Auto-unseal Vault using transit secrets engine

13min
|
Vault
Video

When a Vault server is started, it starts in a sealed state and it does not know how to decrypt data. Before any operation can be performed on the Vault, it must be unsealed. Unsealing is the process of constructing the master key necessary to decrypt the data encryption key.

Challenge

Vault unseal operation requires a quorum of existing unseal keys split by Shamir's Secret sharing algorithm. This is done so that the "keys to the kingdom" won't fall into one person's hand.

Unseal with Shamir's Secret Sharing

However, this process is manual and can become painful when you have many Vault clusters as there are now many different key holders with many different keys.

Solution

Vault supports opt-in automatic unsealing via transit secrets engine. This feature enables operators to delegate the unsealing process to a trusted Vault environment to ease operations.

Unseal with Transit

This tutorial demonstrates how to auto-unseal a Vault with Transit secrets engine.

To auto-unseal your Vault with cloud provider's key, refer to the following tutorials:

Prerequisites

To perform the tasks described in this tutorial, you need Vault 1.1 or later.

Scenario introduction

For the purpose of demonstration, you are going to run two instances of Vault as described in the following diagram:

In reality, the Vault 1 and Vault 2 are two separate stand-alone Vault clusters where one protecting another's root key. Nonetheless, the steps described in this tutorial directly applies to your clustered environment. The main difference would be the location (address) of Vault 1 and Vault 2.

In this scenario, Vault 1 ($VAULT_ADDR) is the encryption service provider, and its transit key protects the Vault 2 server's master key.

Scenario Overview

Step 1: Configure auto-unseal key provider (Vault 1)

Bash script to setup Vault 1

If you prefer running a script instead of manually setting up the Vault 1 server, follow the steps below.

Create a setup script namde autounseal-transit-setup.sh.

autounseal-transit-setup.sh

# Start the Vault 1 server in dev mode
# The system output will be stored in the vault-1.log file
vault server -dev -dev-root-token-id root > vault-1.log 2>&1 &

sleep 1

# Set the environment variables: VAULT_ADDR and VAULT_TOKEN
export VAULT_ADDR=http://127.0.0.1:8200
export VAULT_TOKEN=root

# Enable audit log
vault audit enable file file_path=audit.log

# Enable and configure transit secrets engine
vault secrets enable transit
vault write -f transit/keys/autounseal

# Create an autounseal policy
vault policy write autounseal -<<EOF
path "transit/encrypt/autounseal" {
   capabilities = [ "update" ]
}

path "transit/decrypt/autounseal" {
   capabilities = [ "update" ]
}
EOF

# Create a token for Vault 2 to use for root key encryption
vault token create -orphan -policy="autounseal" -wrap-ttl=120 -period=24h -field=wrapping_token > wrapping-token.txt

Make the file executable.
```
$ chmod +x autounseal-transit-setup.sh
```

Run the script.

$ ./autounseal-transit-setup.sh

Output:

Success! Enabled the file audit device at: file/
Success! Enabled the transit secrets engine at: transit/
Key                       Value
---                       -----
allow_plaintext_backup    false
auto_rotate_period        0s
deletion_allowed          false
derived                   false
exportable                false
imported_key              false
keys                      map[1:1693608276]
latest_version            1
min_available_version     0
min_decryption_version    1
min_encryption_version    0
name                      autounseal
supports_decryption       true
supports_derivation       true
supports_encryption       true
supports_signing          false
type                      aes256-gcm96
Success! Uploaded policy: autounseal

Now, you can skip to Step 2: Configure auto-unseal (Vault 2) section.

Start Vault 1 server

Start a Vault dev server with root as the root token. Output the system log in a file named vault-1.log.
```
$ vault server -dev -dev-root-token-id root > vault-1.log 2>&1 &
```
The Vault dev server defaults to running at 127.0.0.1:8200. The server is also initialized and unsealed.
Insecure operation
Do not run a Vault dev server in production. This approach is only used here to simplify the unsealing process for this demonstration.
Export an environment variable for the vault CLI to address the Vault server.
```
$ export VAULT_ADDR=http://127.0.0.1:8200
```
Export an environment variable for the vault CLI to authenticate with the Vault server.
```
$ export VAULT_TOKEN=root
```

Setup the transit secrets engine

The first step is to enable and configure the transit secrets engine on Vault 1.

Enable an audit device if it hasn't been enabled already so that you can examine the audit log later in Step 3.
```
$ vault audit enable file file_path=audit.log
```
Execute the following command to enable the transit secrets engine.
```
$ vault secrets enable transit
```
Execute the following command to create an encryption key named, "autounseal".
```
$ vault write -f transit/keys/autounseal
```
Note
To learn more about the transit secrets engine, refer to the Encryption as a Service: Transit Secrets Engine.

