Retrieve secrets for AWS applications with Vault Agent

15min
|
Terraform
Vault
Video

You must pass an authentication token with any request from an authenticated endpoint in Vault. This includes all API requests, as well as via the Vault CLI and other libraries.

Vault supports several different authentication methods to enable delivery of the initial token. If you can securely get the first secret from an originator to a consumer, then all secrets later exchanged between them can be authenticated. Getting that first secret to the consumer is the secure introduction challenge.

To that end, Vault provides integration with native authentication capabilities in various environments. For example: IAM in AWS and Google Cloud, Managed Service Identities in Azure, and Service Accounts in Kubernetes are all supported.

Challenge

Although Vault provides the auth method, the client is still responsible for managing the lifecycle of tokens created from the auth method. The challenge then becomes how to enable authentication to Vault, and manage token lifecycle in a standard way without writing custom logic.

Solution

Vault Agent provides a number of different helper features, specifically addressing the following challenges:

Automatic authentication
Secure delivery/storage of tokens
Lifecycle management of these tokens (renewal & re-authentication)

Note

The Vault Agent Auto-Auth functionality addresses the challenges related to obtaining and managing authentication tokens (secret zero).

Note

The Secure Introduction of Vault Clients introduced three basic approaches: Platform Integration, Trusted Orchestrator, and Vault Agent. This tutorial demonstrates how Vault Agent works.

Prerequisites

To perform the tasks described in this tutorial, you need to have:

Terraform v1.0.0 or later installed.
AWS account and associated credentials that allow for the creation of resources.
tree utility for visualizing directory contents (you can use ls -1R instead).

Provision the cloud resources

Clone the demo assets from the HashiCorp Education repository GitHub repository to perform the steps described in this tutorial.
```
$ git clone https://github.com/hashicorp-education/learn-vault-agent-demo
```
This repository has supporting content for all the Vault learn tutorials. You can find the content for this tutorial within a sub-directory.

Be sure to set your working directory to location of the learn-vault-agent-demo/terraform-aws folder.

$ cd learn-vault-agent-demo/terraform-aws

The working directory should contain the provided Terraform files:

$ tree
.
├── aws.tf
├── iam.tf
├── kms.tf
├── network.tf
├── outputs.tf
├── security-groups.tf
├── templates
│   ├── userdata-vault-client.tftpl
│   └── userdata-vault-server.tftpl
├── terraform.tfvars.example
├── variables.tf
├── vault-client.tf
├── vault-server.tf
└── versions.tf

2 directories, 13 files

Note

The example Terraform configuration in this repository is for demonstration purposes, and not suitable for production use. For production deployment, refer the example Terraform in the Vault Guides repository /operations/provision-vault.

Set an AWS_ACCESS_KEY_ID environment variable to hold your AWS access key ID.
```
$ export AWS_ACCESS_KEY_ID = "<YOUR_AWS_ACCESS_KEY_ID>"
```
Set an AWS_SECRET_ACCESS_KEY environment variable to hold your AWS secret access key.
```
$ export AWS_SECRET_ACCESS_KEY = "<YOUR_AWS_SECRET_ACCESS_KEY>"
```
Tip
The above example uses IAM user authentication. You can use any authentication method described in the AWS provider documentation.
Copy the terraform.tfvars.example file and rename it as terraform.tfvars. Within this file, edit the key_name parameter to be the name of your EC2 key pair.
```
$ cp terraform.tfvars.example terraform.tfvars
```
Example:
terraform.tfvars
```
# SSH key name to access EC2 instances (should already exist)
key_name = "vault-test"
```
You should also change the values of the AWS region and availability zone or instance type as necessary.
Tip
If you don't have an EC2 key pair, follow the AWS documentation to create one.

Initialize the Terraform workspace, and download the necessary provider resources.

$ terraform init

Initializing provider plugins...
...snip...
Terraform has been successfully initialized!

Run terraform apply and review the planned actions. Your terminal output should show the plan, resources Terraform will provision.

$ terraform apply

...snip...
Plan: 20 to add, 0 to change, 0 to destroy.

Changes to Outputs:
  + endpoints = (known after apply)

Do you want to perform these actions?
  Terraform will perform the actions described above.
  Only 'yes' will be accepted to approve.

  Enter a value: yes

Enter yes to confirm and resume.

When the apply command completes, the Terraform output will display the public IP address to SSH into your Vault server and client instances.

