LDAP secrets engine
Note: This engine can use external X.509 certificates as part of TLS or signature validation. Verifying signatures against X.509 certificates that use SHA-1 is deprecated and is no longer usable without a workaround starting in Vault 1.12. See the deprecation FAQ for more information.
The LDAP secrets engine provides management of LDAP credentials as well as dynamic creation of credentials. It supports integration with implementations of the LDAP v3 protocol, including OpenLDAP, Active Directory, and IBM Resource Access Control Facility (RACF).
The secrets engine has three primary features:
Setup
Enable the LDAP secret engine:
By default, the secrets engine will mount at the name of the engine. To enable the secrets engine at a different path, use the
-path
argument.Configure the credentials that Vault uses to communicate with LDAP to generate passwords:
Note: it's recommended a dedicated entry management account be created specifically for Vault.
Rotate the root password so only Vault knows the credentials:
Note: it's not possible to retrieve the generated password once rotated by Vault. It's recommended a dedicated entry management account be created specifically for Vault.
Schemas
The LDAP Secret Engine supports three different schemas:
OpenLDAP
By default, the LDAP Secret Engine assumes the entry password is stored in userPassword
.
There are many object classes that provide userPassword
including for example:
Resource access control facility (RACF)
For managing IBM's Resource Access Control Facility (RACF) security system, the secret
engine must be configured to use the schema racf
.
Generated passwords must be 8 characters or less to support RACF. The length of the password can be configured using a password policy:
Active directory (AD)
For managing Active Directory instances, the secret engine must be configured to use the
schema ad
.
Static credentials
Setup
Configure a static role that maps a name in Vault to an entry in LDAP. Password rotation settings will be managed by this role.
Request credentials for the "hashicorp" role:
Password rotation
Passwords can be managed in two ways:
- automatic time based rotation
- manual rotation
Auto password rotation
Passwords will automatically be rotated based on the rotation_period
configured
in the static role (minimum of 5 seconds). When requesting credentials for a static
role, the response will include the time before the next rotation (ttl
).
Auto-rotation is currently only supported for static roles. The binddn
account used
by Vault should be rotated using the rotate-root
endpoint to generate a password
only Vault will know.
Manual rotation
Static roles can be manually rotated using the rotate-role
endpoint. When manually
rotated the rotation period will start over.
Deleting static roles
Passwords are not rotated upon deletion of a static role. The password should be manually rotated prior to deleting the role or revoking access to the static role.
Dynamic credentials
Setup
Dynamic credentials can be configured by calling the /role/:role_name
endpoint:
Note: The rollback_ldif
argument is optional, but recommended. The statements within rollback_ldif
will be
executed if the creation fails for any reason. This ensures any entities are removed in the event of a failure.
To generate credentials:
The distinguished_names
field is an array of DNs that are created from the creation_ldif
statements. If more than
one LDIF entry is included, the DN from each statement will be included in this field. Each entry in this field
corresponds to a single LDIF statement. No de-duplication occurs and order is maintained.
LDIF entries
User account management is provided through LDIF entries. The LDIF entries may be a base64-encoded version of the LDIF string. The string will be parsed and validated to ensure that it adheres to LDIF syntax. A good reference for proper LDIF syntax can be found here.
Some important things to remember when crafting your LDIF entries:
- There should not be any trailing spaces on any line, including empty lines
- Each
modify
block needs to be preceded with an empty line - Multiple modifications for a
dn
can be defined in a singlemodify
block. Each modification needs to close with a single dash (-
)
Active directory (AD)
For Active Directory, there are a few additional details that are important to remember:
To create a user programmatically in AD, you first add
a user object and then modify
that user to provide a
password and enable the account.
Passwords in AD are set using the
unicodePwd
field. This must be proceeded by two (2) colons (::
).When setting a password programmatically in AD, the following criteria must be met:
- The password must be enclosed in double quotes (
" "
) - The password must be in
UTF16LE
format - The password must be
base64
-encoded - Additional details can be found here
- The password must be enclosed in double quotes (
Once a user's password has been set, it can be enabled. AD uses the
userAccountControl
field for this purpose:- To enable the account, set
userAccountControl
to512
- You will likely also want to disable AD's password expiration for this dynamic user account. The
userAccountControl
value for this is:65536
userAccountControl
flags are cumulative, so to set both of the above two flags, add up the two values (512 + 65536 = 66048
): setuserAccountControl
to66048
- See here
for details on
userAccountControl
flags
- To enable the account, set
sAMAccountName
is a common field when working with AD users. It is used to provide compatibility with legacy
Windows NT systems and has a limit of 20 characters. Keep this in mind when defining your username_template
.