Create a policy named autounseal which permits update against transit/encrypt/autounseal and transit/decrypt/autounseal paths.

$ vault policy write autounseal -<<EOF
path "transit/encrypt/autounseal" {
   capabilities = [ "update" ]
}

path "transit/decrypt/autounseal" {
   capabilities = [ "update" ]
}
EOF

Output:

Success! Uploaded policy: autounseal

Create an orphan periodic client token with the autounseal policy attached and response wrap it with TTL of 120 seconds. Store the generated wrapping token value in a file, wrapping-token.txt.
```
$ vault token create -orphan -policy="autounseal" \
   -wrap-ttl=120 -period=24h \
   -field=wrapping_token > wrapping-token.txt
```
Note
Periodic tokens can be renewed within the period. By default, the transit autounseal token is renewed automatically. An orphan token does not have a parent token and will not be revoked when the token that created it expires. Learn more about token hierarchies.
The generated wrapping_token is what you pass to Vault 2 so that its root key can be decrypted to unseal.
```
$ cat wrapping-token.txt
```
Example output:
```
hvs.CAESIJvhwiAb8-Xx3mc23ju8W4Kcp8JBXs6LOwJQ3ILxcaWGGh4KHGh2cy4zVHRKekRWdldyNERtbm11ZE9DTldWa0s
```

Enable an audit device if it hasn't been enabled already so that you can examine the audit log later in Step 3.

$ curl --header "X-Vault-Token: $VAULT_TOKEN" \
   --request PUT \
   --data '{"type":"file", "options":{"file_path":"audit.log"}}' \
   $VAULT_ADDR/v1/sys/audit/file

Be sure to pass your valid token in the X-Vault-Token header.

Enable transit secrets engine and create a key named, "autounseal".

$ curl --header "X-Vault-Token: $VAULT_TOKEN" \
   --request POST \
   --data '{"type":"transit"}' \
   $VAULT_ADDR/v1/sys/mounts/transit

Create an encryption key named, "autounseal".
```
$ curl --header "X-Vault-Token: $VAULT_TOKEN" \
   --request POST \
   $VAULT_ADDR/v1/transit/keys/autounseal
```
Tip
To learn more about the transit secrets engine, refer to the Encryption as a Service: Transit Secrets Engine.

Create the API request payload containing the autounseal policy definition which permits update against transit/encrypt/autounseal and transit/decrypt/autounseal paths.

$ tee payload-policy.json <<EOF
{
  "policy": "path \"transit/encrypt/autounseal\" {\n   capabilities = [ \"update\" ]\n}\n\npath \"transit/decrypt/autounseal\" {\n   capabilities = [ \"update\" ]\n}\n"
}
EOF

Create a policy named autounseal.
```
$ curl --header "X-Vault-Token: $VAULT_TOKEN" --request PUT \
   --data @payload-policy.json \
   $VAULT_ADDR/v1/sys/policies/acl/autounseal
```
Create a policy to permit update against transit/encrypt/<key_name> and transit/decrypt/<key_name> where the <key_name> is the name of the encryption key you created in the previous step.
Create an orphan periodic client token with the autounseal policy attached and response wrap it with TTL of 120 seconds. Store the generated wrapping token value in a file, wrapping-token.txt.
```
$ curl --header "X-Vault-Wrap-TTL: 120" \
   --header "X-Vault-Token: $VAULT_TOKEN" \
   --request POST \
   --data '{"policies":["autounseal"], "period":"24h"}' \
   $VAULT_ADDR/v1/auth/token/create-orphan \
   | jq -r ".wrap_info.token" > wrapped-token.txt
```
Note
Periodic tokens can be renewed within the period. By default the transit autounseal token is renewed automatically. An orphan token does not have a parent token and will not be revoked when the token that created it expires. Learn more about token hierarchies.
The generated token is what you pass to Vault 2 so that its root key can be decrypted to unseal.
```
$ cat wrapping-token.txt
```
Example output:
```
hvs.CAESIJvhwiAb8-Xx3mc23ju8W4Kcp8JBXs6LOwJQ3ILxcaWGGh4KHGh2cy4zVHRKekRWdldyNERtbm11ZE9DTldWa0s
```

Open a web browser and launch the Vault UI (http://127.0.0.1/ui) and then login with token, root.