Example output:

Apply complete! Resources: 20 added, 0 changed, 0 destroyed.

Outputs:

endpoints =
Vault Server IP (public):  54.219.129.15
Vault Server IP (private): 10.0.101.50

For example:
  ssh -i vault-test.pem ubuntu@54.219.129.15

Vault Client IP (public):  54.183.212.51
Vault Client IP (private): 10.0.101.209

For example:
  ssh -i vault-test.pem ubuntu@54.183.212.51

Configure AWS IAM auth method

Note

Perform this step on the Vault server instance.

In this step, you will configure Vault to allow AWS IAM authentication from specific IAM roles.

SSH into the Vault server instance.
At the prompt, enter "yes" to continue.
```
$ ssh -i <path_to_key> ubuntu@<public_ip_of_server>
...
Are you sure you want to continue connecting (yes/no)? yes
```
Tip
If you received Permissions 0664 for '<key_name>.pem' are too open error, be sure to set the file permission appropriately.
```
$ chmod 600 <key_name>.pem
```

To verify your Vault installation, run vault status command and notice that Initialized is false.

$ vault status
Key                      Value
---                      -----
Recovery Seal Type       awskms
Initialized              false
Sealed                   true
Total Recovery Shares    0
Threshold                0
Unseal Progress          0/0
Unseal Nonce             n/a
Version                  1.12.2
Build Date               2022-11-23T12:53:46Z
Storage Type             raft
HA Enabled               true

Run the vault operator init command to initialize the Vault server:
```
$ vault operator init

...snip...

Initial Root Token: s.20JnHBY66EKTj9zyR6SjTMNq

Success! Vault is initialized
...snip...
```
Tip
The Vault server uses auto-unseal with AWS Key Management Service (KMS). When you initialize the Vault, the server is automatically unsealed. To learn how to auto-unseal Vault using AWS KMS, refer to the Auto-unseal using AWS KMS tutorial.
Copy the Initial Root Token value.
Log into Vault. Enter the generated initial root token when prompted.
```
$ vault login -no-print
Token (will be hidden):
```
Examine and then execute the /home/ubuntu/aws_auth.sh script.
```
$ cat aws_auth.sh
```
The script enables key/value v1 secrets engine at secret and writes some test data at secret/myapp/config (at line 1 and 2).
At line 4 through 6, it creates a myapp policy, and line 8 enables the aws auth method.
It creates a role named dev-role-iam which is valid for 24 hours with capabilities set by the myapp policy. (Note: The ${account_id} should be your AWS account ID.) The bound_iam_principal_arn is the Amazon Resource Name (ARN) of the IAM role created by the iam.tf file (at line 11).
aws_auth.sh
```
vault secrets enable -path="secret" kv
vault kv put secret/myapp/config ttl='30s' username='appuser' password='suP3rsec(et!'

echo "path \"secret/myapp/*\" {
    capabilities = [\"read\", \"list\"]
}" | vault policy write myapp -

vault auth enable aws
vault write -force auth/aws/config/client

vault write auth/aws/role/dev-role-iam auth_type=iam bound_iam_principal_arn="arn:aws:iam::${account_id}:role/vault-agent-demo-vault-client-role" policies=myapp ttl=24h 
```

Execute the script.

$ ./aws_auth.sh

Success! Enabled the kv secrets engine at: secret/
Success! Data written to: secret/myapp/config
Success! Uploaded policy: myapp
Success! Enabled aws auth method at: aws/
Success! Data written to: auth/aws/config/client
Success! Data written to: auth/aws/role/dev-role-iam

Check the myapp policy.

$ vault policy read myapp

path "secret/myapp/*" {
    capabilities = ["read", "list"]
}

Check the secrets written in secret/myapp/config.

$ vault kv get secret/myapp/config

====== Data ======
Key         Value
---         -----
password    suP3rsec(et!
ttl         30s
username    appuser

Run Vault Agent with auto-auth

Note

Perform this step on the Vault client instance.

Now, you are going to see how Vault Agent Auto-Auth method works, and write out a token to an arbitrary location on disk. Vault Agent is a client daemon and its Auto-Auth feature allows for easy authentication to Vault.

Open a new terminal and SSH into the Vault Client instance. Output

At the prompt, enter "yes" to continue.

