Apache Kafka - K8s

This is part of the Charmed Apache Kafka K8s Tutorial. Please refer to this page for more information and an overview of the content.

Integrate with client applications

As mentioned in the previous section of the Tutorial, the recommended way to create and manage users is by means of another charm: the Data Integrator Charm. This will let us encode users directly in the Juju model, and - as shown in the following - rotate user credentials rotations with and without application downtime using Relations.

Data Integrator Charm

The Data Integrator Charm is a bare-bones charm that can be used for central management of database users, providing support for different kinds of data platforms (e.g. MongoDB, MySQL, PostgreSQL, Apache Kafka, OpenSearch, etc) with a consistent, opinionated and robust user experience. To deploy the Data Integrator Charm we can use the command juju deploy we have learned above:

juju deploy data-integrator --channel stable --config topic-name=test-topic --config extra-user-roles=admin

Relate to Charmed Apache Kafka

Now that the Database Integrator Charm has been set up, we can relate it to Charmed Apache Kafka. This will automatically create a username, password, and database for the Database Integrator Charm. Relate the two applications with:

juju relate data-integrator kafka-k8s

Wait for juju status --watch 1s to show:

Model     Controller  Cloud/Region        Version  SLA          Timestamp
tutorial  microk8s    microk8s/localhost  3.1.5    unsupported  18:18:16+02:00

App              Version  Status  Scale  Charm            Channel  Rev  Address         Exposed  Message
data-integrator           active      1  data-integrator  stable    13  10.152.183.188  no
kafka-k8s                 active      3  kafka-k8s        3/edge    39  10.152.183.237  no
zookeeper-k8s             active      3  zookeeper-k8s    3/edge    33  10.152.183.134  no

Unit                Workload  Agent  Address     Ports  Message
data-integrator/0*  active    idle   10.1.36.87
kafka-k8s/0         active    idle   10.1.36.78
kafka-k8s/1         active    idle   10.1.36.80
kafka-k8s/2*        active    idle   10.1.36.79
zookeeper-k8s/0     active    idle   10.1.36.84
zookeeper-k8s/1*    active    idle   10.1.36.86
zookeeper-k8s/2     active    idle   10.1.36.85

To retrieve information such as the username, password, and topic:

juju run data-integrator/leader get-credentials 

This should output something like:

Running operation 5 with 1 task
  - task 6 on unit-data-integrator-0

Waiting for task 6...
kafka:
  endpoints: kafka-k8s-2.kafka-k8s-endpoints:9092,kafka-k8s-0.kafka-k8s-endpoints:9092,kafka-k8s-1.kafka-k8s-endpoints:9092
  password: S4IeRaYaiiq0tsM7m2UZuP2mSI573IGV
  tls: disabled
  topic: test-topic
  username: relation-6
  zookeeper-uris: zookeeper-k8s-0.zookeeper-k8s-endpoints:2181,zookeeper-k8s-1.zookeeper-k8s-endpoints:2181,zookeeper-k8s-2.zookeeper-k8s-endpoints:2181/kafka-k8s
ok: "True"

Save the value listed under endpoints, username and password. (Note: your hostnames, usernames, and passwords will likely be different.)

Produce/consume messages

We will now use the username and password to produce some messages to Apache Kafka. To do so, we will first deploy the Apache Kafka Test App (available here): a test charm that also bundles some Python scripts to push data to Apache Kafka, e.g.:

juju deploy kafka-test-app -n1 --channel edge

Once the charm is up and running, you can log into the container

juju ssh kafka-test-app/0 /bin/bash

and make sure that the Python virtual environment libraries are visible:

export PYTHONPATH="/var/lib/juju/agents/unit-kafka-test-app-0/charm/venv:/var/lib/juju/agents/unit-kafka-test-app-0/charm/lib"

Once this is set up, you can use the client.py script that exposes some functionality to produce and consume messages. You can explore the usage of the script:

python3 -m charms.kafka.v0.client --help

usage: client.py [-h] [-t TOPIC] [-u USERNAME] [-p PASSWORD] [-c CONSUMER_GROUP_PREFIX] [-s SERVERS] [-x SECURITY_PROTOCOL] [-n NUM_MESSAGES] [-r REPLICATION_FACTOR] [--num-partitions NUM_PARTITIONS]
                 [--producer] [--consumer] [--cafile-path CAFILE_PATH] [--certfile-path CERTFILE_PATH] [--keyfile-path KEYFILE_PATH] [--mongo-uri MONGO_URI] [--origin ORIGIN]

