Introduction to tokens
Vault issues a token to a client upon successful authentication. A token validates a Vault clients access to Vault and what actions the client can perform. Most actions in Vault require a token. This tutorial provides context for how and why tokens are used in Vault. In later tutorials, you will create tokens using the Vault CLI, HTTP API, UI, and Terraform.
Scenario
HashiCups is preparing for their Vault proof-of-concept (POC) and need to understand how clients, both human and machine, receive authorization to perform operations in a Vault cluster. Alice from the architect team and Oliver from operations meet with HashiCorp to understand the authentication and authorization process for Vault.
HashiCorp will show several important concepts using the Vault CLI. In later tutorials, you will learn to use a number of different interfaces to interact with Vault.
Alice has designed the Vault POC to use the Vault userpass
auth method to
allow teams to authenticate to Vault with a username and password. Workloads
will use the kubernetes
auth method to allow resources such as pods managed by
Kubernetes to authenticate.
What is a Vault token
Before a client, either a human or machine, can perform an operation in Vault that client must authenticate to Vault. Upon successful authentication, Vault issues a token from the Vault auth method used to authenticate with the Vault cluster.
When you start a Vault cluster, like you did in the set up Vault tutorial, it
starts with a root
token (or an admin
token for HCP Vault Dedicated). This
initial token provides full access to Vault, similar to the root
user in
Linux or domain administrator in Active Directory. The token
auth method
generates the token and attaches the root
policy to the token.
Note
A policy authorizes actions an entity can perform in the Vault cluster. You will learn more about policies, and how to author them in a later tutorial.
Root token use needs to be extremely guarded in production environments because it provides full access to the Vault server. Only use the root token for initial configuration of Vault, or for emergency access.
How Vault issues tokens
Alice has selected the userpass
and kubernetes
auth methods for the
HashiCups POC. The userpass
auth method acts similar to an identity provider,
storing a list of usernames, passwords, and policies assigned to the username.
The kubernetes
auth method allows Vault to validate a workload managed by
Kubernetes against the Kubernetes API.
A client will authenticate against an auth method, and upon successful authentication, Vault will issue a token to the client.
The basic authentication flow is:
- A client attempts to authenticate with Vault.
- Vault verifies the identity with a trusted provider. For the HashiCups POC
this would be Vault itself for
userpass
auth method or the Kubernetes API for thekubernetes
auth method. - The identity provider validates the Vault client's identity.
- Vault returns a token to the client.
Regardless of the auth method used, a client must first authenticate with Vault to receive a token.
Token metadata
Every token includes information about that token. Oliver logs into a Vault dev mode server to learn about the information associated with a root token.
Example:
The login information returned includes:
- Token duration: How log the token is valid for, typically expressed as the time-to-live (TTL) value in the format of a time string (i.e. "8h" or "30d"). The value is an infinity symbol for a root token because a root token does not have a TTL and can't be renewed, but is valid forever unless you revoke it.
- Accessor: A unique ID that can be used to lookup, renew, or revoke a token.
- Policies: One or more policies attached to the token that defines the actions authorized to be performed in Vault.
The default Vault TTL is 32 days. You can override the default TTL globally for your Vault cluster, or you can set a specific TTL on a role definition for a specific auth method. You can also set a max TTL which defines how long a token can be renewed for.
In addition to TTL and max TTL, you can set the number of uses for tokens. The tokens with a use limit expire at the end of their last use regardless of their remaining TTLs. On the same note, use limit tokens expire at the end of their TTLs regardless of their remaining uses.
The root token, shown in the code block above, is the only type of Vault token that can be set to never expire.
Token accessors are a unique ID that can be used to perform operations on a token such as renewing a token, looking up additional token metadata, or revoking a token.
Example:
A token accessor, unlike a token, does not provide access to Vault.
Example:
All tokens include one or more policies. In the example above, the root
token
is attached. Another common policy to see attached to a token is the default
policy. You will create policies in later tutorials to provide access to Vault.
What can you use a token accessor for?
Token accessors can be used to perform actions such as renewing or revoking a token, but does not provide access to Vault.
Types of tokens
There are two main types of Vault tokens: service tokens and batch tokens. Vault persists the service tokens in its storage backend. You can renew a service token or revoke it as necessary.
Vault does not persist batch tokens. Batch tokens are encrypted binary large objects (blobs) that carry enough information to perform Vault actions. This makes batch tokens lightweight and scalable but they lack the flexibility and features of service tokens.
Note
There is a recovery token which is a special token used when Vault is running in recovery mode.
You will learn more about recovery tokens in a later tutorial.
Token prefix
Tokens have a prefix that indicates their type.
Token Type | Vault 1.10 and later |
---|---|
Service tokens | hvs.string |
Batch tokens | hvb.string |
Recovery tokens | hvr.string |
Service tokens vs. batch tokens
As the number of machines and apps using Vault for secret management scales, Vault must manage the growing number of client tokens. The creation of service tokens can affect Vault performance since they must be replicated across the primary and secondary clusters. On the other hand, batch tokens are neither persisted to disk nor live in memory; they are not a part of the data replication process.
Due to the lack of features with batch tokens, it's preferable to use service tokens in most use cases. There are, however, use cases for batch tokens. Consider a scenario where you have a large number of containers (e.g. 100,000s) start up and all request a token from Vault. The use of batch tokens can reduce the stress on the storage backend and improve the overall performance.
Batch tokens are designed to be lightweight with limited flexibility. The following table highlights the differences between batch tokens and service tokens.
Token features | Service tokens | Batch tokens |
---|---|---|
Can be root tokens | Yes | No |
Can create child tokens | Yes | No |
Renewable | Yes | No |
Manually Revocable | Yes | No |
Can be periodic | Yes | No |
Can have explicit Max TTL | Yes | No (always uses a fixed TTL) |
Has accessors | Yes | No |
Has Cubbyhole | Yes | No |
Revoked with parent (if not orphan) | Yes | Stops Working |
Dynamic secrets lease assignment | Self | Parent (if not orphan) |
Can be used across Performance Replication clusters | No | Yes (if orphan) |
Creation scales with performance standby node count | No | Yes |
Cost | Heavyweight; multiple storage writes per token creation | Lightweight; no storage cost for token creation |
Using a root token or a token with sudo capabilities can generate periodic tokens. Periodic tokens have a TTL (validity period), but no max TTL; therefore, they may live for an infinite duration of time so long as they are renewed within their TTL. This is useful for long-running services that cannot handle regenerating a token.
Example:
Another type of service token is an orphan token. Like periodic tokens, you must create an orphan token with a root token or a token with sudo capability.
Orphan tokens are not children of their parent; therefore, orphan tokens do not expire when their parent does. Orphan tokens still expire when their own max TTL is reached.
Example:
Renew a token in Vault
A token can be renewed multiple times up to the maximum TTL set for the token.
Example:
The token will be valid for a period
of 1m
and can be renewed for up to 2m
based on the explicit-max-ttl
value.
The token, even if renewed, will not be valid beyond the explicit-max-ttl
value.
Example:
Revoke a token in Vault
A token will no longer be valid once the TTL or max TTL is reached. You can also manually revoke a token. Consider a scenario where a token was commited to source. This would create a security incident because a valid token can be seen in the source code repository. In this type of scenario, you can revoke a token manually.
Example:
You can manually revoke the token using either the token
, or the
token_accessor
.
Summary
Tokens are central to understanding and managing Vault. Vault issues a token upon successful authentication. Tokens include metadata for managing the token, as well as policies that defines the actions a token is able to do in Vault. Vault supports different types of tokens depending on the type of workload authenticating with Vault.