IDSYS documents are JSON-formatted objects and are specific to the back-end identity management system: Active Directory, LDAP, and Linux PAM.
SAML for SSO
With Gateway 7.1, Content UI 7.0, and Swarm UI 3.0 and higher, you can enable SSO (single sign-on) to the Swarm and Content UIs through a third-party identity provider, such as Google. See Enabling SSO with SAML. (v7.1)
Common IDSYS Fields
Below are the common fields within all IDSYS documents. Fields that are specific to the back-end identity management system are broken out into separate sections.
Field | Required | Description |
---|---|---|
name | No | Name of the IDSYS document; value is not used by Gateway |
description | No | Description of the IDSYS document; value is not used by Gateway |
comments | No | Comments about the IDSYS; may be any valid JSON object type |
cookieName | No1 | Cookie name used to store authentication token. Example: "token" |
tokenPath | No1 | URI path used for token authentication. Example: "/.TOKEN/" |
tokenAdmin | No1 | User that is allowed to view and delete authentication tokens for other users. |
1 For details regarding token-based authentication, see Token-Based Authentication.
When a user authenticates to the Gateway using HTTP Basic authentication (that is, not token-based authentication and not S3 HMAC), the user's password is stored in the normal field for LDAP or PAM and it may be hashed in whatever formats are supported by the system. For LDAP, this field is normally userPassword
; for PAM with the traditional Unix authentication mechanism, it is the second field in the /etc/shadow
file.
Password security
Plain-text passwords in both Gateway Configuration and IDSYS are replaced by encrypted versions on startup. Whenever you need to change management passwords, enter your new ones and restart Gateway, which will replace those strings with encrypted versions as part of its startup. (v7.1)
LDAP and AD Examples
These are examples of IDSYS documents for LDAP and Active Directory. They contain fields that are specific to LDAP as well as fields that are common to all IDSYS documents.
{"ldap": { "name" : "idsys-ldap", "description": "LDAP identity management configuration", "protocol" : "ldaps", "ldaphost": ["ldap.example.com", "ldap-sec.example.com"], "ldapport": 636, "adminDN": "uid=YOURUSERNAME,ou=Users,dc=example,dc=com", "adminPassword": "YOURPASSWORD", "userBase": "ou=Users,dc=example,dc=com", "groupBase": "ou=Groups,dc=example,dc=com", "userFilter": "objectclass=account", "groupMemberUidAttr": "memberUid", "cookieName": "token", "tokenPath": "/.TOKEN/", "tokenAdmin": "superuser@admindomain.example.com" } }
The block that begins with "uidAttribute" is what makes this one specific to Active Directory:
{"ldap": { "name": "idsys-ad", "description": "Active Directory example configuration", "protocol" : "ldaps", "ldaphost": "ad.mycompany.com", "ldapport": "636", "adminDN": "cn=BINDUSER,ou=Applications,dc=mycompany,dc=com", "adminPassword": "BINDPASSWORD", "userBase": "ou=Users,dc=mycompany,dc=com", "groupBase": "ou=Groups,dc=mycompany,dc=com", "uidAttribute":"sAMAccountName", "userFilter":"objectclass=*", "groupMemberDNAttr": "member", "cookieName": "token", "tokenPath": "/.TOKEN/", "tokenAdmin" : "caringoadmin@" }
LDAP and AD Fields
These are the fields within the IDSYS document that are specific to the LDAP or Active Directory back-end identity management system.
