Tutorial: Replicating Data Across Clusters

This guide describes how to start two Apache Kafka® clusters and then a Replicator process to replicate data between them. Note that for tutorial purposes, we are running both clusters on the same machine. In order to do that, we jump through a hoop or two to make sure each cluster has its own ports and data directories. You will not need to perform these changes on the ZooKeeper and Broker configuration if you are running in a normal environment where each cluster has its own servers.

../../_images/replicator-quickstart-configuration.png

Replicator Quick Start Configuration

Tip

About install prerequisites and command examples

  • These instructions assume you have a local installation of Confluent Platform, the Confluent CLI, and Java 1.8 or 1.11 (needed for Confluent Platform). If you are new to Confluent Platform, work through the Quick Start for Confluent Platform first, and then return to this tutorial.

  • The commands shown to start servers and Replicator assume you are running from your Confluent Platform home directory or have Confluent Platform in your CLASSPATH. The examples also 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. For example, this command:

    ./bin/zookeeper-server-start etc/kafka/zookeeper.properties
    

    Is meant to be the same as:

    $CONFLUENT_HOME/bin/zookeeper-server-start $CONFLUENT_HOME/etc/kafka/zookeeper.properties
    

Ports for Kafka brokers and Confluent Platform components

The examples in this tutorial define the following port configurations.

Origin Destination
Kafka Brokers 9082 9092
ZooKeeper 2171 2181
Connect Replicator worker   8083 (copies topics from origin -> destination)
Control Center   9021

Start the destination cluster

  1. First, start a ZooKeeper server. In this guide, we are assuming services will run on localhost.

    • Start ZooKeeper by running this command in its own terminal.

      ./bin/zookeeper-server-start etc/kafka/zookeeper.properties
      
  2. Next, start a Kafka broker that will serve as our single node Kafka cluster.

    • Start Kafka by running this command in its own terminal.

      ./bin/kafka-server-start etc/kafka/server.properties
      

For complete details on getting these services up and running see the quick start instructions for Confluent Platform.

Note

The destination cluster should be running the same (or higher) version of Confluent Platform as the source cluster. Replicator runs within a Connect cluster linked to the destination cluster, and reads messages from the source cluster. Therefore, Replicator will not be able to interpret the message format if the destination is running an older version.

Start the origin cluster

While you configured the destination cluster to run on default ports, you will need to run the origin cluster on a different port to avoid collisions. The Kafka broker in the origin cluster is configured on port 9082, ZooKeeper is configured on 2171, as shown in the ports mapping. Copy the configuration files to a temporary location and modify them so they do not conflict with the destination cluster.

  1. Create a “quickstart” directory under $CONFLUENT_HOME for the tutorial properties files; for example, my-examples/.

    Tip

    • $CONFLUENT_HOME represents <path-to-confluent>. On Linux systems, you can use this notation to set a shell environment variable.
    • The rest of the examples in this section assume you are running commands from $CONFLUENT_HOME and using my-examples/ to store your properties files. If you follow this model, you can use the copy-paste options on the code blocks.
  2. Copy the config files to my-examples (or a directory of your choice).

    cp etc/kafka/zookeeper.properties my-examples/zookeeper_origin.properties
    
    cp etc/kafka/server.properties my-examples/server_origin.properties
    
  3. Update the port numbers.

    sed -i '' -e "s/2181/2171/g" my-examples/zookeeper_origin.properties
    sed -i '' -e "s/9092/9082/g" my-examples/server_origin.properties
    sed -i '' -e "s/2181/2171/g" my-examples/server_origin.properties
    sed -i '' -e "s/#listeners/listeners/g" my-examples/server_origin.properties
    sed -i '' -e "s/8090/8091/g" my-examples/server_origin.properties
    sed -i '' -e "s/#confluent.metadata.server.listeners/confluent.metadata.server.listeners/g" my-examples/server_origin.properties
    
  4. Update the broker ID for origin.

    sed -i '' -e "s/broker.id=0/broker.id=1/g" my-examples/server_origin.properties
    

    Tip

    This is not entirely necessary, as the brokers are in two different clusters, but it is nice to have unique broker IDs if you want to manage them with Kafka from the command line.