$ ssh -i <path_to_key> ubuntu@<public_ip_of_client>
...
Are you sure you want to continue connecting (yes/no)? yes

In the client instance, explore the Vault Agent configuration file (/home/ubuntu/vault-agent.hcl).
```
$ cat vault-agent.hcl
```
Starting at line 4, the auto_auth block has two configuration entries: method and sink. In this example, Auto-Auth uses the aws auth method enabled at the auth/aws path on the Vault server. The Vault Agent will use the dev-role-iam role to authenticate.
The sink block specifies the location on disk where to write tokens. You can define the Vault Agent Auto-Auth sink more than once if you want Vault Agent to place the token into additional locations. In this example, the sink path is /home/ubuntu/vault-token-via-agent.
The exit_after_auth parameter is true. This means that the agent will exit with code 0 after a single successful authentication. The default is false and the agent continues to run and automatically renew the client token for you.
vault-agent.hcl
```
exit_after_auth = true
pid_file = "./pidfile"

auto_auth {
  method "aws" {
      mount_path = "auth/aws"
      config = {
          type = "iam"
          role = "dev-role-iam"
      }
  }

  sink "file" {
      config = {
          path = "/home/ubuntu/vault-token-via-agent"
      }
  }
}

vault {
  address = "http://10.0.101.53:8200"
}
```
The vault block points to the Vault server address. This should match to the private IP address of your Vault server host.
Tip
For the full details of Vault Agent configuration parameters, refer to the Vault Agent documentation.
Execute the following command to get help:
```
$ vault agent -h
```

Now, you are ready to run the Vault Agent. Execute the following command:

$ vault agent -config=/home/ubuntu/vault-agent.hcl -log-level=debug
==> Vault agent started! Log data will stream in below:

==> Vault agent configuration:

                     Cgo: disabled
               Log Level: debug
               Version: Vault v1.12.2, built 2022-11-23T12:53:46Z
            Version Sha: 415e1fe3118eebd5df6cb60d13defdc01aa17b03

2023-01-25T16:55:56.473Z [INFO]  sink.file: creating file sink
2023-01-25T16:55:56.473Z [INFO]  sink.file: file sink configured: path=/home/ubuntu/vault-token-via-agent mode=-rw-r-----
2023-01-25T16:55:56.483Z [DEBUG] would have sent systemd notification (systemd not present): notification=READY=1
2023-01-25T16:55:56.483Z [INFO]  template.server: starting template server
2023-01-25T16:55:56.483Z [INFO]  template.server: no templates found
2023-01-25T16:55:56.484Z [INFO]  auth.handler: starting auth handler
2023-01-25T16:55:56.484Z [INFO]  auth.handler: authenticating
2023-01-25T16:55:56.484Z [INFO]  sink.server: starting sink server
2023-01-25T16:55:57.562Z [INFO]  auth.handler: authentication successful, sending token to sinks
2023-01-25T16:55:57.562Z [INFO]  auth.handler: starting renewal process
2023-01-25T16:55:57.562Z [INFO]  sink.file: token written: path=/home/ubuntu/vault-token-via-agent
2023-01-25T16:55:57.562Z [INFO]  sink.server: sink server stopped
2023-01-25T16:55:57.563Z [INFO]  sinks finished, exiting
2023-01-25T16:55:57.563Z [INFO]  template.server: template server stopped
2023-01-25T16:55:57.563Z [INFO]  auth.handler: shutdown triggered, stopping lifetime watcher
2023-01-25T16:55:57.563Z [INFO]  auth.handler: auth handler stopped
2023-01-25T16:55:57.563Z [DEBUG] would have sent systemd notification (systemd not present): notification=STOPPING=1

Because the vault-agent.hcl configuration file contained the line exit_after_auth = true, Vault Agent authenticated and retrieved a token once, wrote it to the defined sink, and exited. Vault Agent can also run in daemon mode where it will continuously renew the retrieved token, and try to re-authenticate if that token becomes invalid. Vault Agent Caching tutorial will show running Vault Agent as a daemon.

Vault Agent writes the token to /home/ubuntu/vault-token-via-agent.

$ more vault-token-via-agent
hvs.CAESIH5hFzxXxBABzyeq78mHlbQZsN8ETqaEQX24XbEZHpdyGh4KHGh2cy5oNjZrMnpyWG02T01xZURkQTVoTDRwbkk

Export a VAULT_ADDR environment variable to address the Vault server directly.