No nested or recursive groups
Field | Default | Required | Description |
---|---|---|---|
ldaphost | Yes | Host name or IP of the LDAP server as a string or a multiple servers as a list. Example: ["1.1.1.1", "1.1.1.2"] | |
ldapport | Yes | Port the LDAP service is running on | |
protocol | ldap | No | Set to "ldap" or "ldaps" |
referrals | follow | No | Set to "follow" or "ignore" to control how referrals are handled |
adminDN | Yes | DN used to bind to the LDAP server for queries | |
adminPassword | Yes | Password for adminDN user | |
userBase | Yes | DN where users are defined | |
groupBase | Yes | DN where groups are defined | |
uidAttribute | uid | No | Attribute name containing user's ID. Examples:
|
userFilter | Yes | Filter for user objects. Example: "objectclass=account" | |
groupMemberUidAttr | Yes1 | Group attribute whose values contain uid of member. Example: "memberUid" if OpenLDAP is configured for groups with "objectclass=posixgroup" | |
groupMemberDNAttr | Yes1 | Group attribute whose values contain DN of member. Example: "member" if OpenLDAP is configured for groups with "objectclass=groupOfNames"; also common with Active Directory | |
s3SecretKeyAttr | No2 | **Deprecated** User attribute whose value contains the user's S3 secret key in plain-text. If "userPassword" is used, you must ensure that it has a plaintext value since this is not the normal handling of this attribute. |
1 The groupMemberUidAttr
and groupMemberDNAttr
parameters are mutually exclusive and you must only define one of them in IDSYS.
2 The s3SecretKeyAttr
parameter is only needed when using S3 Protocol Personality with a user password stored in LDAP. It is not required when using token authentication exclusively.
The adminDN
and adminPassword
parameters define the credentials with which the Gateway binds to the LDAP system in order to perform queries and read records for users and groups. The adminDN
entity within LDAP needs to have read level access (rscdx
privileges) within the LDAP tree. It is not necessary to grant write or manage level access to Gateway.
- A user's name in an access control Policy document is the value of the LDAP attribute named by the
uidAttribute
parameter. By default this will be theuid
attribute of a user's LDAP record. - A group's name in an access control Policy document is the
cn
attribute for the group LDAP entity. The name of this attribute cannot be configured. A group's name may contain spaces and other non-alphanumeric characters.
Important
If you are using LDAPs with self-signed certificates, you must add your signer's public key PEM file to your Java keystore to avoid a SunCertPathBuilderException when Gateway queries the LDAP server.
Although previously this required running java keytool, you should now use the CentOS/RHEL utility update-ca-trust to add any CA to your system.
# cp ca.pem /etc/pki/ca-trust/source/anchors/
# update-ca-trust extract
PAM Example
There are no fields within the IDSYS document that are specific to the PAM back-end identity management system. If you are using PAM, follow this process to implement your identity management:
- Because the
root
user (uid=0) on this Content Gateway server cannot be used to authenticate to the Gateway, create another user (such assuperuser@admindomain.example.com
) on this server for this purpose. Copy and paste this example into your IDSYS document:
/etc/caringo/cloudscaler/idsys.json
.{"pam": { "name" : "idsys-pam", "description": "PAM identity management configuration", "cookieName": "token", "tokenPath": "/.TOKEN/", "tokenAdmin": "superuser@admindomain.example.com" } }
- Update the
tokenAdmin
to match your authentication user.
Modifying IDSYS
The root IDSYS configuration is stored in the idsys.json
file on the Gateway server's disk so that it is always available and so that an administrator can always modify it. In other words, you can't lock yourself out of the storage cluster. Changes to the local file take effect without the need to restart the Gateway.
Important
For modifying a tenant or storage domain's sub-resource through the management API, see Defined ETC Documents.
For details on modifying a storage domain's sub-resource through the storage API, see SCSP Context Sub-resources.
When updating an IDSYS sub-resource through the management API or the storage API, the entire JSON document with all fields must be provided in the update request, even if only one field is being modified.
Caution
IDSYS Precedence Model
The identity system is described by IDSYS documents can exist at the root, tenant, and storage domain within the system. When IDSYS documents exist at multiple levels in the hierarchy, the lowest level overrides the higher levels. When a lower level lacks an IDSYS, it inherits from a higher level.
