Skip to content

This component is complete and can be used. The interfaces will be supported until the function is removed from the project via the deprecation process. There will be ongoing extensions to this function, but it will be done to ensure backward compatibility as far as possible. If there is a need to break backward compatibility, this will be discussed and reviewed in the community, with a documented timeline.

Kafka Open Metadata Topic Connector

Connector details

Apache Kafka provides a publish-subscribe service that allows a producer to publish events to subscribed consumers. Events are organized into topics. A producer publishes an event to a topic and the consumers register a listener to receive all events from a topic.

The Kafka Open Metadata Topic Connector implements an Open Metadata Topic Connector for a PLAINTEXT Apache Kafka topic. It supports both the event producer and event consumer interfaces.

Kafka Open Metadata Topic Connector

Egeria uses this connector as its default event notification technology (known as the Event Bus). You may also use it in your integration connectors is they need to send or receive events from a third party technology.


The connection example shows how to configure the connection for this connector. It is passing properties for both the producer and consumer.

Connection configuration

    "connection" :
        "class": "Connection",
        "qualifiedName": "Kafka Open Metadata Topic Connector",
            "class": "ConnectorType",
            "connectorProviderClassName": "org.odpi.openmetadata.adapters.eventbus.topic.kafka.KafkaOpenMetadataTopicProvider"      
            "class": "Endpoint",
            "address": {{KafkaTopicName}}
                "bootstrap.servers": {{kafkaEndpoint}}
            "": "{{consumerId}}",
                "bootstrap.servers": {{kafkaEndpoint}}
Add the name of the topic in {{topicName}}; a unique consumer identifier in {{consumerId}} and the endpoint for Apache Kafka (for example localhost:9092) in {{kafkaEndpoint}}.

Configuring Egeria

Egeria makes extensive use of events, and this connector is its default connector for sending and receiving events. In order to simplify the configuration of Egeria's OMAG Servers, it is possible to set up the default configuration properties for this connector in the Event Bus Configuration. This configuration is used when configuring the topics for the server's cohorts and Open Metadata Access Services (OMAS)

Default properties for the producer and consumer

The properties for the producer and consumer passed on the connection override the default set of properties coded into the connector implementation.


(see Apache Kafka producer configurations for more information and options)

Property Name Property Value
bootstrap.servers localhost:9092
acks all
retries 1
batch.size 16384 0
buffer.memory 33554432
max.request.size 10485760
key.serializer org.apache.kafka.common.serialization.StringSerializer
value.serializer org.apache.kafka.common.serialization.StringSerializer
bring.up.retries 10
bring.up.minSleepTime 5000


(see Apache Kafka consumer configurations for more information and options)

Property Name Property Value
bootstrap.servers localhost:9092 true 1000 30000
max.partition.fetch.bytes 10485760
key.deserializer org.apache.kafka.common.serialization.StringDeserializer
value.deserializer org.apache.kafka.common.serialization.StringDeserializer
bring.up.retries 10
bring.up.minSleepTime 5000

Controlling the direction of events

By default, this connector supports both the receiving and sending of events on a particular topic. It is possible to turn off, either the ability to listen for events through the consumer, or send events through the producer. This is achieved by setting the eventDirection configuration property.

The name of this property is defined in the OpenMetadataTopicProvider class.

    public static final String EVENT_DIRECTION_PROPERTY_NAME = "eventDirection"; // property name
    public static final String EVENT_DIRECTION_INOUT         = "inOut";          // default value
    public static final String EVENT_DIRECTION_OUT_ONLY      = "outOnly";        // do not start the consumer
    public static final String EVENT_DIRECTION_IN_ONLY       = "inOnly";         // do not start the producer
Below is a code fragment that uses these values:
        Map<String, Object>    configurationProperties = new HashMap<>();

        configurationProperties.put(OpenMetadataTopicProvider.EVENT_DIRECTION_PROPERTY_NAME, OpenMetadataTopicProvider.EVENT_DIRECTION_OUT_ONLY);
Using JSON, the property would be set as follows:
      "eventDirection" : "outOnly"

Do not use eventDirection for Egeria OMAG Server configuration

If you are configuring the event bus configuration in a configuration document using the REST API, do not set the eventDirection property because the configuration helper functions use the event bus configuration to set up all types of topics and will assign the appropriate values for eventDirection.


