Configure MDS to Manage Centralized Audit Logs

Starting in Confluent Platform 6.0, you can use Centralized audit logging to dynamically update an audit log configuration. Changes made through the Confluent CLI are pushed from the MDS (metadata service) out to all registered clusters, allowing for the centralized management of the audit log configuration, and assurance that all registered clusters publish their audit log events to the same destination Kafka cluster.

The MDS uses an admin client to connect to the destination Kafka cluster and inspect, create, and alter destination topics in response to certain API requests.

Before you can use centralized audit logging, you must configure one of your Kafka clusters to run the metadata service (MDS), which provides API endpoints to register a list of the Kafka clusters in your organization and to centrally manage the audit log configurations of those clusters. This audit log configuration API pushes out to all registered clusters the rules governing which events are captured and where they are sent. It also creates missing destination topics, and keeps the retention time policies of the destination topics in sync with the audit log configuration policy.

Until configured otherwise, the MDS operates on the assumption that it is a lone cluster, using its own internal admin client to configure destination topics on itself, and leaving the bootstrap servers unspecified so that audit log destination topics are on the same cluster. Because the default behavior requires less configuration, it is useful for the initial setup in a development environment.

In a production setting, you should have all of the Kafka clusters publish their audit logs to a single, central destination cluster. The configuration file for each managed cluster must include the destination cluster’s connection and credential information. However, it should disable auto-creation of destination topics, and leave confluent.security.event.router.config unspecified.

The following sections explain how to configure Kafka clusters and the MDS to manage centralized audit logs.

Prerequisites

  • Migrate legacy audit log configurations from all of your Kafka clusters into a combined JSON policy.

    Important

    You must satisfy this prerequisite prior to registering all of the Kafka clusters.

  • Register all of your Kafka clusters, including the MDS cluster, in the Cluster Registry.

    Note

    MDS cluster registration does not occur by default. You must explicitly register the MDS cluster in the cluster registry before registering other clusters.

  • Configure all of your registered clusters to use the same MDS for RBAC.

  • The MDS cluster uses the admin client to communicate with registered clusters (managed clusters). Ensure that the MDS admin client can connect to all of your registered clusters by having them expose an authentication token listener (for example, listener.name.external.sasl.enabled.mechanisms=OAUTHBEARER), and registering that listener’s port in the cluster registry. When using SASL_SSL, only use TLS keys that are verifiable by certificates in your client trust stores.

  • Configure a cluster to receive the audit logs. Set up an audit log writer user (with a name like auditlogwriter) on that cluster with the ability to write to the destination topics. For example, grant the DeveloperWrite role on the topic prefix confluent-audit-log-events. For details, refer to Configure the Audit Log Writer to the Destination Cluster.

  • Grant the AuditAdmin role on all your Kafka clusters to users or groups who will be managing the audit log configuration.

    Note

    The recommended way to grant permissions is for the audit log administrator to run any of the Confluent CLI confluent audit-log commands. The error message returns a list of recommended role bindings to grant to the user:

    confluent login --url "http://mds.example.com:8090" # authenticate as user "alice"
    confluent audit-log config describe
    Error: 403 Forbidden
    User:alice not permitted to DescribeConfigs on one or more clusters. Fix it:
    confluent iam rbac role-binding create --role AuditAdmin --principal User:alice --kafka-cluster-id DBS26_qTQ-mT23p5opUK_g
    confluent iam rbac role-binding create --role AuditAdmin --principal User:alice --kafka-cluster-id prz9a_-xqqlRgmekDoLw4U
    

Required Configuration Updates

To use centralized audit logging, you must:

Configure the MDS Cluster

You must make the following configuration updates for the MDS cluster:

Propagate the Audit Log Configuration to the Managed Clusters

MDS propagates audit log configuration updates to the managed clusters. For MDS to do this, every managed cluster must have its OAUTHBEARER endpoint registered with cluster registry (refer to Configure Listener Connections from the MDS Cluster).

Assuming the managed clusters’ listeners are configured for SASL_SSL, you must configure TLS client settings in the MDS cluster’s /etc/kafka/server.properties file. For example:

# Configuration used when propagating changes to managed clusters with SASL_SSL endpoints
ssl.confluent.metadata.server.truststore.location=<path-to-truststore.jks>
ssl.confluent.metadata.server.truststore.password=<trust-store-password>

If you haven’t already, you must grant the AuditAdmin role on all of your Kafka clusters to users or groups who will be managing the audit log configuration.