$ export VAULT_ADDR=$(grep address vault-agent.hcl | awk '{print $3}' | tr -d '"')

Try an API call using the token that Vault Agent retrieved to test:

$ curl \
    --silent \
    --header "X-Vault-Token: $(cat /home/ubuntu/vault-token-via-agent)" \
    $VAULT_ADDR/v1/secret/myapp/config | jq -r ".data"

Output:

{
  "password": "suP3rsec(et!",
  "ttl": "30s",
  "username": "appuser"
}

Response wrap the token

Note

Perform this step on the Vault client instance.

It may not be ideal to write the token as a plaintext depending on the firewall around the client instance. Vault Agent supports response-wrapping of the token to offer an additional layer of protection for the token. You can wrap tokens by either the auth method or by the sink configuration, with each approach solving for different challenges as described in Response-Wrapping Tokens.

In the client instance, explore the /home/ubuntu/vault-agent-wrapped.hcl file which supports response-wrapping of the token.

$ cat /home/ubuntu/vault-agent-wrapped.hcl

exit_after_auth = true
pid_file = "./pidfile"

auto_auth {
    method "aws" {
        mount_path = "auth/aws"
        config = {
            type = "iam"
            role = "dev-role-iam"
        }
    }

    sink "file" {
        wrap_ttl = "5m"
        config = {
            path = "/home/ubuntu/vault-token-via-agent"
        }
    }
}

vault {
  address = "http://10.0.101.53:8200"
}

Notice the wrap_ttl parameter in the sink block. This configuration uses the sink method to response-wrap the retrieved tokens.

From the client, run the Vault Agent again and inspect the output:

$ vault agent -config=/home/ubuntu/vault-agent-wrapped.hcl -log-level=debug

==> Vault agent started! Log data will stream in below:

==> Vault agent configuration:

                     Cgo: disabled
               Log Level: debug
               Version: Vault v1.12.2, built 2022-11-23T12:53:46Z
            Version Sha: 415e1fe3118eebd5df6cb60d13defdc01aa17b03

2023-01-25T17:03:13.056Z [INFO]  sink.file: creating file sink
2023-01-25T17:03:13.056Z [INFO]  sink.file: file sink configured: path=/home/ubuntu/vault-token-via-agent mode=-rw-r-----
2023-01-25T17:03:13.065Z [DEBUG] would have sent systemd notification (systemd not present): notification=READY=1
2023-01-25T17:03:13.065Z [INFO]  template.server: starting template server
2023-01-25T17:03:13.065Z [INFO]  template.server: no templates found
2023-01-25T17:03:13.065Z [INFO]  auth.handler: starting auth handler
2023-01-25T17:03:13.065Z [INFO]  auth.handler: authenticating
2023-01-25T17:03:13.066Z [INFO]  sink.server: starting sink server
2023-01-25T17:03:14.121Z [INFO]  auth.handler: authentication successful, sending token to sinks
2023-01-25T17:03:14.121Z [INFO]  auth.handler: starting renewal process
2023-01-25T17:03:14.128Z [INFO]  auth.handler: renewed auth token
2023-01-25T17:03:14.144Z [INFO]  sink.file: token written: path=/home/ubuntu/vault-token-via-agent
2023-01-25T17:03:14.144Z [INFO]  sink.server: sink server stopped
2023-01-25T17:03:14.144Z [INFO]  sinks finished, exiting
2023-01-25T17:03:14.144Z [INFO]  template.server: template server stopped
2023-01-25T17:03:14.144Z [INFO]  auth.handler: shutdown triggered, stopping lifetime watcher
2023-01-25T17:03:14.144Z [INFO]  auth.handler: auth handler stopped
2023-01-25T17:03:14.145Z [DEBUG] would have sent systemd notification (systemd not present): notification=STOPPING=1

Instead of a token value, you now have a JSON object containing a wrapping token as well as some additional metadata. To get to the true token, you need to first perform an unwrap operation.

Unwrap the response-wrapped token and save it to a VAULT_TOKEN environment variable that other applications can use:
```
$ export VAULT_TOKEN=$(vault unwrap -field=token $(jq -r '.token' /home/ubuntu/vault-token-via-agent))
```

View the unwrapped token value.

