ssm-cache 2.2

AWS System Manager Parameter Store caching client for Python

This module wraps the AWS Parameter Store and adds a caching and grouping layer with max-age invalidation.

You can use this module with AWS Lambda to read and refresh sensitive parameters. Your IAM role will require ssm:GetParameters permissions (optionally, also kms:Decrypt if you use SecureString params).

How to install

Install the module with pip:

pip install ssm-cache

How to use it

Simplest use case

A single parameter, configured by name.

from ssm_cache import SSMParameter
param = SSMParameter('my_param_name')
value = param.value

With cache invalidation

You can configure the max_age in seconds, after which the values will be automatically refreshed.

from ssm_cache import SSMParameter
param_1 = SSMParameter('param_1', max_age=300)  # 5 min
value_1 = param.value

param_2 = SSMParameter('param_2', max_age=3600)  # 1 hour
value_2 = param_2.value

With multiple parameters

You can configure more than one parameter to be fetched/cached (and decrypted or not) together.

from ssm_cache import SSMParameterGroup
group = SSMParameterGroup(max_age=300)
param_1 = group.parameter('param_1')
param_2 = group.parameter('param_2')

value_1 = param_1.value
value_2 = param_2.value

Explicit refresh

You can manually force a refresh on a parameter or parameter group. Note that if a parameter is part of a group, refreshing it will refresh the entire group.

from ssm_cache import SSMParameter
param = SSMParameter('my_param_name')
value = param.value
param.refresh()
new_value = param.value

from ssm_cache import SSMParameterGroup
group = SSMParameterGroup()
param_1 = group.parameter('param_1')
param_2 = group.parameter('param_2')

value_1 = param_1.value
value_2 = param_2.value

group.refresh()
new_value_1 = param_1.value
new_value_2 = param_2.value

param_1.refresh()
new_new_value_1 = param_1.value
new_new_value_2 = param_2.value # one parameter refreshes the whole group

Without decryption

Decryption is enabled by default, but you can explicitly disable it on either an SSMParameter or SSMGroup.

from ssm_cache import SSMParameter
param = SSMParameter('my_param_name', with_decryption=False)
value = param.value

Usage with AWS Lambda

Your AWS Lambda code will look similar to the following snippet.

from ssm_cache import SSMParameter
param = SSMParameter('my_param_name')

def lambda_handler(event, context):
    secret_value = param.value
    return 'Hello from Lambda with secret %s' % secret_value

Complex invalidation based on “signals”

You may want to explicitly refresh the parameter cache when you believe the cached value expired.

In the example below, we refresh the parameter value when an InvalidCredentials exception is detected (see the decorator utility for a simpler version!).

from ssm_cache import SSMParameter
from my_db_lib import Client, InvalidCredentials  # pseudo-code

param = SSMParameter('my_db_password')
my_db_client = Client(password=param.value)

def read_record(is_retry=False):
    try:
        return my_db_client.read_record()
    except InvalidCredentials:
        if not is_retry:  # avoid infinite recursion
            param.refresh()  # force parameter refresh
            my_db_client = Client(password=param.value)  # re-configure db client
            return read_record(is_retry=True)  # let's try again :)

def lambda_handler(event, context):
    return {
        'record': read_record(),
    }

Decorator utility

The retry logic shown above can be simplified with the decorator method provided by each SSMParameter and SSMParameterGroup object.

The @refresh_on_error decorator will intercept errors (or a specific error_class, if given), refresh the parameters values, and attempt to re-call the decorated function. Optionally, you can provide a callback argument to implement your own logic (in the example below, to create a new db client with the new password).

from ssm_cache import SSMParameter
from my_db_lib import Client, InvalidCredentials  # pseudo-code

param = SSMParameter('my_db_password')
my_db_client = Client(password=param.value)

def on_error_callback():
    my_db_client = Client(password=param.value)

@param.refresh_on_error(InvalidCredentials, on_error_callback)
def read_record(is_retry=False):
    return my_db_client.read_record()

def lambda_handler(event, context):
    return {
        'record': read_record(),
    }

Optionally, you can also customize the is_retry argument name. refresh_on_error supports the following arguments:

• error_class (default: Exception)

• error_callback (default: None)

• retry_argument (default: "is_retry")

Replacing the SSM client

If you want to replace the default boto3 SSM client, SSMParameter and SSMParameterGroup both support calling set_ssm_client with an object that implements the SSM get_parameters method.

For example, here’s how you could inject a Placebo client for local tests:

import placebo, boto3
from ssm_cache import SSMParameter

# create regular boto3 session
session = boto3.Session()
# attach placebo to the session
pill = placebo.attach(session, data_path=PLACEBO_PATH)
pill.playback()
# create special boto3 client
client = session.client('ssm')
# inject special client into SSMParameter or SSMParameterGroup
SSMParameter.set_ssm_client(client)

How to contribute

Clone this repository, create a virtualenv and install all the dev dependencies:

git clone https://github.com/alexcasalboni/ssm-cache-python.git
cd ssm-cache-python
virtualenv env
source env/bin/activate
pip install -r requirements-dev.txt

You can run tests with nose:

nosetests

Generate a coverage report:

nosetests --with-coverage --cover-erase --cover-html --cover-package=ssm_cache
open cover/index.html

Run pylint:

pylint ssm_cache

Note: when you open a new PR, GitHub will run tests on multiple Python environments and verify the new coverage for you, but we highly recommend you run these tasks locally as well before submitting new code.

What’s new?

• version 2.2: client replacement and boto3/botocore minimum requirements

• version 2.1: group refresh bugfix

• version 2.0: new interface, SSMParameterGroup support

• version 1.3: Python3 support

• version 1.0: initial release

References and articles

• You should use SSM Parameter Store over Lambda env variables by Yan Cui (similar Node.js implementation)

• AWS System Manager Parameter Store doc