  5. Update data directories.

    sed -i '' -e "s/zookeeper/zookeeper_origin/g" my-examples/zookeeper_origin.properties
    sed -i '' -e "s/kafka-logs/kafka-logs-origin/g" my-examples/server_origin.properties
    
  6. From here, you can start up the origin cluster.

    • Start ZooKeeper by running this command in its own terminal.

      ./bin/zookeeper-server-start my-examples/zookeeper_origin.properties
      
    • Start Kafka by running this command in its own terminal.

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

Create a topic

Bring up a new command window to run Kafka commands.

Create a topic named “test-topic” in the origin cluster with the following command:

kafka-topics --create --topic test-topic --replication-factor 1 --partitions 1 --bootstrap-server localhost:9082

Tip

Depending on your environment, you may have to use the .sh extension (for example, kafka-topics.sh) for Kafka commands.

You can verify that the topic was created as follows:

kafka-topics --list --bootstrap-server localhost:9082

Your output should look similar to this (the _confluent topics are internal topics):

__confluent.support.metrics
_confluent-command
test-topic

When we configure and run Replicator, this “test-topic” will get replicated to the destination cluster (on port 2181) with the exact configuration we defined above. For the sake of this example, the test topic was created with just one partition. Replicator will work with any number of topics and partitions.

Configure and run Replicator

Confluent Replicator can run as an executable or as a Connector in the Kafka Connect framework. For this quick start, start Replicator as an executable.

Create consumer, producer, and replicator configuration files

The Replicator executable script expects three configuration files:

  • Configuration for the origin cluster
  • Configuration for the destination cluster
  • Replicator configuration

Tip

You can place these config files in any directory you choose, but these steps assume a path of $CONFLUENT_HOME/my-examples/, so as not to conflict with the default properties files that ship with Confluent Platform. (The default configs for producer.properties and consumer.properties are in etc/kafka/, and replicator.properties is in etc/kafka-connect-replicator/.)

Create the following files in $CONFLUENT_HOME/my-examples/:

  1. Configure the origin cluster in a new file named consumer.properties. Edit the file and make sure it contains the addresses of brokers from the origin cluster. The default broker list will match the origin cluster you started earlier.

    # Origin cluster connection configuration
    bootstrap.servers=localhost:9082
    
  2. Configure the destination cluster in a new file named producer.properties. Edit the file and make sure it contains the addresses of brokers from the destination cluster. The default broker list will match the destination cluster you started earlier.

    # Destination cluster connection configuration
    bootstrap.servers=localhost:9092
    
  3. Define the Replicator configuration in a new file named replication.properties for the Connect worker. This quick start shows a configuration for topic.rename.format but any of the Confluent Replicator Configuration Properties that are not connection related can be supplied in this file.

    # Replication configuration
    topic.rename.format=${topic}.replica
    replication.factor=1
    config.storage.replication.factor=1
    offset.storage.replication.factor=1
    status.storage.replication.factor=1
    confluent.topic.replication.factor=1
    

Tip

  • If no port is defined in replication.properties, this worker runs on its default port 8083, which is the desired config for this deployment.
  • The replication factor properties (all set to 1) are used because these test clusters are small. The recommended minimum cluster size in production is 3 and this is the default for these properties.

Start the replicator

After you have created the necessary configuration files, start Replicator executable in its own terminal with the command below (assuming the properties files are in my-examples).

