrets=None) -> AWSResponse:
    """
    Reboot a given EC2 instance.
    """
    client = aws_client('ec2', configuration, secrets)
    return client.reboot_instances(InstanceIds=[instance_id], DryRun=dry_run)

As you can see, the actual code is straightforward. You first create a EC2 client and simply call the appropriate method on that client with the expected arguments. We return the action as-is so that it can be logged by the chaostoolkit, or even be used as part of a steady-state hypothesis probe (if this was a probe, not action that is).

You could decide to make more than one AWS API call but, it is better to keep it simple so that composition is easier from the experiment. Nonetheless, you may also compose those directly into a single action as well for specific use-cases.

Please refer to the Chaos Toolkit documentation to learn more about the configuration and secrets objects.

Once you have implemented that action, you must create at least one unit test for it in the tests/ec2/test_ec2_actions.py test module. For example:

from chaosaws.ec2.actions import reboot_instancex

@patch('chaosaws.ec2.actions.aws_client', autospec=True)
def test_reboot_instance(aws_client):
    client = MagicMock()
    aws_client.return_value = client
    inst_id = "i-1234567890abcdef0"
    response = reboot_instance(inst_id)
    client.reboot_instances.assert_called_with(
        InstanceIds=[inst_id], DryRun=False)

By using the built-in Python module to mock objects, we can mock the EC2 client and assert that we do indeed call the appropriate method with the right arguments. You are encouraged to write more than a single test for various conditions.

Finally, should you choose to add support for a new AWS API resource altogether, you should create the according sub-package.

Services not supported by boto (new AWS features)

If the support you want to provide is for a new AWS service that boto does not support yet, this requires direct call to the API endpoint via the requests package. Say we have a new service, not yet supported by boto3

from chaoslib.types import Configuration, Secrets

from chaosaws import signed_api_call
from chaosaws.types import AWSResponse

def terminate_worker_node(worker_node_id: str,
                          configuration: Configuration=None,
                          secrets: Secrets=None) -> AWSResponse:
    """
    Terminate a worker node.
    """
    params = {
        "DryRun": True,
        "WorkerNodeId.1": worker_node_id
    }
    response = signed_api_call(
        'some-new-service-name', path='/2018-01-01/worker/terminate',
        method='POST', params=params,
        configuration=configuration, secrets=secrets)
    return response.json()

Here is an example on existing API call (as a more concrete snippet):

from chaoslib.types import Configuration, Secrets

from chaosaws import signed_api_call

def stop_instance(instance_id: str, configuration: Configuration=None,
                  secrets: Secrets=None) -> str:
    response = signed_api_call(
        'ec2',
        configuration=configuration,
        secrets=secrets,
        params={
            "Action": "StopInstances",
            "InstanceId.1": instance_id,
            "Version": "2013-06-15"
        }
    )

    # this API returns XML, not JSON
    return response.text

When using the signed_api_call, you are responsible for the right way of passing the parameters. Basically, look at the AWS documentation for each API call.

WARNING: It should be noted that, whenever boto3 implements an API, this package should be updated accordingly, as boto3 is much more versatile and solid.

Make your new sub-package discoverable

Finally, if you have created a new sub-package entirely, you need to make its capability discoverable by the chaos toolkit. Simply amend the discover function in the chaosaws/__init__.py. For example, assuming a new eks sub-package, with actions and probes:

    activities.extend(discover_actions("chaosaws.eks.actions"))
    activities.extend(discover_probes("chaosaws.eks.probes"))