# Repeat for each Kafka cluster
confluent iam rbac role-binding create --role AuditAdmin --principal User:<audit-admin-user> --kafka-cluster-id <managed-kafka-cluster-id>
# Alternatively, repeat for each Kafka cluster, but specify the role binding for the group
confluent iam rbac role-binding create --role AuditAdmin --principal Group:<audit-admin-group> --kafka-cluster-id <managed-kafka-cluster-id>

Configure Audit Log Topic Management on the Destination Cluster

MDS manages the audit log topics on the destination cluster, creating missing topics, and keeping the retention time policies of those topics in sync with the audit log configuration policy. For MDS to do this, you must configure the admin client used by MDS to connect to the destination cluster.

Use the confluent.security.event.logger.destination.admin. prefix when configuring the admin client in the MDS cluster’s server.properties file. Other than the prefix requirement, this configuration is similar to other admin client configurations. This connection must be consistent with the producer configuration on this and all of the managed clusters. For details about the properties specified here, refer to AdminClient Configurations and Kafka AdminClient.

SASL_SSL Configuration

confluent.security.event.logger.destination.admin.bootstrap.servers=<logs1.example.com:9092,logs2.example.com:9092>
confluent.security.event.logger.destination.admin.security.protocol=SASL_SSL
confluent.security.event.logger.destination.admin.sasl.mechanism=PLAIN
confluent.security.event.logger.destination.admin.sasl.jaas.config=org.apache.kafka.common.security.plain.PlainLoginModule required \
      username="<audit-log-admin-username>" \
      password="<audit-log-admin-password>";
confluent.security.event.logger.destination.admin.ssl.truststore.location=<path-to-truststore.jks>
confluent.security.event.logger.destination.admin.ssl.truststore.password=<truststore-password>

Note

The recommended security protocol for the admin client configuration is SASL_SSL. If the destination cluster uses a different security protocol (for example, SASL_PLAINTEXT), then the value of confluent.security.event.logger.destination.admin.security.protocol should reflect that. In the case of SASL_PLAINTEXT, you would not need to specify the TLS/SSL trust store properties shown above.

The servers you specify in confluent.security.event.logger.destination.admin.bootstrap.servers will only be used as a fallback, until the audit log configuration JSON object stored in MDS specifies its destination bootstrap_servers, overriding this setting.

Configure the Audit Log Writer to the Destination Cluster

Because MDS also generates audit log events, you must configure the producer to connect to the audit log destination cluster.

# Configure the producer from the MDS cluster to export its own audit log events
confluent.security.event.logger.enable=true
confluent.security.event.logger.exporter.kafka.topic.create=false
confluent.security.event.logger.exporter.kafka.bootstrap.servers=<logs1.example.com:9092,logs2.example.com:9092>
confluent.security.event.logger.exporter.kafka.security.protocol=SASL_SSL
confluent.security.event.logger.exporter.kafka.sasl.mechanism=PLAIN
confluent.security.event.logger.exporter.kafka.sasl.jaas.config=org.apache.kafka.common.security.plain.PlainLoginModule required \
      username="<audit-log-writer-username>" \
      password="<audit-log-writer-password>";
confluent.security.event.logger.exporter.kafka.ssl.truststore.location=<path-to-truststore.jks>
confluent.security.event.logger.exporter.kafka.ssl.truststore.password=<truststore-password>

Note

The user specified in confluent.security.event.logger.exporter.kafka.sasl.jaas.config must be a principal who has the DeveloperWrite write role on all topics with the prefix confluent-audit-log-events on the destination cluster. However, it is not necessary for all of the registered clusters to use the same principal specified here. If required they can all use different principals, but you must grant all of those principals the appropriate permissions.

The centralized configuration feature is responsible for creating and updating the destination topics, so we recommend that you turn off the auto-create mechanism by setting confluent.security.event.logger.exporter.kafka.topic.create=false on all Kafka clusters (including the MDS cluster).

The servers you specify in confluent.security.event.logger.exporter.kafka.bootstrap.servers will only be used as a fallback, until the audit log configuration JSON object stored in MDS specifies its destination bootstrap_servers, overriding this setting.

Note

If you are migrating an existing Confluent Platform 5.4 or 5.5 audit log configuration to Confluent Platform 6.0 and intend to use centralized audit logging, but you still want to have access to the old audit logs, you should either: consume from both the original clusters and the destination cluster for the period of time you want access, or copy the old logs from the original topic in the cluster into the destination cluster topics.

Configure Each Registered Cluster

You must make the following configuration updates for each registered cluster:

Configure Listener Connections from the MDS Cluster

You must configure listeners from each managed registered cluster to use SASL_SSL (or SASL_PLAINTEXT) authentication protocols with OAUTHBEARER specified in sasl.enabled.mechanisms. For details about this configuration, refer to Configure Brokers.