Select Enable new engine from the Secrets engines page.
Select the Transit radio button and click Next.
Leave the Path as transit, and click Enable Engine.
Select Create encryption key and enter autounseal in the Name field.
Click Create encryption key to complete.
Select Policies, and then select Create ACL policy.
Enter autounseal in the Name field, and then enter the following policy in the Policy text field.
```
path "transit/encrypt/autounseal" {
  capabilities = [ "update" ]
}

path "transit/decrypt/autounseal" {
  capabilities = [ "update" ]
}
```
Create a policy to permit update against transit/encrypt/<key_name> and transit/decrypt/<key_name> where the <key_name> is the name of the encryption key you created in the previous step.
Click Create policy to complete.
Click the Vault CLI shell icon (>_) to open a command shell, and run the following command to geerate a token.
```
$ vault write auth/token/create-orphan policies=autounseal -wrap-ttl=120 -field=token
```
Copy the returned value, and store it in a file, wrapping-token.txt.

Step 2: Configure auto-unseal (Vault 2)

Now, start a second Vault instance which listens to port 8100. The server configuration file should define a seal stanza with parameters properly set based on the tasks you performed in Step 1.

Scenario Overview

Execute the following command to unwrap the secrets passed from Vault 1 and retrieve the token value.

$ vault unwrap -field=token $(cat wrapping-token.txt)

Example output:

hvs.CAESIErEuLhua677_F9Uma0cZ30bpWn1-WafzoQmTGRDxKyaGh4KHGh2cy5VenI2RGw4cHJBVElTUm9MNUYwaUhUNlY

Open a terminal where you will run the Vault 2 server, and set VAULT_TOKEN environment variable whose value is the client token you just unwrapped.
Example:
```
$ export VAULT_TOKEN="hvs.CAESIErEuLhua677_F9Uma0cZ30bpWn1-WafzoQmTGRDxKyaGh4KHGh2cy5VenI2RGw4cHJBVElTUm9MNUYwaUhUNlY"
```
Create a server configuration file (config-autounseal.hcl) to start a second Vault instance (Vault 2).
```
$ tee config-autounseal.hcl <<EOF
disable_mlock = true
ui=true

storage "raft" {
   path    = "./vault/vault-2"
   node_id = "vault-2"
}

listener "tcp" {
  address     = "127.0.0.1:8100"
  tls_disable = "true"
}

seal "transit" {
  address = "$VAULT_ADDR"
  disable_renewal = "false"
  key_name = "autounseal"
  mount_path = "transit/"
  tls_skip_verify = "true"
}

api_addr = "http://127.0.0.1:8100"
cluster_addr = "https://127.0.0.1:8101"
EOF
```
Notice that the address points to the Vault server listening to port 8200 (Vault 1). The key_name and mount_path match to what you created in Step 1.
Note
The seal stanza does not set the token value since it's already set as VAULT_TOKEN environment variable.
Warning
Although the listener stanza disables TLS (tls_disable = "true") for this tutorial, Vault should always be used with TLS in production to provide secure communication between clients and the Vault server. It requires a certificate file and key file on each Vault host.
Create the raft path directory as configured in the storage stanza.
```
$ mkdir -p vault/vault-2
```

Start the server using the configuration.

$ vault server -config=config-autounseal.hcl

Open another terminal and initialize your second Vault server (Vault 2).
```
$ VAULT_ADDR=http://127.0.0.1:8100 vault operator init
```
By passing the VAULT_ADDR, the subsequent command gets executed against the second Vault server (http://127.0.0.1:8100).
Example output:
```
Recovery Key 1: iz1XWxe4CM+wrOGqRCx8ex8kB2XvGJEdfjhXFC+MA6Rc
Recovery Key 2: rKZETr6IAy686IxfO3ZBKXPDAOkkwkpSepIME+bjeUT7
Recovery Key 3: 4XA/KJqFOm+jzbBkKQuRVePEYPrQe3H3TmFVmdlUjRFv
Recovery Key 4: lfnaYoZufP0uhooO3mHDAKGNZB5HLP9HYYb+LAfKkUmd
Recovery Key 5: L169hHj3DMpphGsOnS8TEz3Febvdx3vsG3Xr8kGWdUtW

Initial Root Token: s.AWnDagUkKNNbvkENiL72wysn

Success! Vault is initialized

Recovery key initialized with 5 key shares and a key threshold of 3. Please
securely distribute the key shares printed above.
```
Note
The initialization generates recovery keys (instead of unseal keys) when using auto-unseal. Some of the Vault operations still require Shamir keys. For example, to regenerate a root token, each key holder must enter their recovery key. Similar to unseal keys, you can specify the number of recovery keys and the threshold using the -recovery-shares and -recovery-threshold flags. It is strongly recommended to initialize Vault with PGP.

