AWS Compute Blog

Using the AWS Parameter and Secrets Lambda extension to cache parameters and secrets

This post is written by Pal Patel, Solutions Architect, and Saud ul Khalid, Sr. Cloud Support Engineer.

Serverless applications often rely on AWS Systems Manager Parameter Store or AWS Secrets Manager to store configuration data, encrypted passwords, or connection details for a database or API service.

Previously, you had to make runtime API calls to AWS Parameter Store or AWS Secrets Manager every time you wanted to retrieve a parameter or a secret inside the execution environment of an AWS Lambda function. This involved configuring and initializing the AWS SDK client and managing when to store values in memory to optimize the function duration, and avoid unnecessary latency and cost.

The new AWS Parameters and Secrets Lambda extension provides a managed parameters and secrets cache for Lambda functions. The extension is distributed as a Lambda layer that provides an in-memory cache for parameters and secrets. It allows functions to persist values through the Lambda execution lifecycle, and provides a configurable time-to-live (TTL) setting.

When you request a parameter or secret in your Lambda function code, the extension retrieves the data from the local in-memory cache, if it is available. If the data is not in the cache or it is stale, the extension fetches the requested parameter or secret from the respective service. This helps to reduce external API calls, which can improve application performance and reduce cost. This blog post shows how to use the extension.


The following diagram provides a high-level view of the components involved.

High-level architecture showing how parameters or secrets are retrieved when using the Lambda extension

The extension can be added to new or existing Lambda. It works by exposing a local HTTP endpoint to the Lambda environment, which provides the in-memory cache for parameters and secrets. When retrieving a parameter or secret, the extension first queries the cache for a relevant entry. If an entry exists, the query checks how much time has elapsed since the entry was first put into the cache, and returns the entry if the elapsed time is less than the configured cache TTL. If the entry is stale, it is invalidated, and fresh data is retrieved from either Parameter Store or Secrets Manager.

The extension uses the same Lambda IAM execution role permissions to access Parameter Store and Secrets Manager, so you must ensure that the IAM policy is configured with the appropriate access. Permissions may also be required for AWS Key Management Service (AWS KMS) if you are using this service. You can find an example policy in the example’s AWS SAM template.

Example walkthrough

Consider a basic serverless application with a Lambda function connecting to an Amazon Relational Database Service (Amazon RDS) database. The application loads a configuration stored in Parameter Store and connects to the database. The database connection string (including user name and password) is stored in Secrets Manager.

This example walkthrough is composed of:

  • A Lambda function.
  • An Amazon Virtual Private Cloud (VPC).
  • Multi-AZ Amazon RDS Instance running MySQL.
  • AWS Secrets Manager database secret that holds database connection.
  • AWS Systems Manager Parameter Store parameter that holds the application configuration.
  • An AWS Identity and Access Management (IAM) role that the Lambda function uses.

Lambda function

This Python code shows how to retrieve the secrets and parameters using the extension

import pymysql
import urllib3
import os
import json

### Load in Lambda environment variables
aws_session_token = os.environ['AWS_SESSION_TOKEN']
env = os.environ['ENV']
app_config_path = os.environ['APP_CONFIG_PATH']
creds_path = os.environ['CREDS_PATH']
full_config_path = '/' + env + '/' + app_config_path
http = urllib3.PoolManager()
### Define function to retrieve values from extension local HTTP server cachce
def retrieve_extension_value(url): 
    url = ('http://localhost:' + port + url)
    headers = { "X-Aws-Parameters-Secrets-Token": os.environ.get('AWS_SESSION_TOKEN') }
    response = http.request("GET", url, headers=headers)
    response = json.loads(   
    return response  

def lambda_handler(event, context):
    ### Load Parameter Store values from extension
    print("Loading AWS Systems Manager Parameter Store values from " + full_config_path)
    parameter_url = ('/systemsmanager/parameters/get/?name=' + full_config_path)
    config_values = retrieve_extension_value(parameter_url)['Parameter']['Value']
    print("Found config values: " + json.dumps(config_values))

    ### Load Secrets Manager values from extension
    print("Loading AWS Secrets Manager values from " + creds_path)
    secrets_url = ('/secretsmanager/get?secretId=' + creds_path)
    secret_string = json.loads(retrieve_extension_value(secrets_url)['SecretString'])
    #print("Found secret values: " + json.dumps(secret_string))

    rds_host =  secret_string['host']
    rds_db_name = secret_string['dbname']
    rds_username = secret_string['username']
    rds_password = secret_string['password']
    ### Connect to RDS MySQL database
        conn = pymysql.connect(host=rds_host, user=rds_username, passwd=rds_password, db=rds_db_name, connect_timeout=5)
        raise Exception("An error occurred when connecting to the database!")

    return "DemoApp sucessfully loaded config " + config_values + " and connected to RDS database " + rds_db_name + "!"

