Worker configuration

Workers authenticate to Boundary using an activation token. They require an accessible directory defined by auth_storage_path for credential storage and rotation. Transport level communication between the worker & Controller is secured through PKI.

Example (not safe for production!):

worker {
  auth_storage_path="/var/lib/boundary"
  initial_upstreams = ["10.0.0.1"]
}

Authorization methods

There are three mechanisms that can be used to initially register a worker to the cluster, Controller-Led, Worker-Led, and registration through an external KMS.

Controller-led authorization flow

In this flow, the operator fetches an activation token from the controller's workers:create:controller-led action (on the CLI, this is via boundary workers create controller-led). That activation token is given to the worker via the controller_generated_activation_token parameter. This can be done either directly or via an env var or file by using env:// or file:// syntax:

worker {
  auth_storage_path="/var/lib/boundary"
  initial_upstreams = ["10.0.0.1"]
  controller_generated_activation_token = "neslat_........."
  # controller_generated_activation_token = "env://ACT_TOKEN"
  # controller_generated_activation_token = "file:///tmp/worker_act_token"
}

Once the worker starts, it reads this token and uses it to authorize to the cluster. Note that this token is one-time-use; it is safe to keep it here even after the worker has successfully authorized and authenticated, as it will be unusable at that point.

Note: If this value is not present at worker startup time and the worker is not authorized, it will print and write out suitable information for the worker-led flow, described below. If the worker-led flow has not been used to authorize the worker, and the controller-generated activation token is provided and the worker restarted, it will make use of it.

Worker-led authorization flow

In this flow, the worker prints out an authorization request token to two places: the startup information printed to stdout, and a file called auth_request_token in the base of the configured auth_storage_path. This token can be submitted to a controller at the workers:create:worker-led path; on the CLI this would be via boundary workers create worker-led -worker-generated-auth-token. No values are needed in the configuration file.

KMS-led authorization & authentication flow

In this flow, the worker authenticates upstream, either to a controller or worker, using a shared KMS provided by the customer. This mechanism auto-registers the worker in addition to authenticating it, and does not require on-disk storage for credentials since each time it connects, it re-authenticates using the trusted KMS.

Optionally with the Multi-Hop workers feature, trusted Workers can authenticate downstream nodes using a separate KMS.

Workers using KMS-led authorization require a name field. This specifies a unique name of this worker within the Boundary cluster and must be unique across workers. The name value can be:

• a direct name string (must be all lowercase)
• a reference to a file on disk (file://) from which the name is read
• an env var (env://) from which the name is read.

Workers using KMS-led authorization accept an optional description field. The description value can be:

• a direct description string
• a reference to a file on disk (file://) from which the name is read
• an env var (env://) from which the name is read.

worker {
  name = "example-worker"
  description = "An example worker"
  public_addr = "5.1.23.198"
}

Workers using the KMS authorization flow also require a KMS block designated for worker-auth. This is the KMS configuration for authentication between the workers and controllers and must be present. Example (not safe for production!):

kms "aead" {
    purpose = "worker-auth"
    aead_type = "aes-gcm"
    key = "X+IJMVT6OnsrIR6G/9OTcJSX+lM9FSPN"
    key_id = "global_worker-auth"
}

The upstream controller or worker must have a kms block that references the same key and purpose. If both a controller and worker are running as the same server process, only one stanza is needed.

For Multi-Hop workers, It is also possible to specify a kms block with the downstream-worker-auth purpose. If specified, this will be a separate KMS that can be used for authenticating new downstream nodes. Blocks with this purpose can be specified multiple times. This allows a single upstream node to authenticate with one key to its own upstream (via the worker-auth purpose) and then serve as an authenticating upstream to nodes across various networks, each with their own separate KMS system or key:

kms "aead" {
    purpose = "downstream-worker-auth"
    aead_type = "aes-gcm"
    key = "XthZVtFtBD1Bw1XwAWhZKVrIwRhR7HcZ"
    key_id = "iot-nodes-auth"
}

kms "aead" {
    purpose = "downstream-worker-auth"
    aead_type = "aes-gcm"
    key = "OLFhJNbEb3umRjdhY15QKNEmNXokY1Iq"
    key_id = "production-nodes-auth"
}

In the examples above we are encoding key bytes directly in the configuration file. This is because we are using the aead method where you directly supply a key; in production you'd want to use a KMS such as AWS KMS, GCP CKMS, Azure KeyVault, or HashiCorp Vault. For a complete guide to all available KMS types, see our KMS documentation.

Complete configuration example

listener "tcp" {
    purpose = "proxy"
    tls_disable = true
    address = "127.0.0.1"
}

worker {
  # Name attr must be unique across workers
  name = "demo-worker-1"
  description = "A default worker created for demonstration"

  # Workers must be able to reach upstreams on :9201
  initial_upstreams = [
    "10.0.0.1",
    "10.0.0.2",
    "10.0.0.3",
  ]

  public_addr = "myhost.mycompany.com"

  tags {
    type   = ["prod", "webservers"]
    region = ["us-east-1"]
  }
}

# must be same key as used on controller config
kms "aead" {
    purpose = "worker-auth"
    aead_type = "aes-gcm"
    key = "X+IJMVT6OnsrIR6G/9OTcJSX+lM9FSPN"
    key_id = "global_worker-auth"
}

initial_upstreams are used to connect to upstream Boundary clusters.

Resources

For more on how tags{} in the above configuration are used to facilitate routing to the correct target, refer to the Worker Tags page.

KMS configuration

When using Controller or Worker-led Authentication, a worker’s generated activation token is stored in clear-text on disk. Using an external KMS, a Workers' credentials can be encrypted by including an optional KMS stanza with the purpose worker-auth-storage.

Example (not safe for production!):

kms "aead" {
  purpose = "worker-auth-storage"
    aead_type = "aes-gcm"
    key = "X+IJMVT6OnsrIR6G/9OTcJSX+lM9FSPN"
    key_id = "worker-auth-storage"
}

Session recording

Session recording requires at least one Worker with access to local and remote storage. Workers used for session recording require an accessible directory defined by recording_storage_path for storing in-progress session recordings. On session closure, a local session recording is moved to remote storage and deleted locally.

The recording_storage_minimum_available_capacity value determines the minimum amount of storage space that is required for workers to perform session recording operations. If a worker is at or below the storage threshold, Boundary does not use the worker to record sessions or play back recordings.

Development example:

worker {
  auth_storage_path="/var/lib/boundary"
  initial_upstreams = ["10.0.0.1"]
  recording_storage_path="/local/storage/directory"
  recording_storage_minimum_available_capacity="500MB"
}