README.adoc

Using kafkacat to export data from one Kafka cluster and load to another

Sometimes it’s useful to be able to copy the contents of a topic from one Kafka cluster to another. For example, a set of data to support a demo, or some data against which to test an application.

kafkacat is a commandline tool for interacting with Kafka, including producing and sending messages.

You can run all the steps in this example using Docker to simulate two single-node Kafka clusters, using kafkacat to dump data from one and import to the other. Using kafkacat to 'sync' a topic via a

Demo setup

Populate some test data in Cluster "A":

docker run --rm -i --network export-import-with-kafkacat_default confluentinc/cp-kafkacat \
  kafkacat -b kafka-a:29092 -t test_topic -P -K: <<EOF
1:{"hello":"world"}
2:{"foo":"bar"}
EOF

Check it’s there:

docker run --rm -it --network export-import-with-kafkacat_default confluentinc/cp-kafkacat \
  kafkacat -b kafka-a:29092 -t test_topic -C -f '\nKey (%K bytes): %k\t\nValue (%S bytes): %s\nTimestamp: %T\tPartition: %p\tOffset: %o\n--\n'

Should look like:

Key (1 bytes): 1
Value (17 bytes): {"hello":"world"}
Timestamp: 1542086750978        Partition: 0    Offset: 0
--

Key (1 bytes): 2
Value (13 bytes): {"foo":"bar"}
Timestamp: 1542086769945        Partition: 0    Offset: 1
--
% Reached end of topic test_topic [0] at offset 2

Export

docker run --rm -it --network export-import-with-kafkacat_default confluentinc/cp-kafkacat \
  kafkacat -b kafka-a:29092 -t test_topic -C -K: -e -q > export.json

Generated export file should look like:

$ cat export.json
1:{"hello":"world"}
2:{"foo":"bar"}

Import

Import to Cluster B the exported data from Cluster A:

$ docker run --rm -it --volume $PWD:/data --network export-import-with-kafkacat_default confluentinc/cp-kafkacat \
    kafkacat -b kafka-b:29092 -t test_topic -P -K: -l /data/export.json

(if you’re not using Docker, you’ll need to amend the path in -l accordingly)

Verify that the data is there in Cluster B:

docker run --rm -it --network export-import-with-kafkacat_default confluentinc/cp-kafkacat \
  kafkacat -b kafka-b:29092 -t test_topic -C -f '\nKey (%K bytes): %k\t\nValue (%S bytes): %s\nTimestamp: %T\tPartition: %p\tOffset: %o\n--\n'

Should look like:

Key (1 bytes): 1
Value (18 bytes): {"hello":"world"}
Timestamp: 1542087361395        Partition: 0    Offset: 0
--

Key (1 bytes): 2
Value (14 bytes): {"foo":"bar"}
Timestamp: 1542087361395        Partition: 0    Offset: 1
--
% Reached end of topic test_topic [0] at offset 2

Sync via netcat

Since kafkacat can stream and consume messages using pipes, we can pair it with netcat (nc) to stream messages between Kafka clusters. This is obviously a very inferior method of replicating data compared to MirrorMaker or Confluent’s Replicator tool!

• On the target machine:

  nc -k -l 42000 | kafkacat -b localhost:19092 -t test_topic -P -K:

  Where:

  • -k tells nc not to hang up when the connection is closed

  • -l defines the port on which to listen

  • localhost:19092 is the Kafka broker on the target cluster

• On the source machine:

  kafkacat -b localhost:9092 -t test_topic -C -K: -u -q | nc localhost 42000

  Where:

  • localhost 42000 is the host and port on which the other nc is listening

  • localhost:9092 is the Kafka broker on the source cluster