In the global scope the environment variable PARAMETERS_SECRETS_EXTENSION_HTTP_PORT is retrieved, which defines the port the extension HTTP server is running on. This defaults to 2773.

The retrieve_extension_value function calls the extension’s local HTTP server, passing in the X-Aws-Parameters-Secrets-Token as a header. This is a required header that uses the AWS_SESSION_TOKEN value, which is present in the Lambda execution environment by default.

The Lambda handler code uses the extension cache on every Lambda invoke to obtain configuration data from Parameter Store and secret data from Secrets Manager. This data is used to make a connection to the RDS MySQL database.


  1. Git installed
  2. AWS SAM CLI version 1.58.0 or greater.

Deploying the resources

  1. Clone the repository and navigate to the solution directory:
    git clone
  2. Build and deploy the application using following command:
    sam build
    sam deploy --guided

This template takes the following parameters:

  • pVpcCIDR — IP range (CIDR notation) for the VPC. The default is
  • pPublicSubnetCIDR — IP range (CIDR notation) for the public subnet. The default is
  • pPrivateSubnetACIDR — IP range (CIDR notation) for the private subnet A. The default is
  • pPrivateSubnetBCIDR — IP range (CIDR notation) for the private subnet B, which defaults to
  • pDatabaseName — Database name for DEV environment, defaults to devDB
  • pDatabaseUsername — Database user name for DEV environment, defaults to myadmin
  • pDBEngineVersion — The version number of the SQL database engine to use (the default is 5.7).

Adding the Parameter Store and Secrets Manager Lambda extension

To add the extension:

  1. Navigate to the Lambda console, and open the Lambda function you created.
  2. In the Function Overview pane. select Layers, and then select Add a layer.
  3. In the Choose a layer pane, keep the default selection of AWS layers and in the dropdown choose AWS Parameters and Secrets Lambda Extension
  4. Select the latest version available and choose Add.

The extension supports several configurable options that can be set up as Lambda environment variables.

This example explicitly sets an extension port and TTL value:

Lambda environment variables from the Lambda console

Testing the example application

To test:

  1. Navigate to the function created in the Lambda console and select the Test tab.
  2. Give the test event a name, keep the default values and then choose Create.
  3. Choose Test. The function runs successfully:

Lambda execution results visible from Lambda console after successful invocation.

To evaluate the performance benefits of the Lambda extension cache, three tests were run using the open source tool Artillery to load test the Lambda function. This can use the Lambda URL to invoke the function. The Artillery configuration snippet shows the duration and requests per second for the test:

  target: ""
      duration: 60
      arrivalRate: 10
      rampTo: 40

          url: ""
  • Test 1: The extension cache is disabled by setting the TTL environment variable to 0. This results in 1650 GetParameter API calls to Parameter Store over 60 seconds.
  • Test 2: The extension cache is enabled with a TTL of 1 second. This results in 106 GetParameter API calls over 60 seconds.
  • Test 3: The extension is enabled with a TTL value of 300 seconds. This results in only 18 GetParameter API calls over 60 seconds.

In test 3, the TTL value is longer than the test duration. The 18 GetParameter calls correspond to the number of Lambda execution environments created by Lambda to run requests in parallel. Each execution environment has its own in-memory cache and so each one needs to make the GetParameter API call.

In this test, using the extension has reduced API calls by ~98%. Reduced API calls results in reduced function execution time, and therefore reduced cost.


After you test this example, delete the resources created by the template, using following commands from the same project directory to avoid continuing charges to your account.

sam delete


Caching data retrieved from external services is an effective way to improve the performance of your Lambda function and reduce costs. Implementing a caching layer has been made simpler with this AWS-managed Lambda extension.

For more information on the Parameter Store, Secrets Manager, and Lambda extensions, refer to:

For more serverless learning resources, visit Serverless Land.