Last updated May 30, 2023

This article describes how to use AWS PrivateLink to create a secure connection between an AWS VPC and a Heroku Postgres database running in a Private Space or a Shield Private Space. This process involves three high-level steps:

1. Creating an Endpoint Service on your Heroku Postgres database
2. Creating an Endpoint Network Interface in your AWS VPC
3. Establishing the secure connection between the two endpoints

As part of setting up the connection, you can specify a list of approved accounts to limit access to your Heroku Postgres database from the VPC.

Heroku Prerequisites

The following Heroku resources are required to set up a PrivateLink endpoint:

• A Private Space. This article describes how to create a Private Space using either the Heroku Dashboard or the Heroku CLI.

• A Shield Private Space. This article describes how to create a Shield Private Space using either the Heroku Dashboard or the Heroku CLI.

• A Heroku app running in the Private Space or Shield Private Space with an attached Heroku Postgres database. All Heroku Postgres instances running in a Private Space or a Shield Private Space use one of the private or shield plan types, respectively.

Provisioning the Heroku endpoint

Step 1: Install the Heroku Data via PrivateLink CLI Plugin

$ heroku plugins:install @heroku-cli/plugin-data-privatelink

Step 2: Obtain Your AWS Account ID

You can obtain your AWS account ID with the AWS CLI like so:

$ aws sts get-caller-identity --output text --query 'Account'

123456789101

The example command returns an account ID of 123456789101.

You can also obtain your account ID from the My Account page of your AWS account. The Account ID is shown in the Account Settings section:

Step 3: Create a PrivateLink endpoint

Create a PrivateLink endpoint using the following Heroku CLI command (note the values to substitute):

$ heroku data:privatelink:create POSTGRESQL_ADDON_NAME --aws-account-id ACCOUNT_ID --app APP_NAME

• Replace POSTGRESQL_ADDON_NAME with the name of your Heroku Postgres database (for example, postgresql-sushi-12345).
• Replace APP_NAME with your app’s name.
• Replace ACCOUNT_ID with the AWS account that receives access to your Heroku Postgres database. This ID can match any of the following patterns:
  • account-id
  • account-id:user/username
  • account-id:role/rolename

Here’s an example command with accompanying output:

$ heroku data:privatelink:create postgresql-sushi-12345 --aws-account-id 123456789101:user/abc.xyz --app privatelink-vpc-endpoint-demo
Creating privatelink... done

Service Name: Provisioning
Status:       Provisioning

The privatelink is now being provisioned for postgresql-sushi-12345.
Run heroku data:privatelink:wait postgresql-sushi-12345 -a APP to check the creation process.

Step 4: Obtain Your Endpoint’s Service Name

When the PrivateLink endpoint finishes provisioning, use the following command to view its details:

$ heroku data:privatelink POSTGRESQL_ADDON_NAME --app APP_NAME

Replace POSTGRESQL_ADDON_NAME with the name of your Heroku Postgres database, and replace APP_NAME with your app’s name.

Here’s an example command with accompanying output:

$ heroku data:privatelink postgresql-sushi-12345 --app privatelink-vpc-endpoint-demo
=== privatelinks for postgresql-sushi-12345
Service Name:         com.amazonaws.vpce.us-east-1.vpce-svc-0410a2e25933fe8ec
Status:               Operational

=== Allowed Accounts
ARN                                    Status
arn:aws:iam::123456789101:user/abc.xyz Active

Your privatelink is now operational.
You must now copy the Service Name and follow the rest of the steps listed in https://devcenter.heroku.com/articles/heroku-postgres-via-privatelink.

Copy the value of the Service Name field from the command’s output (in the example above, the value is com.amazonaws.vpce.us-east-1.vpce-svc-0410a2e25933fe8ec). You need this value to provision the Amazon VPC endpoint.

Provisioning the Amazon VPC Endpoint

You perform the steps in this section from your Amazon VPC dashboard.

Step 1: Create and Configure a Security Group

Your endpoint requires a security group with appropriate ingress security rules. Click Create security group in the Security Groups tab of your VPC dashboard:

Specify an appropriate security group name and description and select your desired VPC before clicking Create:

Select your newly created security group from the list and click Actions > Edit inbound rules:

Enable TCP access to ports 5432–5433 from any valid IP address and click Save rules:

Step 2: Create the endpoint

Navigate to the Endpoints tab of your VPC dashboard and click Create Endpoint:

In the Create Endpoint form that appears, select the Find service by name option and paste the Service Name value you obtained earlier.

