Tutorial: Share Data Across Topics Using Cluster Linking for Confluent Platform

This tutorial provides a practical first look at how to use Cluster Linking for a topic data sharing use case.

What the tutorial covers

By the end of this tutorial, you will have configured two clusters and successfully used Cluster Linking to create a mirror topic and share topic data across the clusters. You will also learn how to stop mirroring to make the topic writable, and verify that the two topics have diverged.

../../_images/kafka-basics-multi-cluster.png

About prerequisites and command examples

  • These instructions assume you have a local installation of Confluent Platform 7.0.0 or later, the Confluent CLI, and Java 8, 11, or 17 (recommended). For details on Java requirements, see Java in System Requirements for Confluent Platform. If you are new to Confluent Platform, you may want to work through the Quick Start for Confluent Platform first, and then return to this tutorial.
  • The examples assume that your properties files are in the default locations on your Confluent Platform installation, except as otherwise noted. This should make it easier to copy/paste example commands directly into your terminal in most cases.
  • With Confluent Platform is installed, Confluent CLI commands themselves can be run from any directory (kafka-topics, kafka-console-producer, kafka-console-consumer), but for commands that access properties files in $CONFLUENT_HOME (kafka-server-start), the examples show running these from within that directory. A reference for these open source utilities is provided in Kafka Command-Line Interface (CLI) Tools. A reference for Confluent premium command line tools and utilities is provided in CLI Tools for Confluent Platform.
  • Confluent CLI commands can specify the bootstrap server at the beginning or end of the command: kafka-topics --list --bootstrap-server localhost:9092 is the same as kafka-topics --bootstrap-server localhost:9092 --list. In these tutorials, the target bootstrap server is specified at the end of commands.

The rest of the tutorial refers to $CONFLUENT_HOME to indicate your Confluent Platform install directory. Set this as an environment variable, for example:

export CONFLUENT_HOME=$HOME/confluent-7.6.1
PATH=$CONFLUENT_HOME/bin:$PATH

KRaft and ZooKeeper

Important

As of Confluent Platform 7.5, ZooKeeper is deprecated for new deployments. Confluent recommends KRaft mode for new deployments. To learn more about running Kafka in KRaft mode, see KRaft Overview for Confluent Platform and the KRaft steps in the Platform Quick Start.

This tutorial provides examples for both KRaft mode and ZooKeeper mode.

For KRaft, the examples show a combined mode configuration, where for each cluster the broker and controller run on the same server. Currently, combined mode is not intended for production use but is shown here to simplify the tutorial. If you want to run controllers and brokers on separate servers, use KRaft in isolated mode. To learn more, see KRaft Overview for Confluent Platform and Kraft mode under Configure Confluent Platform for production.

Ports and configuration mapping

The example deployment in this tutorial uses the following port and feature configurations, and assumes that services will run on localhost.

Source Destination
Kafka brokers 9092 (location of original demo topic) 9093 (location of demo mirror topic)
KRaft controllers 9094 9095
HTTP listeners 8090 8091
Cluster link   Cluster link is enabled on the destination

Configure Kafka brokers, controllers, and ZooKeeper files

Configure the following files to set up the source and destination clusters. You can copy and modify the existing properties files to use as a starting point.

Tip

  • Note that you must add confluent.http.server.listeners and configure it to be unique to each broker.
  • In KRaft mode, you must set confluent.cluster.link.enable=true on the destination, as shown in the examples for configuring the destination cluster.
  • Pre 7.0.0 Confluent Platform releases in ZooKeeper mode required adding confluent.cluster.link.enable=true to the destination server properties file to enable Cluster Linking. Starting with Confluent Platform 7.0.0, Cluster Linking is enabled by default, so this setting is no longer needed and not shown in the configs above.

Configure the source cluster

The source cluster configurations (ports, data directories, and so on) are largely based on the defaults for the template properties files, with only a few changes. The source cluster must be Confluent Platform 5.4 or later (Apache Kafka® 2.4).

  1. Change directories to the location where Confluent Platform is installed:

    cd $CONFLUENT_HOME
    
  2. Create a directory to be used for all of your example files:

    mkdir my-examples
    
  3. Copy etc/kafka/kraft/server.properties into the examples directory and rename it to match its purpose:

    cp etc/kafka/kraft/server.properties my-examples/server-src.properties
    
  4. Update the port numbers.

    sed -i '' -e "s/9093/9094/g" my-examples/server-src.properties
    
  5. Append the following line at the end of server-src.properties to set the HTTP listener on the source cluster to 8090:

    echo "confluent.http.server.listeners=http://localhost:8090" >> my-examples/server-src.properties
    

Configure the destination cluster

While you configured the source cluster to run on default ports, you must set up the destination cluster on different ports to avoid collisions, as summarized in ports mapping. Copy the configuration files to your examples directory and modify them as shown below.