Configure the Producer to Publish to the Audit Log Destination Cluster

You must configure the producer for each registered cluster to connect and publish to the audit log destination cluster for centralized audit logging. Specify the following SASL_SSL configuration in the server.properties file:

# Publish to the audit log destination cluster
confluent.security.event.logger.enable=true
confluent.security.event.logger.exporter.kafka.topic.create=false
confluent.security.event.logger.exporter.kafka.bootstrap.servers=<logs1.example.com:9092,logs2.example.com:9092>
confluent.security.event.logger.exporter.kafka.security.protocol=SASL_SSL
confluent.security.event.logger.exporter.kafka.sasl.mechanism=PLAIN
confluent.security.event.logger.exporter.kafka.sasl.jaas.config=org.apache.kafka.common.security.plain.PlainLoginModule required \
    username="<audit-log-writer-username>" \
    password="<audit-log-writer-password>";
confluent.security.event.logger.exporter.kafka.ssl.truststore.location=<path-to-truststore.jks>
confluent.security.event.logger.exporter.kafka.ssl.truststore.password=<truststore-password>

Configure the Audit Log Destination Cluster

You must make the following configuration updates for the audit log destination cluster:

Grant User Permissions for Audit Log Topics

To configure the audit log destination cluster for centralized audit logging you must grant the following user permissions:

  1. Specify a principal or principals who can write logs to the destination cluster. The principal should be the same one that appears in the confluent.security.event.logger.exporter.kafka.sasl.jaas.config property of the server.properties file of each Kafka cluster.

    confluent iam rbac role-binding create --principal User:<audit-log-writer> --role DeveloperWrite --resource Topic:confluent-audit-log-events --prefix --kafka-cluster-id <kafka-destination-cluster-id>
    
  2. Specify a principal who can administer audit log topics on the destination cluster. The principal should be the same one that appears in the confluent.security.event.logger.destination.admin.sasl.jaas.config property of the server.properties file of the MDS cluster.

    confluent iam rbac role-binding create --principal User:<audit-log-admin> --role ResourceOwner --resource Topic:confluent-audit-log-events --prefix --kafka-cluster-id <kafka-destination-cluster-id>
    

Configure Listener Connections from the MDS and Managed Registered Clusters

The destination server must have a listener that allows connections from all of the clients:

listeners=INTERNAL://:9092,BROKER://:9091,EXTERNAL://:9093
listener.security.protocol.map=INTERNAL:SASL_SSL,BROKER:SASL_SSL,EXTERNAL:SASL_SSL
...
listener.name.external.sasl.enabled.mechanisms=PLAIN
listener.name.external.plain.sasl.jaas.config=org.apache.kafka.common.security.plain.PlainLoginModule required \
  username="admin" password="admin-secret" \
  user_audit-log-admin-username="<audit-log-admin-password>" \
  user_audit-log-writer-username="<audit-log-writer-password>"
listener.name.external.ssl.key.password=<key-password>
listener.name.external.ssl.keystore.location=<path-to-keystore.jks>
listener.name.external.ssl.keystore.password=<key-store-password>
listener.name.external.ssl.truststore.location=<path-to-truststore.jks>
listener.name.external.ssl.truststore.password=<trust-store-password>

For details about this configuration, refer to Configure Brokers.

Configure the Default Properties for Topics on the Audit Log Destination Cluster

Any topics created on the destination cluster will use the default properties as specified in the destination cluster’s server.properties file (for example: num.partitions, min.insync.replicas, default.replication.factor).

Next Steps

At this point you have completed the required configuration updates, and can now use the Confluent Platform CLI confluent audit-log config commands or the Confluent MDS API to configure audit logs for all of the registered clusters. For additional command examples and guidance on troubleshooting, see Troubleshooting the MDS Configuration for Centralized Audit Logging.

Troubleshooting the MDS Configuration for Centralized Audit Logging

This section includes suggestions for troubleshooting the MDS cluster configuration used for managing centralized audit logs.

Verify the Audit Log Retention Setting

These procedures only affect your retention policy. It is recommended that you make minor changes only.