Then click Verify to display the list of available subnets:

Attach the security group you created earlier to the VPC Endpoint and click Create endpoint:

The endpoint is created with an initial status of pending acceptance, which transitions to available after 5-10 minutes:

Connecting the Heroku and Amazon VPC Endpoints

After the Amazon VPC endpoint becomes available, you can obtain the URLs from the config vars that Heroku creates to allow your VPC to communicate with your Heroku Postgres database (DATABASE_ENDPOINT_ID_URL) and PgBouncer (DATABASE_ENDPOINT_ID_PGBOUNCER_URL).

First, obtain your PrivateLink endpoint’s Endpoint ID and extract the 17-character string that appears at the end of it. Convert that string to upper case and use it in the following command.

For example, if the Endpoint ID is vpce-01c87ae3c05563935, the Endpoint ID is 01C87AE3C05563935.

Run the following command, substituting the obtained string where indicated:

$ heroku config --app your_app_name | grep ENDPOINT_ID_HERE

This command displays the AWS VPC Endpoint connection URL and the corresponding connection string for your Heroku Postgres database. The connection string has the following format:

postgres://user:password@vpc-endpoint-dns-name:5432/database

You can now use this connection string to connect the applications in your AWS VPC to your private Heroku Postgres database and PgBouncer instance. Here’s an example command with accompanying output:

$ heroku config --app privatelink-vpc-endpoint-demo | grep 01C87AE3C05563935

DATABASE_ENDPOINT_01C87AE3C05563935_URL:           postgres://abcdefghijklmn:abcdefghijklmnopqrstuvwxyz123456789101112131415161718192021222324@ec2-3-83-63-168.compute-1.amazonaws.com:5432/dd0k757ojc5qt
DATABASE_ENDPOINT_01C87AE3C05563935_PGBOUNCER_URL: postgres://abcdefghijklmn:abcdefghijklmnopqrstuvwxyz123456789101112131415161718192021222324@ec2-3-83-63-168.compute-1.amazonaws.com:5433/dd0k757ojc5qt

Connecting to Heroku Postgres from EC2 via VPC Endpoints

After you configure your VPC endpoints, you can create an EC2 instance in your AWS VPC to connect to Heroku Postgres.

Click Launch Instance in your EC2 dashboard and select your AMI and Instance Type. In this example, an Ubuntu t2.micro instance is created:

When configuring the instance’s details, select the VPC network with the security group you created earlier and pick an appropriate subnet. Click Review and Launch and launch the EC2 instance.

After the instance’s status transitions to running and all status checks have passed, connect to the instance using the SSH key pair you specified during instance creation.

After it’s connected, install Postgres on your EC2 instance with the following commands:

$ sudo apt-get -qq update && sudo apt-get install -y curl ca-certificates
$ curl -s https://www.postgresql.org/media/keys/ACCC4CF8.asc | sudo apt-key add -
$ sudo sh -c 'echo "deb http://apt.postgresql.org/pub/repos/apt/ $(lsb_release -cs)-pgdg main" > /etc/apt/sources.list.d/pgdg.list'
$ sudo apt-get -qq update
$ sudo apt-get install -y postgresql-client-10

You can then connect to your Heroku Postgres database from the EC2 instance using the following command:

$ psql postgres://user:password@vpc-endpoint-dns-name:5432/database

The following screenshot shows the connection string being used to connect to a Heroku Postgres database from within a sample EC2 instance:

Limitations

• The Amazon VPC Endpoint you create must be provisioned in a subnet that is in the same region as your Heroku Postgres database.
• It is your responsibility to verify the security of your VPC to ensure fully secure access to your Heroku Postgres database.

Comparison to Private Space Trusted IP Ranges

Private Spaces support trusted IP ranges for data services as a beta feature. This feature is an option if you must connect to a Heroku Postgres database from outside the Private Space boundary. However, you must contact Heroku to enable this feature, and granular access control isn’t available for it.

By connecting via AWS PrivateLink, your database is treated as part of your own VPC, and you can restrict access to a set of users and roles. Consequently, this method is recommended whenever it’s available for your use case.