./bin/replicator --cluster.id replicator --consumer.config my-examples/consumer.properties --producer.config my-examples/producer.properties --replication.config my-examples/replication.properties --whitelist 'test-topic'

Some Replicator executable parameters can be passed on the command line:

  • --cluster.id - An identifier used to determine which Replicator cluster this executable should join. Multiple Replicator executable instances with the same cluster.id will work together.
  • --consumer.config - The path to the origin cluster configuration
  • --producer.config - The path to the destination cluster configuration
  • --replication.config - The path to a file containing any non connection specific configuration. Command line arguments will override these configurations.
  • --whitelist - A list of topics to replicate from origin to destination

For a full list of command line options see Command Line Parameters of Replicator Executable.

Look for success messages related to starting the source task and creating the replicated topic that indicate Replicator is up and running, and copying topics.

Verify topic replication across the clusters

When Replicator finishes initialization, it checks the origin cluster for topics that need to be replicated.

In this case, it finds test-topic and creates the corresponding topic in the destination cluster. You can verify this with the following command.

./bin/kafka-topics --describe --topic test-topic.replica --bootstrap-server localhost:9092

Note that you are checking the existence of test-topic.replica because test-topic was renamed when it was replicated to the destination cluster, according to your configuration.

Your output should look similar to this:

./bin/kafka-topics --describe --topic test-topic.replica --bootstrap-server localhost:9092
Topic: test-topic.replica    PartitionCount: 1       ReplicationFactor: 1    Configs: message.timestamp.type=CreateTime,segment.bytes=1073741824
      Topic: test-topic.replica      Partition: 0    Leader: 0       Replicas: 0     Isr: 0  Offline: 0

You can also list and describe the topics on the destination cluster. Replicated topics, like test-topic.replica will be listed.

./bin/kafka-topics --list --bootstrap-server localhost:9092

Tip

  • To list topics on the origin cluster, run kafka-topics --list against localhost:9082.
  • To view a description of the original topic, run kafka-topics --describe but look for test-topic and target localhost:9082.

At any time after you’ve created the topic in the origin cluster, you can begin sending data to it using a Kafka producer to write to test-topic in the origin cluster. You can then confirm that the data has been replicated by consuming from test-topic.replica in the destination cluster. For example, to send a sequence of numbers using Kafka’s console producer, run the following command in a new terminal window:

seq 10000 | ./bin/kafka-console-producer --topic test-topic --broker-list localhost:9082

You can confirm delivery in the destination cluster using the console consumer in its own terminal window:

./bin/kafka-console-consumer --from-beginning --topic test-topic.replica --bootstrap-server localhost:9092

If the numbers 1 to 10,000 appear in the consumer output, this indicates that you have successfully created multi-cluster replication.

Press Ctl-C to end the consumer readout and return to the command prompt.

Use Control Center to monitor replicators

You can use Control Center to monitor the replicators in your current deployment:

  1. Stop Replicator and brokers on both the origin and destination clusters, and then stop the ZooKeeper instances (in that order).

    Press Ctl-C in the each command window to stop the processes, but keep the windows open to make it easy to restart each one.

  2. Activate the monitoring extension for Replicator by doing the following, as fully described in Replicator Monitoring Extension.

    • Add the full path to replicator-rest-extension-<version>.jar to your CLASSPATH.
    • Add rest.extension.classes=io.confluent.connect.replicator.monitoring.ReplicatorMonitoringExtension to my-examples/replication.properties.
  3. Uncomment or add the following lines to the Kafka configuration files for both the destination and origin, etc/kafka/server.properties and my-examples/server_origin.properties, respectively. The configuration for confluent.metrics.reporter.bootstrap.servers must point to localhost on port 9092 in both files, so you may need to edit one or both of these port numbers. (Searching on confluent.metrics will take you to these lines in the files.)

    confluent.metrics.reporter.topic.replicas=1
    metric.reporters=io.confluent.metrics.reporter.ConfluentMetricsReporter
    confluent.metrics.reporter.bootstrap.servers=localhost:9092
    
    • The first line indicates to Control Center that your deployment is in development mode, using a replication factor of 1.
    • The other two lines enable metrics reporting on Control Center, and provide access to the Confluent internal topic that collects and stores the monitoring data.

