skew

Skew is a package for identifying and enumerating cloud resources. The name is a homonym for SKU (Stock Keeping Unit). Skew allows you to define different SKU schemes which are a particular encoding of a SKU. Skew then allows you to use this scheme pattern and regular expressions based on the scheme pattern to identify and enumerate a resource or set of resources.

At the moment, the the only available scheme is the ARN scheme. The ARN scheme uses the basic structure of Amazon Resource Names (ARNs) to assign a unique identifier to every AWS resource.

An example ARN pattern would be:

arn:aws:ec2:us-west-2:123456789012:instance/i-12345678

This pattern identifies a specific EC2 instance running in the us-west-2 region under the account ID 123456789012. The account ID is the 12-digit unique identifier for a specific AWS account as described here. To allow skew to find your account number, you need to create a skew YAML config file. By default, skew will look for your config file in ~/.skew but you can use the SKEW_CONFIG environment variable to tell skew where to find your config file if you choose to put it somewhere else. The basic format of the skew config file is:

---
  accounts:
    "123456789012":
      profile: dev
    "234567890123":
      profile: prod

Within the accounts section, you create keys named after your 12-digit account ID (as a string). Within that, you must have an entry called profile that lists the profile name this account maps to within your AWS credential file.

The main purpose of skew is to identify resources or sets of resources across services, regions, and accounts and to quickly and easily return the data associated with those resources. For example, if you wanted to return the data associated with the example ARN above:

from skew import scan

arn = scan('arn:aws:ec2:us-west-2:123456789012:instance/i-12345678')
for resource in arn:
    print(resource.data)

The call to scan returns an ARN object which implements the iterator pattern and returns a Resource object for each AWS resource that matches the ARN pattern provided. The Resource object contains all of the data associated with the AWS resource in dictionary under the data attribute.

Any of the elements of the ARN can be replaced with a regular expression. The simplest regular expression is * which means all available choices. So, for example:

arn = scan('arn:aws:ec2:us-east-1:*:instance/*')

would return an iterator for all EC2 instances in the us-east-1 region found in all accounts defined in the config file.

To find all DynamoDB tables in all US regions for the account ID 234567890123 you would use:

arn = scan('arn:aws:dynamodb:us-.*:234567890123:table/*')

CloudWatch Metrics

In addition to making the metadata about a particular AWS resource available to you, skew also tries to make it easy to access the available CloudWatch metrics for a given resource.

For example, assume that you had did a scan on the original ARN above and had the resource associated with that instance available as the variable instance. You could do the following:

>>> instance.metric_names
['CPUUtilization',
 'NetworkOut',
 'StatusCheckFailed',
 'StatusCheckFailed_System',
 'NetworkIn',
 'DiskWriteOps',
 'DiskReadBytes',
 'DiskReadOps',
 'StatusCheckFailed_Instance',
 'DiskWriteBytes']
>>>

The metric_names attribute returns the list of available CloudWatch metrics for this resource. The retrieve the metric data for one of these:

>>> instance.get_metric_data('CPUUtilization')
[{'Average': 0.134, 'Timestamp': '2014-09-29T14:04:00Z', 'Unit': 'Percent'},
 {'Average': 0.066, 'Timestamp': '2014-09-29T13:54:00Z', 'Unit': 'Percent'},
 {'Average': 0.066, 'Timestamp': '2014-09-29T14:09:00Z', 'Unit': 'Percent'},
 {'Average': 0.134, 'Timestamp': '2014-09-29T13:34:00Z', 'Unit': 'Percent'},
 {'Average': 0.066, 'Timestamp': '2014-09-29T14:19:00Z', 'Unit': 'Percent'},
 {'Average': 0.068, 'Timestamp': '2014-09-29T13:44:00Z', 'Unit': 'Percent'},
 {'Average': 0.134, 'Timestamp': '2014-09-29T14:14:00Z', 'Unit': 'Percent'},
 {'Average': 0.066, 'Timestamp': '2014-09-29T13:29:00Z', 'Unit': 'Percent'},
 {'Average': 0.132, 'Timestamp': '2014-09-29T13:59:00Z', 'Unit': 'Percent'},
 {'Average': 0.134, 'Timestamp': '2014-09-29T13:49:00Z', 'Unit': 'Percent'},
 {'Average': 0.134, 'Timestamp': '2014-09-29T13:39:00Z', 'Unit': 'Percent'}]
>>>

You can also customize the data returned rather than using the default settings:

>>> instance.get_metric_data('CPUUtilization', hours=8, statistics=['Average', 'Minimum', 'Maximum'])
[{'Average': 0.132,
  'Maximum': 0.33,
  'Minimum': 0.0,
  'Timestamp': '2014-09-29T10:54:00Z',
  'Unit': 'Percent'},
 {'Average': 0.134,
  'Maximum': 0.34,
  'Minimum': 0.0,
  'Timestamp': '2014-09-29T14:04:00Z',
  'Unit': 'Percent'},
  ...,
 {'Average': 0.066,
  'Maximum': 0.33,
  'Minimum': 0.0,
  'Timestamp': '2014-09-29T08:34:00Z',
  'Unit': 'Percent'},
 {'Average': 0.134,
  'Maximum': 0.34,
  'Minimum': 0.0,
  'Timestamp': '2014-09-29T08:04:00Z',
  'Unit': 'Percent'}]
>>>

Filtering Data

Each resource that is retrieved is a Python dictionary. Some of these (e.g. an EC2 Instance) can be quite large and complex. Skew allows you to filter the data returned by applying a jmespath query to the resulting data. If you aren't familiar with jmespath, check it out. Its a very powerful query language for JSON data and has full support in Python as well as a number of other languages such as Ruby, PHP, and Javascript. It is also the query language used in the AWSCLI so if you are familiar with the --query option there, you can use the same thing with skew.

To specify a query to be applied to results of a scan, simply append the query to the end of the ARN, separated by a | (pipe) character. For example:

arn:aws:ec2:us-west-2:123456789012:instance/i-12345678|InstanceType

Would retrieve the data for this particular EC2 instance and would then filter the returned data through the (very) simple jmespath query to which retrieves the value of the attribute InstanceType within the data. The filtered data is available as the filtered_data attribute of the Resource object. The full, unfiltered data is still available as the data attribute.

Multithreaded Usage

Skew is single-threaded by default, like most Python libraries. In order to speed up the enumeration of matching resources, you can use multiple threads:

import skew

class Worker(Thread):
   def __init__(self, arn):
       Thread.__init__(self)
       self.arn = arn
       self.name = arn

   def run(self):
       for i in skew.scan(self.arn):
           # now do something with the stuff

arn = skew.ARN()

for service in arn.service.choices():
    uri = 'arn:aws:' + service + ':*:*:*/*'
    worker = Worker(uri);
    worker.start()

(thanks to @alFReD-NSH for the snippet)

More Examples

Find Unattached Volumes

Audit Security Groups

Find Untagged Instances