The destination cluster must be Confluent Platform 7.0.0 or later.

  1. Change directories to the location where Confluent Platform is installed:

    cd $CONFLUENT_HOME
    
  2. Copy my-examples/server-src.properties (the file you just created) in the examples directory and rename the copy to match its purpose:

    cp my-examples/server-src.properties my-examples/server-dst.properties
    
  3. Update the port numbers.

    sed -i '' -e "s/9092/9093/g" my-examples/server-dst.properties
    
    sed -i '' -e "s/9094/9095/g" my-examples/server-dst.properties
    
    sed -i '' -e "s/8090/8091/g" my-examples/server-dst.properties
    
  4. Update data directories.

    sed -i '' -e "s/kraft-combined-logs/kraft-combined-logs-dst/g" my-examples/server-dst.properties
    
  5. Add the following line to server-dst.properties to set the replication factor to 1 for the cluster link topic metadata. (This configuration is set to simplify this demo. The recommended replication factor in production is 3, which is the default.)

    echo "confluent.cluster.link.metadata.topic.replication.factor=1" >> my-examples/server-dst.properties
    
  6. Enable Cluster Linking on the destination.

    sed -i '' -e "s/confluent.cluster.link.enable=false/confluent.cluster.link.enable=true/g" my-examples/server-dst.properties
    

Start the source cluster

Run the following commands and start the source cluster from $CONFLUENT_HOME.

  1. Generate a random-uuid using the kafka-storage tool:

    KAFKA_CLUSTER_ID="$(bin/kafka-storage random-uuid)"
    
  2. Format the log directories for this server:

    ./bin/kafka-storage format -t $KAFKA_CLUSTER_ID -c my-examples/server-src.properties
    

    Tip

    The kafka-storage command is run only once per broker/controller. You cannot use this command to update an existing cluster. If you make a mistake in configurations at this point, you must recreate the directories from scratch, and work through the steps again.

  3. Start the server for the source cluster by running this command in its own terminal.

    ./bin/kafka-server-start my-examples/server-src.properties
    

Start the destination cluster

Run the following commands and start the destination cluster from $CONFLUENT_HOME.

  1. Generate a random-uuid using the kafka-storage tool:

    KAFKA_CLUSTER_ID="$(bin/kafka-storage random-uuid)"
    
  2. Format the log directories for this server:

    ./bin/kafka-storage format -t $KAFKA_CLUSTER_ID -c my-examples/server-dst.properties
    
  3. Start the server for the source cluster by running this command in its own terminal.

    ./bin/kafka-server-start my-examples/server-dst.properties
    

Populate the source cluster

  1. Create a topic on the source cluster.

    kafka-topics --create --topic demo --bootstrap-server localhost:9092
    

    You should get confirmation that the topic was successfully created.

    Created topic demo.
    

    Tip

    You can get a list of existing topics as follows:

    kafka-topics --list --bootstrap-server localhost:9092
    

    And get detailed information on a topic with the --describe option:

    kafka-topics --describe --topic demo --bootstrap-server localhost:9092
    
  2. Run the producer in a new command window to send some messages to the demo topic on the source cluster, and fill it with data.

    kafka-console-producer --topic demo --bootstrap-server localhost:9092
    

    When the producer starts, you will get a prompt >.

    Type some messages at the prompt. Press return after each message. The producer window will look like this:

    >first
    >second
    >third
    

    Tip

    Do not copy-paste the above example into your producer, but rather type the messages and press return after each message.

  3. Consume from the topic on the source cluster.

    In a new terminal, run a consumer to consume messages from the demo topic.

    kafka-console-consumer --topic demo --from-beginning --bootstrap-server localhost:9092
    

    If the topic successfully consumes the messages, your output will be:

    first
    second
    third
    

Check the replica status on the destination

  1. Run the following command to monitor the replicas on the destination cluster.

    kafka-replica-status --topics demo --include-linked --bootstrap-server localhost:9093
    

    Your output should resemble the following example.

