Fork me on GitHub
Teleport

Locking

Session and Identity Locking

Teleport provides a locking mechanism to restrict access to a computing environment. System administrators can disable a compromised user or prevent access during cluster maintenance.

When a lock is in force, all interactions like SSH/DB/k8s connections and certificate requests matched by the lock's target are rejected. A lock can target the following objects or attributes:

  • a Teleport user by the user's name;
  • a Teleport RBAC role by the role's name;
  • an MFA device by the device's UUID;
  • an OS/UNIX login;
  • a Teleport node by the node's UUID (effectively unregistering it from the cluster).
Warning
Locking support was introduced in Teleport v7.1. To fully enforce the locks at all components, you should update all teleport binaries in your deployment.

Prerequisites

Verify that your Teleport client is connected:

$ tctl status

# Cluster  tele.example.com
# Version  7.1.3
# CA pin   sha256:sha-hash-here
Connecting to the cloud

To try this flow in the cloud, login into your cluster using tsh, then use tctl remotely:

$ tsh login --proxy=myinstance.teleport.sh
$ tctl status

Creating a new lock

Locks are modeled as resources with kind: lock. To create a new lock, one can run the tctl lock command:

tctl lock [email protected] --message="Suspicious activity." --ttl=10h

Created a lock with name "dc7cee9d-fe5e-4534-a90d-db770f0234a1".

Note that without specifying --ttl or --expires the created lock remains in force until explicitly removed with tctl rm. Refer to tctl lock --help for the list of all supported parameters.

Under the hood, tctl lock creates a resource:

kind: lock
version: v2
metadata:
  name: dc7cee9d-fe5e-4534-a90d-db770f0234a1
spec:
  target:
    user: [email protected]
  message: "Suspicious activity."
  expires: "2021-08-14T22:27:00Z"  # RFC3339 format

The kind: lock resources can also be created and updated using tctl create as per usual, see the Admin Guide for more details.

Lock in force

With a lock in force, all established connections involving the lock's target get terminated while any new requests are rejected.

Errors returned and warnings logged in this situation feature a message of the form:

lock targeting User:"[email protected]" is in force: Suspicious activity.

Locking mode

If a Teleport node or proxy cannot properly synchronize its local lock view with the backend, there is a decision to be made about whether to rely on the last known locks. This decision strategy is encoded as one of the two modes:

  • strict mode causes all interactions to be terminated when the locks are not guaranteed to be up to date;
  • best_effort mode keeps relying on the most recent locks.

The cluster-wide mode defaults to best_effort. You can set up default locking mode via API or CLI using resource cluster_auth_preference or static configuration file:

Create a YAML file cap.yaml or get the existing file using tctl get cap

kind: cluster_auth_preference
metadata:
  name: cluster-auth-preference
spec:
  locking_mode: best_effort
version: v2

Create a resource:

tctl create -f cap.yaml

cluster auth preference has been updated

It is also possible to configure the locking mode for a particular role:

kind: role
version: v4
metadata:
    name: example-role-with-strict-locking
spec:
    options:
       lock: strict

When none of the roles involved in an interaction specify the mode or when there is no user involved, the mode is taken from the cluster-wide setting.

With multiple potentially conflicting locking modes (the cluster-wide default and the individual per-role settings) a single occurrence of strict suffices for the local lock view to become evaluated strictly.

Have a suggestion or can’t find something?
IMPROVE THE DOCS