    Tip

    • For this example, the metrics reporter must point to the cluster that Confluent Control Center bootstraps to, which is the destination cluster. If this is not set properly, metrics on source topics will not show up in Control Center. This is why etc/kafka/server.properties and my-examples/server_origin.properties must have the same configuration for confluent.metrics.reporter.bootstrap.servers=localhost:9092.
    • When adapting these steps to more complex, real-world environments, you may decide to use a different approach. For example, in a deployment with multiple instances of Control Center for source and destination, each monitoring its own respective cluster, confluent.metrics.reporter.bootstrap.servers should point to source or destination, as appropriate. To learn more, see the scenarios for Multi-cluster configuration with Control Center, Monitoring Replicator, and Multi-Region Clusters.
  4. Edit my-examples/producer.properties to add the monitoring interceptor for the producer:

    # Monitoring interceptor for producer
    interceptor.classes=io.confluent.monitoring.clients.interceptor.MonitoringProducerInterceptor
    
  5. Edit my-examples/consumer.properties to add the monitoring interceptor for the consumer:

    # Monitoring interceptor for consumer
    interceptor.classes=io.confluent.monitoring.clients.interceptor.MonitoringConsumerInterceptor
    
  6. Edit etc/confluent-control-center/control-center-dev.properties to add the following two lines that specify origin and destination bootstrap servers for Control Center, as is required for monitoring multiple clusters. (A convenient place to add these lines is near the top of the file under “Control Center Settings”, immediately after the line that specifies confluent.controlcenter.id.)

    # multi-cluster monitoring
    confluent.controlcenter.kafka.origin.bootstrap.servers=localhost:9082
    confluent.controlcenter.kafka.destination.bootstrap.servers=localhost:9092
    

    Tip

    Control Center requires the host and port of the Connect REST endpoint to know where to look for Replicator monitoring metrics. In the config file used for this example (control-center-dev.properties), this is configured for you on the default port, and so works out-of-the-box:

    # A comma separated list of Connect host names
    confluent.controlcenter.connect.cluster=http://localhost:8083
    

    The production-ready config file (control-center-production.properties) has the default commented out. If you use this file instead, have multiple Connectors, or want to configure Connect clusters differently, you must specify the Connect endpoint(s), either by uncommenting the default or specifying hosts for your own Connect clusters. To learn more, see Control Center Configuration Reference descriptions for confluent.controlcenter.connect.<connect-cluster-name>.cluster and confluent.controlcenter.connect.cluster (deprecated).

  7. Restart the ZooKeeper instances on the destination and origin clusters with the same commands used above, for example:

    ./bin/zookeeper-server-start etc/kafka/zookeeper.properties
    
    ./bin/zookeeper-server-start my-examples/zookeeper_origin.properties
    
  8. Restart the brokers on the destination and origin clusters with the same commands used above, for example:

    ./bin/kafka-server-start etc/kafka/server.properties
    
    ./bin/kafka-server-start my-examples/server_origin.properties
    
  9. Restart Replicator and the Connect worker with the same command as above. For example:

    ./bin/replicator --cluster.id replicator --consumer.config my-examples/consumer.properties --producer.config my-examples/producer.properties --replication.config my-examples/replication.properties --whitelist 'test-topic'
    
  10. Launch Control Center with the following command.

    ./bin/control-center-start etc/confluent-control-center/control-center-dev.properties
    

    If no port is defined in control-center-dev.properties, Control Center runs by default on port 9021, as described in Confluent Control Center. This is the desired config for this deployment.

  11. Open Control Center at http://localhost:9021/ in your web browser.

    The clusters are rendered on Control Center with auto-generated names, based on your configuration.

    ../../_images/c3-replicators-multi-cluster.png
  12. (Optional) On Control Center, edit the cluster names to suit your use case, as described in Origin and Destination clusters in “Replicators” in the Control Center User Guide.