For example, if only a Root IDSYS exists, all tenants and all storage domains will inherit from the Root IDSYS. In this case, there will only be one identity management system with one set of users and groups. If, however, the tenants each defined their own IDSYS, each tenant and the storage domains owned by them would share their own identity system that would be separate from the Root IDSYS. In this second case, the storage domains would inherit from their Tenant IDSYS.
The IDSYS inheritance also works at the field level, meaning that tenant and storage domain IDSYS documents can choose to override specific fields. For example, if tokenAdmin is defined in the Root IDSYS and not in the tenant or domain IDSYS, its value will be inherited by the tenant and domain levels. Similarly, the Root IDSYS may define the LDAP adminDN and adminPassword and let the tenant and domain IDSYS documents override the userBase and groupBase values.
- Single Company
- In this scenario, the company has one identity management system, there is one tenant per business unit, and each business unit has one or more storage domains. This scenario is likely with a private cloud that serves a single company. The configuration in this scenario is that the Root IDSYS defines the configuration of the identity management system and there are no IDSYS definitions for the tenants and storage domains. Therefore, the tenants and storage domains will inherit from the Root IDSYS using a single source of users and groups.
- Service Provider / Distributed Company
- In this scenario, a storage MSP, or a large company that has business units each with their own identity management systems, has multiple user/group sources. The configuration in this scenario is that the Root IDSYS defines the cluster administrator users and groups and the Tenant IDSYS documents define the users and groups for each customer or business unit. The storage domains do not define an IDSYS so that they inherit the definition from the tenant and share the users and groups with the other storage domains owned by their tenant.
- Service Provider with Resellers
- This is an extension of the previous scenario except that each tenant could be a reseller offering storage domains to separate, unrelated companies. In this case, each storage domain would define an IDSYS that would override the Tenant IDSYS allowing a different set of users and groups for each storage domain. This scenario is not mutually exclusive with the previous one: a hybrid of the two is possible where some domains override the IDSYS of their tenant, and others do not.
Qualifying User and Group Names
You may need to fully qualify your user and group principal names to ensure correct policy resolution. In access control policies and x-owner-meta
headers, a "fully qualified" principal has its tenant name or storage domain appended directly to its name:
user | non-qualified | Principal from the same IDSYS scope as the content |
---|---|---|
user@domain | fully qualified | Principal from a specific storage domain's IDSYS scope |
user+tenant | fully qualified | Principal from a specific tenant's IDSYS scope |
user@ | fully qualified | Principal from root scope only |
If a principal (user/group) authenticates from the same IDSYS as the resource they are accessing, their name in your policies may remain unqualified (no @domain
or +tenant
suffix on principal names).
If a principal authenticates from a different IDSYS from the one used by the resource, Gateway uses the default assignment of the x-owner-meta
header value to fully qualify the principal (such as user@domain
or user+tenant
). Applications can also assign object ownership across domains, where the IDSYS of the storage domain differs from that of the user from another domain. There is no limit on the number of cross-domain relationships you can have, but all must be within the same Swarm cluster.
Tokens — Use care with tokens, because they are bound to the IDSYS of the context both where and when they were created. That is, if you create a tenant-level token but that tenant doesn't have an IDSYS, then the token has to take the root scope, meaning that all requests using this token will authenticate using the root IDSYS (and likely fail, not finding the user there), even if a correct tenant-level IDSYS is added later. And if you create a tenant-level token with a tenant IDSYS, the token must ignore any domain-level IDSYS, current or future. If you want to have domain-level controls over tokens, either create a tenant-level IDSYS and use inherit at the domain-level, or create tokens at the domain-level.
Best practices
Fully qualify any token administrators that are defined in an IDSYS document. Because token administrator is a privileged permission, this practice avoids ambiguity if a storage domain inherits its IDSYS from the tenant or root scope.
Fully qualify user/group names in your policies if there is more than one IDSYS involved, to ensure there are no problems with policy resolution.