    Topic Partition Replica ClusterLink IsLeader IsObserver IsIsrEligible IsInIsr IsCaughtUp LastCaughtUpLagMs LastFetchLagMs LogStartOffset LogEndOffset
    demo  0         0       -           true     false      true          true    true       0                 0              0              3
    demo  0         0       demo-link   true     false      true          true    true       2                 2              0              3
    

    The output shows both the source and destination’s replicas as distinguished by the ClusterLink field, where - means local and the link name (demo-link in this case) indicates the replica(s) over the cluster link.

  2. Run the following command to populate values for “MirrorState”, “MirrorLastFetchTimeMs”, “MirrorLastFetchHighWatermark”:

    kafka-replica-status --topics demo --include-mirror --bootstrap-server localhost:9093
    

    Your output should resemble the following:

    Topic Partition Replica IsLeader IsObserver IsIsrEligible IsInIsr IsCaughtUp LastCaughtUpLagMs LastFetchLagMs LogStartOffset LogEndOffset LeaderEpoch MirrorState MirrorLastFetchTimeMs MirrorLastFetchHighWatermark
    demo  0         1       true     false      true          true    true       0                 0              0              3            0           ACTIVE      -1                    -1
    

Verify that the mirror topic is read-only

Attempt to produce messages to the mirror topic on the destination as a way of verifying that the mirror topic is read-only.

  1. Start a producer, targeting the mirror topic.

    kafka-console-producer --topic demo --bootstrap-server localhost:9093
    
  2. Type a message at the > prompt, and press return.

    You should get an error similar to the following example because the mirror topic is not writable.

    >hi
    2020-08-13 16:55:53,571] ERROR Error when sending message to topic demo with key: null, value: 2 bytes with error: (org.apache.kafka.clients.producer.internals.ErrorLoggingCallback)
    

Change the Source Topic Configuration

  1. Describe the mirror topic to verify the destination and the mirror topic configs.

    kafka-configs --describe --topic demo --bootstrap-server localhost:9093
    

    The output will show parameters that are set for strictly mirrored configs.

    Dynamic configs for topic demo are:
    compression.type=producer sensitive=false synonyms={DYNAMIC_TOPIC_CONFIG:compression.type=producer, DEFAULT_CONFIG:compression.type=producer}
    cleanup.policy=delete sensitive=false synonyms={DYNAMIC_TOPIC_CONFIG:cleanup.policy=delete, DEFAULT_CONFIG:log.cleanup.policy=delete}
    segment.bytes=1073741824 sensitive=false synonyms={DYNAMIC_TOPIC_CONFIG:segment.bytes=1073741824, STATIC_BROKER_CONFIG:log.segment.bytes=1073741824, DEFAULT_CONFIG:log.segment.bytes=1073741824}
    max.message.bytes=1048588 sensitive=false synonyms={DYNAMIC_TOPIC_CONFIG:max.message.bytes=1048588, DEFAULT_CONFIG:message.max.bytes=1048588}
    
  2. Modify the source set retention.ms.

    This is one of the configurations that will mirror only if it is explicitly set on the source. (Note that retention.ms was not included in the output for the previous step.)

    kafka-configs --alter --topic demo --add-config retention.ms=123456890 --bootstrap-server localhost:9092
    

    The confirmation message looks like this:

    Completed updating config for topic demo.
    
  3. Rerun the kafka-configs command to verify that retention.ms is now on the mirror topic.

    kafka-configs --describe --topic demo --bootstrap-server localhost:9093
    

    The dynamic configurations now include retention.ms.

    Dynamic configs for topic demo are:
    compression.type=producer sensitive=false synonyms={DYNAMIC_TOPIC_CONFIG:compression.type=producer, DEFAULT_CONFIG:compression.type=producer}
    cleanup.policy=delete sensitive=false synonyms={DYNAMIC_TOPIC_CONFIG:cleanup.policy=delete, DEFAULT_CONFIG:log.cleanup.policy=delete}
    segment.bytes=1073741824 sensitive=false synonyms={DYNAMIC_TOPIC_CONFIG:segment.bytes=1073741824, STATIC_BROKER_CONFIG:log.segment.bytes=1073741824, DEFAULT_CONFIG:log.segment.bytes=1073741824}
    retention.ms=123456890 sensitive=false synonyms={DYNAMIC_TOPIC_CONFIG:retention.ms=123456890}
    max.message.bytes=1048588 sensitive=false synonyms={DYNAMIC_TOPIC_CONFIG:max.message.bytes=1048588, DEFAULT_CONFIG:message.max.bytes=1048588}
    

    Tip

    It may take up to 5 seconds to propagate the change made in the previous step to set retention.ms on the source topic.