See here for additional details.
Since the default username_template
is longer than 20 characters which follows the template of v_{{.DisplayName}}_{{.RoleName}}_{{random 10}}_{{unix_time}}
, we recommend customising the username_template
on the role configuration to generate accounts with names less than 20 characters. Please refer to the username templating document for more information.
With regard to adding dynamic users to groups, AD doesn't let you directly modify a user's memberOf
attribute.
The member
attribute of a group and memberOf
attribute of a user are
linked attributes. Linked attributes are
forward link/back link pairs, with the forward link able to be modified. In the case of AD group membership, the
group's member
attribute is the forward link. In order to add a newly-created dynamic user to a group, we also
need to issue a modify
request to the desired group and update the group membership with the new user.
Active directory LDIF example
The various *_ldif
parameters are templates that use the go template
language. A complete LDIF example for creating an Active Directory user account is provided here for reference:
Service account Check-Out
Service account check-out provides a library of service accounts that can be checked out by a person or by machines. Vault will automatically rotate the password each time a service account is checked in. Service accounts can be voluntarily checked in, or Vault will check them in when their lending period (or, "ttl", in Vault's language) ends.
The service account check-out functionality works with various schemas, including OpenLDAP, Active Directory, and RACF. In the following usage example, the secrets engine is configured to manage a library of service accounts in an Active Directory instance.
First we'll need to enable the LDAP secrets engine and tell it how to securely connect to an AD server.
Our next step is to designate a set of service accounts for check-out.
In this example, the service account names of fizz@example.com
and buzz@example.com
have
already been created on the remote AD server. They've been set aside solely for Vault to handle.
The ttl
is how long each check-out will last before Vault checks in a service account,
rotating its password during check-in. The max_ttl
is the maximum amount of time it can live
if it's renewed. These default to 24h
, and both use duration format strings.
Also by default, a service account must be checked in by the same Vault entity or client token that
checked it out. However, if this behavior causes problems, set disable_check_in_enforcement=true
.
When a library of service accounts has been created, view their status at any time to see if they're available or checked out.
To check out any service account that's available, simply execute:
If the default ttl
for the check-out is higher than needed, set the check-out to last
for a shorter time by using:
This can be a nice way to say, "Although I can have a check-out for 24 hours, if I haven't checked it in after 30 minutes, I forgot or I'm a dead instance, so you can just check it back in."
If no service accounts are available for check-out, Vault will return a 400 Bad Request.
To extend a check-out, renew its lease.
Renewing a check-out means its current password will live longer, since passwords are rotated
anytime a password is checked in either by a caller, or by Vault because the check-out ttl
ends.
To check a service account back in for others to use, call:
Most of the time this will just work, but if multiple service accounts are checked out by the same caller, Vault will need to know which one(s) to check in.
To perform a check-in, Vault verifies that the caller should be able to check in a given service account. To do this, Vault looks for either the same entity ID used to check out the service account, or the same client token.
If a caller is unable to check in a service account, or simply doesn't try,
Vault will check it back in automatically when the ttl
expires. However, if that is too long,
service accounts can be forcibly checked in by a highly privileged user through:
Or, alternatively, revoking the secret's lease has the same effect.
Password generation
This engine previously allowed configuration of the length of the password that is generated
when rotating credentials. This mechanism was deprecated in Vault 1.5 in favor of
password policies. This means the length
field should
no longer be used. The following password policy can be used to mirror the same behavior
that the length
field provides:
LDAP password policy
The LDAP secret engine does not hash or encrypt passwords prior to modifying values in LDAP. This behavior can cause plaintext passwords to be stored in LDAP.
To avoid having plaintext passwords stored, the LDAP server should be configured with an LDAP password policy (ppolicy, not to be confused with a Vault password policy). A ppolicy can enforce rules such as hashing plaintext passwords by default.
The following is an example of an LDAP password policy to enforce hashing on the
data information tree (DIT) dc=hashicorp,dc=com
:
API
The LDAP secrets engine has a full HTTP API. Please see the LDAP secrets engine API docs for more details.