  1. Use the Confluent CLI to modify the audit log configuration and update the retention_ms of one or more destination topics:

    # Capture the current configuration from MDS
    confluent audit-log config describe > /tmp/audit-log-config.json
    
    # View what was captured
    cat /tmp/audit-log-config.json
    {
      "destinations": {
        "bootstrap_servers": [
          "logs1.example.com:9092",
          "logs2.example.com:9092"
        ],
        "topics": {
          "confluent-audit-log-events": {
            "retention_ms": 7776000000
          }
        }
      },
      "default_topics": {
        "allowed": "confluent-audit-log-events",
        "denied": "confluent-audit-log-events"
      }
    }
    
    # Make a small change
    vim /tmp/audit-log-config.json # e.g. - change 7776000000 to 7776000001
    
    # Post the change back to MDS
    confluent audit-log config update < /tmp/audit-log-config.json
    
  2. Verify that the topic’s retention.ms setting reflects the new value on the destination cluster:

    cat /tmp/destination-cluster-admin-client.properties
    
    bootstrap.servers=<logs1.example.com:9092>
    security.protocol=SASL_SSL
    sasl.jaas.config=org.apache.kafka.common.security.plain.PlainLoginModule required \
      username="<destination-cluster-admin-client-username>" \
      password="<destination-cluster-admin-client-password>";
    ssl.endpoint.identification.algorithm=https
    sasl.mechanism=PLAIN
    truststore.location=<path-to-truststore.jks>
    truststore.password=<trust-store-password>
    
    kafka-topics --bootstrap-server <logs1.example.com:9092> \
    --command-config /tmp/destination-cluster-admin-client.properties \
    --describe --topic <confluent-audit-log-events>
    
    Topic: confluent-audit-log-events  PartitionCount: 6 ReplicationFactor: 3
       Configs: min.insync.replicas=2,cleanup.policy=delete,retention.ms=7776000001
    Topic: confluent-audit-log-events Partition: 0  Leader: 2 Replicas: 2,1,0 Isr: 2,0,1
    Topic: confluent-audit-log-events Partition: 1  Leader: 1 Replicas: 1,0,2 Isr: 2,0,1
    Topic: confluent-audit-log-events Partition: 2  Leader: 0 Replicas: 0,2,1 Isr: 2,0,1
    Topic: confluent-audit-log-events Partition: 3  Leader: 2 Replicas: 2,0,1 Isr: 2,0,1
    Topic: confluent-audit-log-events Partition: 4  Leader: 1 Replicas: 1,2,0 Isr: 2,0,1
    Topic: confluent-audit-log-events Partition: 5  Leader: 0 Replicas: 0,1,2 Isr: 2,0,1
    
  3. Alter the retention.ms value of one of the destination topics directly on the destination cluster:

    kafka-topics --bootstrap-server <destination-cluster>:9092 \
    --command-config /tmp/destination-cluster-admin-client.properties \
    --alter --topic confluent-audit-log-events \
    --config retention.ms=7776000002
    
  4. Verify that the audit log configuration shows the new retention_ms setting:

    confluent audit-log config describe
    
    {
      "destinations": {
        "bootstrap_servers": [
          "logs1.example.com:9092",
          "logs2.example.com:9092"
        ],
        "topics": {
          "confluent-audit-log-events": {
            "retention_ms": 7776000002
          }
        }
      },
      "default_topics": {
        "allowed": "confluent-audit-log-events",
        "denied": "confluent-audit-log-events"
      }
    }
    

If this troubleshooting procedure doesn’t work (for example, if audit logging is not configured properly, you will get an error when you attempt to run the describe command), check to ensure that the connection and credentials in your MDS broker properties (the properties prefixed by confluent.security.event.logger.destination.admin.) are working. Also verify that you’ve granted sufficient permissions to the admin client principal on the destination cluster. Note that the minimum role binding should grant the ResourceOwner role on topics with the prefix confluent-audit-log-events on the destination cluster. Finally, confirm that the destination cluster is reachable and listening for connections from the MDS cluster’s network address.

Verify the Audit Log Configuration is Synchronized to Registered Clusters

Use the following command to verify that the audit log configuration is synchronized to registered clusters:

kafka-configs --bootstrap-server <mds-managed-cluster>:9092 \
--command-config /tmp/managed-cluster-admin-client.properties \
--entity-type brokers \
--entity-default \
--describe \
| grep confluent.security.event.router.config

You should see the same JSON audit log configuration you get when you run confluent audit-log config describe. It is possible that retention_ms values may differ if the audit topics have been altered directly on the destination cluster, in which case the metadata in the JSON may also be different. Everything else should be the same. If this verification fails check the following for the MDS cluster registry:

  • The clusters expose an auth token listener (listener.name.<example>.sasl.enabled.mechanism=OAUTHBEARER)
  • The clusters’ TLS keys are verifiable by certificates in the MDS server’s trust store.

Also look for error status messages when making an audit log API update request to MDS.