Use Predefined RBAC Roles in Confluent Platform

Confluent Platform provides predefined roles to help implement granular permissions for specific resources and to simplify access control across the Confluent Platform. A predefined role is a Confluent-defined job function that is assigned a set of permissions required to perform specific actions or operations on Confluent resources. Each role is bound to a principal and Confluent resources. Users can have multiple roles assigned to them. You cannot use a predefined role to override denial-of-access (DENY) that is configured in an ACL.

When a role is assigned at the cluster-level (Kafka cluster, Schema Registry cluster, ksqlDB cluster, or Connect cluster) it means that users who are assigned this role have access to all resources in a cluster. For example, the ClusterAdmin of a Kafka cluster has access to Confluent Control Center alerts. There are corresponding resource types for each cluster type. For example, you can assign the ResourceOwner role to the resource types KsqlCluster:ksql-cluster or Cluster:kafka-cluster to provide a user all the ResourceOwner privileges for a ksqlDB or Kafka cluster.

When a role is assigned at the resource-level it means that users assigned this role only have access to specific resources as defined in the role binding.

The resource types for which you can assign RBAC roles and role bindings are:

• Kafka cluster

• Topic

• Consumer group

• TransactionalID

• Schema Registry cluster

• Schema Registry subject

• ksqlDB cluster

• Connect cluster

• Connector

• Kek (CSFLE key resource in Schema Registry)

Confluent Platform provides the following predefined roles:

Role Name	Level of Role Scope	View Role Bindings of Others	Manage Role Bindings	Monitor	Resource Read	Resource Write	Resource Manage
super.user	Cluster	Yes	Yes	Yes	Yes	Yes	Yes
SystemAdmin	Cluster	Yes	Yes	Yes	Yes	Yes	Yes
ClusterAdmin	Cluster	No	No	Yes	No	No	Yes
UserAdmin	Cluster	Yes	Yes	No	No	No	No
SecurityAdmin	Cluster	Yes	No	No	No	No	No
AuditAdmin	Cluster	No	No	No	No	No	Yes (1)
Operator	Cluster	No	No	Yes	No	No	Yes (2)
ResourceOwner	Resource	Yes	Yes	No	Yes	Yes	Yes
DeveloperRead	Resource	No	No	No	Yes	No	No
DeveloperWrite	Resource	No	No	No	No	Yes	No
DeveloperManage	Resource	No	No	No	No	No	Yes

Notes:

1. The AuditAdmin role provides sufficient access for creating and managing the audit log configuration.

2. For Operator Resource Manage, Operators can only pause, resume, and scale Connectors.

super.user

The purpose of super.user is to have a bootstrap user who can initially grant another user the SystemAdmin role. Technically speaking, super.user is not a predefined role. It is a server.properties attribute that defines a user who has full access to all resources within a Metadata Service (MDS) cluster. A super.user has no access to resources in other clusters (unless also configured as a super.user on other clusters).

Permissions granted by super.user apply only to the broker where the super.user attribute is specified. No authorization is enforced on users defined as super.user. It is strongly recommended that this role is assigned only to a limited number of users (for example, 1-2 users who are responsible for bootstrapping).

The super.user permission does not apply to other brokers, clusters, or Confluent Platform components such as Schema Registry, ksqlDB, and so forth. To be a true super user on these components the user must be listed in the super.users configuration and be a SystemAdmin. Using SystemAdmin in combination with super.users for component-specific tasks can lead to excessive privileges and is generally only required for initial setup or broad cluster-level operations.

SystemAdmin

Provides full access to all scoped resources in the cluster (ksqlDB cluster, Kafka cluster, or Schema Registry cluster).

For managing specific components or resources, SystemAdmin is unnecessary, only the relevant RBAC roles are necessary ClusterAdmin, ResourceOwner, KsqlAdmin` for ksqlDB, DeveloperManage for Schema Registry, and so forth. These more granular roles provide targeted permissions, adhering to the principle of least privilege, which enhances security and simplifies auditing.