  13. On Control Center, select the destination cluster, click Replicators on the navigation panel, and use Control Center to monitor replication performance and drill down on source and replicated topics.

    ../../_images/c3-replicators-all.png

    To see messages produced to both the original and replicated topic on Control Center, try out kafka-consumer-perf-test in its own command window to auto-generate test data to test-topic.

    kafka-producer-perf-test \
       --producer-props bootstrap.servers=localhost:9082 \
       --topic test-topic \
       --record-size 1000 \
       --throughput 1000 \
       --num-records 3600000
    

    The command provides status output on messages sent, as shown:

    4999 records sent, 999.8 records/sec (0.95 MB/sec), 1.1 ms avg latency, 240.0 ms max latency.
    5003 records sent, 1000.2 records/sec (0.95 MB/sec), 0.5 ms avg latency, 4.0 ms max latency.
    5003 records sent, 1000.2 records/sec (0.95 MB/sec), 0.6 ms avg latency, 5.0 ms max latency.
    5001 records sent, 1000.2 records/sec (0.95 MB/sec), 0.3 ms avg latency, 3.0 ms max latency.
    5001 records sent, 1000.0 records/sec (0.95 MB/sec), 0.3 ms avg latency, 4.0 ms max latency.
    5000 records sent, 1000.0 records/sec (0.95 MB/sec), 0.8 ms avg latency, 24.0 ms max latency.
    5001 records sent, 1000.2 records/sec (0.95 MB/sec), 0.6 ms avg latency, 3.0 ms max latency.
    ...
    

    Like before, you can consume these messages from the command line, using kafka-console-consumer to verify that the replica topic is receiving them:

    ./bin/kafka-console-consumer --from-beginning --topic test-topic.replica --bootstrap-server localhost:9092
    

    You can also verify this on Control Center. Navigate to test-topic on the origin cluster to view messages on the original topic, and to test-topic.replica on the destination to view messages on the replicated topic.

    ../../_images/c3-replicator-topic-drilldown-messages.png
  14. To learn more about monitoring Replicators in Control Center, see “Replicators” in Control Center User Guide.

  15. When you have completed your experiments with the tutorial, be sure to perform clean up as follows:

    • Stop any producers and consumers using Ctl-C in the each command window.
    • Use Ctl-C in each command window to stop each service in reverse order to which you started them (stop Control Center first, then Replicator, Kafka brokers, and finally ZooKeepers).

Troubleshooting

If you run into trouble at any point with getting things up and running or with Control Center monitoring of replicators, here are some things to check:

  • Make sure the configurations in all properties files are correct, and port numbers match origin and destination ports as described in Ports for Kafka brokers and Confluent Platform components.
  • For monitoring with Control Center, make sure that your configurations match the monitoring requirements per the steps in Use Control Center to monitor replicators. If you are using the production-ready Control Center configuration file instead of the “dev” version shown in this example, make sure you have specified the Connect endpoint per the tip in Use Control Center to monitor replicators.
  • Verify that the monitoring extension is installed per Replicator Monitoring Extension and is in your CLASSPATH, especially in the shells where you start the Kafka brokers. Check this by running echo $CLASSPATH in open command windows.
  • If you are using the systemctl command to start the monitoring service, make sure you follow the steps in systemctl Command Configurations. If you don’t configure the environment variables properly, Connect will fail to start.
  • To retry the tutorial:
    1. Use Ctl-C in each command window to stop each service in reverse order to which you started them (stop Control Center first, then Replicator, Kafka brokers, and finally ZooKeepers).
    2. Delete stale log and data files in /tmp that may conflict with a new run of the clusters and topics. For example, remove these files: /tmp/confluent/control-center/, /tmp/zookeeper, /tmp/zookeeper_origin, /tmp/kafka-logs, /tmp/kafka-logs-origin, /tmp/control-center-logs
    3. From here, you can start again with your current install of Confluent Platform or even reinstall Confluent Platform and try the entire process from scratch.