$ echo $VAULT_TOKEN
hvs.CAESIM0bghylBC3fTJ2RWV-9ILquhTEgUaf1pZOdzIGyP1hnGh4KHGh2cy5rME90YzBHUkxHeVNpZXNhdXhEN3BtdUo

Test to make sure that the token has the read capability on secret/myapp/config.
```
$ curl \
    --silent \
    --header "X-Vault-Token: $VAULT_TOKEN" $VAULT_ADDR/v1/secret/myapp/config \
   | jq -r ".data"
```
Notice that the value saved to the VAULT_TOKEN is not the same as the token value in the /home/ubuntu/vault-token-via-agent file. The value in VAULT_TOKEN is the unwrapped token retrieved by Vault Agent.
If you try to unwrap that same value again or wait longer than 5 minutes (wrap_ttl), it will throw an error:
```
$ export VAULT_TOKEN=$(vault unwrap -field=token $(jq -r '.token' /home/ubuntu/vault-token-via-agent))

Error unwrapping: Error making API request.

URL: PUT http://active.vault-va-demo.service.dc1.consul:8200/v1/sys/wrapping/unwrap
Code: 400. Errors:

* wrapping token is not valid or does not exist
```
You can unwrap a response-wrapped token just once. Additional attempts to unwrap an already-unwrapped token will result in triggering an error.
Note
In this tutorial, you learned about the basic mechanics of Vault Agent. Read the Additional discussion section about the next step.

Clean up

Return to the first terminal where you created the cluster and use Terraform to destroy the cluster.

Destroy the AWS resources provisioned by Terraform.

$ terraform destroy -auto-approve

Delete the state file.

$ rm *tfstate*

Additional discussion

In the above examples, you manually ran Vault Agent to show how it works. How you actually integrate Vault Agent into your application deployment workflow will vary based on several factors. Some questions to ask to help decide appropriate usage:

What is the lifecycle of my application? Is it more ephemeral or long-lived?
What are the lifecycles of my authentication tokens? Are they long-lived and requiring repetitive renewal to show liveliness of a service or do you want to enforce periodic re-authentications?
Do I have a group of applications running on my host which each need their own token? Can I use a native authentication capability (for example, AWS IAM, Kubernetes, Azure MSI, Google Cloud IAM, etc.)?

The answers to these questions will help you decide if Vault Agent should run as a daemon or as a prerequisite of a service configuration. Take for example the following Systemd service definition for running Nomad:

[Unit]
Description=Nomad Agent
Requires=consul-online.target
After=consul-online.target

[Service]
KillMode=process
KillSignal=SIGINT
Environment=VAULT_ADDR=http://active.vault.service.consul:8200
Environment=VAULT_SKIP_VERIFY=true
ExecStartPre=/usr/local/bin/vault agent -config /etc/vault-agent.d/vault-agent.hcl
ExecStart=/usr/bin/nomad-vault.sh
ExecReload=/bin/kill -HUP $MAINPID
Restart=on-failure
RestartSec=2
StartLimitBurst=3
StartLimitIntervalSec=10
LimitNOFILE=65536

[Install]
WantedBy=multi-user.target

Notice the ExecStartPre directive that runs Vault Agent before the desired service starts. The service startup script expects a Vault token value set as shown in the /usr/bin/nomad-vault.sh startup script:

#!/usr/bin/env bash

if [ -f /mnt/ramdisk/token ]; then
  exec env VAULT_TOKEN=$(vault unwrap -field=token $(jq -r '.token' /mnt/ramdisk/token)) \
    /usr/local/bin/nomad agent \
      -config=/etc/nomad.d \
      -vault-tls-skip-verify=true
else
  echo "Nomad service failed due to missing Vault token"
  exit 1
fi

Applications which expect Vault tokens typically look for a VAULT_TOKEN environment variable. Here, you're using Vault Agent to get a token and write it out to a RAM disk and as part of the Nomad startup script. You read the response-wrapped token from the RAM disk, and save it to your VAULT_TOKEN environment variable before starting Nomad.

Vault Agent in Vault version 1.3.0 or higher also supports secret templates for use in configuration files, and other similar use cases. You can learn more about this feature in Vault Agent Templates.

Help and reference

Blog post: Why Use the Vault Agent for Secrets Management?
Video: Streamline Secrets Management with Vault Agent and Vault 0.11
Secure Introduction of Vault Clients
Vault Agent Auto-Auth documentation
AWS Auth Method documentation