It is strongly recommended that you restrict SystemAdmin to a limited number of users (one or two per cluster) who need full permission for initial setup or to address urgent issues when absolutely necessary in production instances. You may wish to assign this role more liberally in small test and development use cases, or when working in ksqlDB clusters that are primarily single tenant. Otherwise, it is recommended that you do not assign the SystemAdmin role.

ClusterAdmin

Sets up clusters (ksqlDB cluster, Kafka cluster, or Schema Registry cluster).

Responsible for setting up and managing Kafka clusters, brokers, networking, ksqlDB clusters, Connect clusters, and adding or removing nodes and performing upgrades. The ClusterAdmin typically creates topics and sets the properties of those topics, for example performance and capacity, but cannot read or write to topics, and has no access to data. For monitoring applications, it is recommended that this role is delegated to the operator who monitors your applications. Typically, the ClusterAdmin user does not possess knowledge about the content of the cluster data and he/she delegates the ownership responsibility of those resources to users assigned the ResourceOwner role. For example, after creating topics the ClusterAdmin can set ownership to a specific user familiar with the topic data.

UserAdmin

Manages role bindings for users and groups in all clusters managed by MDS.

Manages users and groups in a cluster, including the mapping of users and groups to roles. Has no access to any other resources. Typically, users with the UserAdmin role are tasked with setting up access to resources. Users granted this role should be extremely trustworthy because they can grant roles to themselves and others. You can monitor the actions of the UserAdmin using audit logs.

SecurityAdmin

Enables management of platform-wide security initiatives.

Sets up security-related features (for example, encryption, tracking of audit logs, and watching for abnormal behavior). Provides a dedicated set of users for the initial setup and ongoing management of security functions.

AuditAdmin

Users or groups assigned this role on the MDS cluster and every registered Kafka cluster can manage the audit log configuration using the Confluent Metadata API.

Operator

Provides operational management of clusters and scale applications as needed.

Monitors the health of applications and clusters, including monitoring uptime. This role cannot create applications, nor does it allow you to view or edit the content of the topics. However, you can view what topics and partitions exist.

ResourceOwner

Transfers the ownership of critical resources and to scale the ability to manage authorizations for those resources.

Owns the resource and has full access to it, including read, write, and list. ResourceOwner can grant permission to others who need access to resources. Owner cannot change some of the configurations, for example the number of partitions. Must own the resource to grant others access to it. Enables scaling of authorization for critical resources.

DeveloperRead, DeveloperWrite, DeveloperManage

Allows developers to drive the implementation of applications they are working on and manage the content.

RBAC role use cases

These use cases are based on a new project where security is managed using RBAC predefined roles as follows:

Predefined Role	Plan
super.user	Sam is granted full access to all project resources and operations. He will create the initial set of roles for the project.
ResourceOwner	Ryan will own all topics with the prefix finance_. He can grant others permission to access and use this resource. In this use case, he is the ResourceOwner for the finance topics.
UserAdmin	Uri will manage the users and groups for the project.
Operator	Olivia will be responsible for the operational and health management of the platform and applications.
ClusterAdmin	Cindy is a member of the Kafka cluster central team.
DeveloperRead, DeveloperWrite, DeveloperManage	David will be responsible for developing and managing the application.

The use cases below guide you through the application of predefined roles to illustrate one possible way you can leverage and use RBAC predefined roles.

Project started by super.user

During the initial configuration, Sam is granted super.user access through the configuration file. Now Sam has access to all the resources within the Kafka cluster that is hosting MDS, and also can create role bindings to any resource managed by MDS.

Roles: super.user

Invite (using corporate tools like JIRA) users to build Confluent Platform

Sam sets up role bindings for users and groups:

• Uri granted UserAdmin role to manage users/groups

• If needed, Sam can grant the SystemAdmin role for the cluster to others (only if absolutely necessary).

Roles: UserAdmin, SystemAdmin

UserAdmin defines and assigns roles

Uri takes the following actions:

• If needed, Uri creates new groups in corporate LDAP (for example, Kafka developer group) making it easier to manage role bindings for multiple users at the same time.

• Grants Ryan ResourceOwner role on topics that use the prefix s.

• Grants Olivia Operator role on a Connect cluster so she can manage operations.

• Grants Cindy the ClusterAdmin role. Now Cindy can manage clusters.