Handler for running a Apache Kafka client

options:
  -h, --help            show this help message and exit
  -t TOPIC, --topic TOPIC
                        Kafka topic provided by Kafka Charm
  -u USERNAME, --username USERNAME
                        Kafka username provided by Kafka Charm
  -p PASSWORD, --password PASSWORD
                        Kafka password provided by Kafka Charm
  -c CONSUMER_GROUP_PREFIX, --consumer-group-prefix CONSUMER_GROUP_PREFIX
                        Kafka consumer-group-prefix provided by Kafka Charm
  -s SERVERS, --servers SERVERS
                        comma delimited list of Kafka bootstrap-server strings
  -x SECURITY_PROTOCOL, --security-protocol SECURITY_PROTOCOL
                        security protocol used for authentication
  -n NUM_MESSAGES, --num-messages NUM_MESSAGES
                        number of messages to send from a producer
  -r REPLICATION_FACTOR, --replication-factor REPLICATION_FACTOR
                        replcation.factor for created topics
  --num-partitions NUM_PARTITIONS
                        partitions for created topics
  --producer
  --consumer
  --cafile-path CAFILE_PATH
  --certfile-path CERTFILE_PATH
  --keyfile-path KEYFILE_PATH
  --mongo-uri MONGO_URI
  --origin ORIGIN

Using this script, you can therefore start producing messages:

python3 -m charms.kafka.v0.client \
  -u relation-6 -p S4IeRaYaiiq0tsM7m2UZuP2mSI573IGV \
  -t test-topic \
  -s "kafka-k8s-2.kafka-k8s-endpoints:9092,kafka-k8s-0.kafka-k8s-endpoints:9092,kafka-k8s-1.kafka-k8s-endpoints:9092" \
  -n 10 --producer \
  -r 3 --num-partitions 1

and consume them

python3 -m charms.kafka.v0.client \
  -u relation-6 -p S4IeRaYaiiq0tsM7m2UZuP2mSI573IGV \
  -t test-topic \
  -s "kafka-k8s-2.kafka-k8s-endpoints:9092,kafka-k8s-0.kafka-k8s-endpoints:9092,kafka-k8s-1.kafka-k8s-endpoints:9092" \
  --consumer \
  -c "cg"

Charm client applications

Actually, the Data Integrator is only a very special client charm, that implements the kafka_client relation for exchanging data with Charmed Apache Kafka and user management via relations.

For example, the steps above for producing and consuming messages to Apache Kafka have also been implemented in the kafka-test-app charm (that also implements the kafka_client relation) providing a fully integrated charmed user experience, where producing/consuming messages can simply be achieved using relations.

Producing messages

To produce messages to Apache Kafka, we need to configure the kafka-test-app to act as a producer, publishing messages to a specific topic:

juju config kafka-test-app topic_name=test_kafka_app_topic role=producer num_messages=20

To start to produce messages to Apache Kafka, we JUST simply relate the Apache Kafka Test App with Kafka:

juju relate kafka-test-app kafka-k8s

Note This will both take care of creating a dedicated user (as much as done for the data-integrator) as well as start a producer process publishing messages to the test_kafka_app_topic topic.

After some time, the juju status output should show:

Model     Controller  Cloud/Region        Version  SLA          Timestamp
tutorial  microk8s    microk8s/localhost  3.1.5    unsupported  18:58:47+02:00

App              Version  Status  Scale  Charm            Channel  Rev  Address         Exposed  Message
...
kafka-test-app            active      1  kafka-test-app   edge       8  10.152.183.60   no       Topic test_kafka_app_topic enabled with process producer
...

Unit                Workload  Agent  Address     Ports  Message
...
kafka-test-app/0*   active    idle   10.1.36.88         Topic test_kafka_app_topic enabled with process producer
...

indicating that the process has started. To make sure that this is indeed the case, you can check the logs of the process:

juju exec --application kafka-test-app "tail /tmp/*.log"

To stop the process (although it is very likely that the process has already stopped given the low number of messages that were provided) and remove the user, you can just remove the relation

juju remove-relation kafka-test-app kafka-k8s

Consuming messages

Note that the kafka-test-app charm can also similarly be used to consume messages by changing its configuration to

juju config kafka-test-app topic_name=test_kafka_app_topic role=consumer consumer_group_prefix=cg

After configuring the Apache Kafka Test App, just relate it again with the Charmed Apache Kafka. This will again create a new user and start the consumer process.

What’s next?

In the next section, we will learn how to rotate and manage the passwords for the Apache Kafka users, both the admin one and the ones managed by the Data Integrator.