Check the Vault 2 server status. It is now successfully initialized and unsealed.

$ VAULT_ADDR=http://127.0.0.1:8100 vault status

Key                      Value
---                      -----
Recovery Seal Type       shamir
Initialized              true
Sealed                   false
Total Recovery Shares    5
Threshold                3
# ...snip...

Notice that it shows Total Recovery Shares instead of Total Shares. The transit secrets engine is solely responsible for protecting the root key of Vault 2.

Step 3: Verify auto-unseal

When you stop and start the Vault 2 server, it comes up in the unsealed state and ready for operations.

To verify that Vault 2 gets automatically unseal, press Ctrl + C to stop the Vault 2 server where it is running.

...snip...
[INFO]  core.cluster-listener: rpc listeners successfully shut down
[INFO]  core: cluster listeners successfully shut down
[INFO]  core: vault is sealed

Note that Vault 2 is now sealed.

When you try to check the Vault status, it returns the "connection refused" message.

$ VAULT_ADDR=http://127.0.0.1:8100 vault status

Error checking seal status: Get "http://127.0.0.1:8100/v1/sys/seal-status": dial
tcp 127.0.0.1:8100: connect: connection refused

Press the upper-arrow key, and execute the vault server -config=config-autounseal.hcl command again to start Vault 2 and see what happens.

$ vault server -config=config-autounseal.hcl

==> Vault server configuration:

                Seal Type: transit
          Transit Address: $VAULT_ADDR
        Transit Key Name: autounseal
      Transit Mount Path: transit/
                      Cgo: disabled
              Listener 1: tcp (addr: "0.0.0.0:8100", cluster address: "0.0.0.0:8101", max_request_duration: "1m30s", max_request_size: "33554432", tls: "disabled")
                Log Level: info
                    Mlock: supported: true, enabled: false
                  Storage: file
                  Version: Vault v1.1.0
              Version Sha: 36aa8c8dd1936e10ebd7a4c1d412ae0e6f7900bd

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

[WARN]  no `api_addr` value specified in config or in VAULT_API_ADDR; falling back to detection if possible, but this value should be manually set
[INFO]  core: stored unseal keys supported, attempting fetch
[INFO]  core: vault is unsealed
# ...snip...

Notice that the Vault server is already unsealed. The Transit Address is set to your Vault 1 which is listening to port 8200 ($VAULT_ADDR).

Check the Vault 2 server status.

$ VAULT_ADDR=http://127.0.0.1:8100 vault status

Key                      Value
---                      -----
Recovery Seal Type       shamir
Initialized              true
Sealed                   false
Total Recovery Shares    5
Threshold                3
# ...snip...

Vault 2 is automatically unsealed.

Now, examine the audit log in Vault 1.

$ tail -f audit.log | jq

# ...snip...
"request": {
  "id": "a46719eb-eee0-92a4-2da6-6c7de77fd410",
  "operation": "update",
  "client_token": "hmac-sha256:ce8613487054dadb36a9d08da1f5a4bbee2fbfc1ef1ec5ebdeec696df7823e69",
  "client_token_accessor": "hmac-sha256:f3b6cb798605835e8a00bafa9e0e16fc0534b8923b31e499f2c8e694f6b69158",
  "namespace": {
    "id": "root",
    "path": ""
  },
  "path": "transit/decrypt/autounseal",
    # ...snip...
  "remote_address": "127.0.0.1",
  "wrap_ttl": 0,
  "headers": {}
},
# ...snip...
}

You should see an update request against the transit/decrypt/autounseal path. The remote_address is 127.0.0.1 in this example since Vault 1 and Vault 2 are both running locally. If the Vault 2 is running on a different host, the audit log will show the IP address of the Vault 2 host.

Warning

If a security incident forces you to seal the Vault server using the vault operator seal command, it requires the threshold number of recovery keys to unseal Vault and bring it back to operation.

Clean up

Execute the following command to stop the Vault servers.
```
$ pgrep -f vault | xargs kill
```
Unset the VAULT_TOKEN and VAULT_ADDR environment variables.
```
$ unset VAULT_TOKEN VAULT_ADDR
```

Delete the generated files.

$ rm config-autounseal.hcl vault-1.log audit.log wrapping-token.txt

Delete the raft directory.
```
$ rm -r vault/
```

Help and reference

Harden server deployments

Recover from lost quorum

This tutorial also appears in:

4 tutorials

Auto unseal
Automatically unsealing Vault reduces the operational complexity of keeping the Vault unseal keys secure.
- Vault