Roles: UserAdmin, ResourceOwner, ClusterAdmin, Operator

ClusterAdmin creates topics and other resources

Cindy joins the project to manage a Kafka cluster. She:

• Manages brokers and updates configurations as needed.

• Receives a request to create topics for the finance development team:

• She creates topics with the finance_ prefix.

• She can’t read or write to these topics.

Roles: ClusterAdmin

Data governance is implemented

• Ryan has ResourceOwner rights to all the topics prefixed with finance_, so he can view all these topics.

• Ryan grants appropriate access rights for all topics with the prefix finance_.

• Ryan is notified through the existing corporate workflow that there is a request for access to topics and other critical resources to which he can grant or deny access.

Roles: ResourceOwner

Developer creates applications

David needs to create a new application that produces to a topic (finance_report).

• If David has not already been added as a developer (through user/group mapping by Uri), then he must request access from Ryan.

• Ryan (ResourceOwner) grants David the DeveloperWrite role on the topic finance_report.

• Ryan (ResourceOwner) can grant David the DeveloperRead or DeveloperManage roles too, depending on what kind of operations David’s application needs to perform.

Roles: DeveloperWrite, DeveloperRead, DeveloperManage

Monitor health of the cluster and take proper actions

Olivia is responsible for managing the health of the resources in the cluster and takes actions when necessary. For example, if a connector has performance issues then she can increase the number of instances or scale them up.

Roles: Operator

RBAC role use cases for Confluent Platform for Apache Flink

Confluent Manager for Apache Flink supports all available RBAC authentication options to authenticate with the Confluent Platform metadata service. The following examples cover some authentication methods for the metadata service.

OAuth example

The following YAML file shows how you might configure mTLS authentication for CMF with Oauth authentication with MDS.

# authorization-values.yaml
 cmf:
   authorization:
     mdsRestConfig:
       endpoint: #Mandatory: replace with the endpoint for the Metadata Service
       authentication:
         type: oauth
         config:
           confluent.metadata.http.auth.credentials.provider: OAUTHBEARER
           confluent.metadata.oauthbearer.token.endpoint.url: #URL of your configured IDP used to store the users
           confluent.metadata.oauthbearer.login.client.id: #Client id for the user talking to MDS, needs to be a user that can request permissions for all users
           confluent.metadata.oauthbearer.login.client.secret: #Client secret for the user talking to MDS
           # Required when MDS uses HTTPS: truststore to trust the MDS server certificate.
           confluent.metadata.ssl.truststore.location: #truststore path to trust the MDS server certificate
           confluent.metadata.ssl.truststore.password: #optional password if the truststore is password protected

When you make the helm install call, use the -f flag to pass the YAML file with the security information like the following:

helm upgrade --install cmf confluent/confluent-manager-for-apache-flink \
  -f mtls-values.yaml \
  -f authorization-values.yaml

Note

When the MDS endpoint uses HTTPS, you must include confluent.metadata.ssl.truststore.location in the authentication configuration so that CMF can trust the MDS server certificate. You only need to include confluent.metadata.ssl.truststore.password if the truststore is password-protected. These settings are specific to the CMF-to-MDS connection and are separate from any server SSL configuration on CMF itself.

mTLS example

The following example shows how you configure mTLS authentication for CMF with mTLS authentication with MDS.

# authorization-values.yaml
cmf:
  authorization:
    mdsRestConfig:
      endpoint: #Mandatory: replace with the endpoint for the Metadata Service
      authentication:
        type: mtls
        config:
          confluent.metadata.ssl.truststore.location: #truststore path containing certificates needed to talk to metadata service
          confluent.metadata.ssl.truststore.password: #optional password if truststore is encrypted
          confluent.metadata.ssl.keystore.location: #keystore path containing keys/certificates needed to talk to metadata service
          confluent.metadata.ssl.keystore.password: #optional password if keystore is encrypted
          confluent.metadata.ssl.key.password:

When you make the helm install call, use the -f flag to pass the YAML file with the security information like the following:

helm upgrade --install cmf confluent/confluent-manager-for-apache-flink \
  -f mtls-values.yaml \
  -f authorization-values.yaml