Change the source topic’s partitions

  1. Alter the number of partitions on the source topic.

    kafka-topics --alter --topic demo --partitions 8 --bootstrap-server localhost:9092
    
  2. Verify the change on the source topic.

    kafka-topics --describe --topic demo --bootstrap-server localhost:9092
    

    The output will be:

    Topic: demo       PartitionCount: 8       ReplicationFactor: 1    Configs: segment.bytes=1073741824,retention.ms=123456890
      Topic: demo     Partition: 0    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 1    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 2    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 3    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 4    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 5    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 6    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 7    Leader: 0       Replicas: 0     Isr: 0  Offline:
    
  3. Verify the change on the mirror topic.

    kafka-topics --describe --topic demo --bootstrap-server localhost:9093
    

    The output of this same command for the mirror topic should match the original topic exactly.

    Topic: demo       PartitionCount: 8       ReplicationFactor: 1    Configs: segment.bytes=1073741824,retention.ms=123456890
      Topic: demo     Partition: 0    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 1    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 2    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 3    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 4    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 5    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 6    Leader: 0       Replicas: 0     Isr: 0  Offline:
      Topic: demo     Partition: 7    Leader: 0       Replicas: 0     Isr: 0  Offline:
    

    Tip

    This change may a few minutes to propagate because the default metadata refresh is at 5 minutes. This can be improved by setting config metadata.max.age.ms when creating the cluster link. For example:

    ./bin/kafka-cluster-links --bootstrap-server localhost:9093 --create --link demo-link \
    --config bootstrap.servers=localhost:9092,metadata.max.age.ms=5000
    

List mirror topics

To list mirror topics run kafka-cluster-links --list with the --include-topics option.

kafka-cluster-links --list --link demo-link --include-topics --bootstrap-server localhost:9093

Your output should resemble the following.

Link name: 'demo-link', link ID: '123-some-link-id', remote cluster ID: '123-some-cluster-id', local cluster ID: '456-some-other-cluster-id', remote cluster available: 'true', topics: [demo]

Cut over the mirror topic to make it writable

  1. Stop the mirror topic for the demo topic on the destination.

    kafka-mirrors --promote --topics demo --bootstrap-server localhost:9093
    

    You will get the following confirmation.

    Calculating max offset and ms lag for mirror topics: [demo]
    Finished calculating max offset lag and max lag ms for mirror topics: [demo]
    Request for stopping topic inventory.arrivals's mirror was successfully scheduled.
    Use the describe command with the --pending-stopped-only option to monitor progress.
    
  2. Verify that the mirror topic has stopped mirroring.

    kafka-mirrors --describe --topics demo --pending-stopped-only --bootstrap-server localhost:9093
    

    You will get the following confirmation that the topic has entered the STOPPED state.

    Topic: demo  LinkName: demo-link  LinkId: 123-some-link-id MirrorTopic: demo  State: STOPPED StateTime: 2021-11-15 14:06:26
    

Produce to both topics to verify divergence

  1. Run a producer to send messages to the original topic on the source. (Or switch back to this command window if you still have this producer running.)

    kafka-console-producer --topic demo --bootstrap-server localhost:9092
    

    Type the message old at the prompt.

    > old
    
  2. Run a producer to send messages to the topic on the destination, which is no longer a mirror topic.

    kafka-console-producer --topic demo --bootstrap-server localhost:9093
    

    Type the message new at the prompt.

    > new
    
  3. Run consumers to read from both topics. Your output will show they have diverged.

    • Run a consumer to read messages from the original topic on the source. (Or simply view the output on this window, if you still have this consumer running)

      kafka-console-consumer --topic demo --from-beginning --bootstrap-server localhost:9092
      

      The output on the source topic now includes the old message.

      first
      second
      third
      old
      
    • Run a consumer to read messages from the topic on the destination. (Or simply view the output on this window, if you still have this consumer running.)

      kafka-console-consumer --topic demo --from-beginning --bootstrap-server localhost:9093
      

      The output on the destination topic includes the new message.

      first
      second
      third
      new
      

Teardown

Run shutdown and cleanup tasks.

  1. Stop Replicator with Ctl-C in its command window.

  2. Stop all of the other components with Ctl-C in their respective command windows, in reverse order in which you started them.

    • Stop consumers and producers.
    • For KRaft mode, stop the destination server first, then stop the source server.
    • For ZooKeeper mode, stop the Kafka brokers, then their respective ZooKeepers.
  3. Delete the log directories from /tmp.

    This will clear out the metadata from your system and enable you to configure and run new local deployments with no collisions with the legacy metadata.