By default, kafka security is not configured. The exact configuration may depend on the specific kafka service being used. Service specific notes are below. They may work for other providers, and feedback is welcome so that this documentation can be updated accordingly.

Example: IBM Event Streams on IBM Cloud

There are 2 key pieces of information that are provided in the documentation for your configured cloud service

  • List of brokers - try to specify at least 3. The hosts in the examples below will need updating
  • API key - referred to as MYAPIKEY below

Given these, configure kafka properties for both provider and consumer as follows:

"broker.list: ",,,,,"
"sasl.jaas.config":" required username='token' password='MYAPIKEY';",
An example of a use of this configuration can be found in the virtual data connector helm charts. See odpi-egeria-vdc helm chart

Handling Kafka Cluster Bring Up Issues

It's strongly recommended that you ensure Kafka is started before starting any OMAG servers, for example in Kubernetes you may add an init container into a pod running Egeria's OMAG Server Platform to validate Kafka is ok, before the main container is launched. The Helm charts that are provided by the Egeria team typically take this approach. You should also ensure that you have a highly available Kafka deployment with multiple brokers.

However, in some environments users have encountered issues where the Kafka Cluster hasn't become fully available. To mitigate this, the Kafka Open Topic Connector provides a mechanism that verifies that the Kafka Cluster brokers are available before attempting to connect.

This mechanism is controlled by two producer/consumer properties.

  • bring.up.retries defaults to 10 and specifies the number of times the Egeria KafkaTopicConnector will retry verification before reporting a failure.
  • bring.up.minSleepTime is set to 5000ms by default and is the minimum amount of time to wait before attempting a verification retry. If a Kafka verification attempt takes longer than this value the KafkaTopicConnector does not pause before retrying the verification.

Different Kafka issues will result in differing overall times. For example, if the Kafka brokers are not resolvable (typical if waiting for a Kubernetes service to start), the Kafka library API calls will immediately fail, and so the connector will impose the minimum delay of 5s, each iteration will therefore take 5s by default. If instead the network address is unreachable, the Kafka client library will typically wait for 60s. As this is greater than 5s, no additional wait will be added.

If any capabilities of Egeria require Kafka and the topic connector cannot initialize, then the initialization of that capability will fail. For example attempting to start a server which is a member of a Cohort will fail if the Kafka Cluster is not accessible.

Topic Creation

Many enterprise Apache Kafka services do not allow automatic topic creation and so you need to ensure that any topics being used by Egeria are defined in Apache Kafka.

By default, Egeria generates its topic names using the following naming convention where BASE_TOPIC_NAME is the value used for topicURLRoot when configuring the egeria event bus. For example, the default value is egeria.

Cohort topics

For each cohort being used (such as cocoCohort):

  • BASE_TOPIC_NAME.omag.openmetadata.repositoryservices.cohort.COHORT_NAME.OMRSTopic
OMAS Topics

Open Metadata Access Services OMASs use topics to support event based interaction.

These need to be done FOR EACH SERVER configured to run one or more OMASs. (For example for Coco Pharmaceuticals this might include cocoMDS1, cocoMDS2, cocoMDS3 etc).

Then, FOR EACH OMAS configured (eg Asset Consumer OMAS, Data Platform OMAS, Governance Engine OMAS etc.) that supports topics, you will need to create the topic it is using.

This is the default naming convention:

  • BASE_TOPIC_NAME.omag.server.SERVER_NAME.omas.OMAS_NAME.InTopic
  • BASE_TOPIC_NAME.omag.server.SERVER_NAME.omas.OMAS_NAME.OutTopic

One way to create the list of kafka topics is to initially run against a kafka service which allows auto topic creation, then make note of the kafka topics that have been created - so that they can be replicated on the restricted setup.

In addition, review the Egeria Audit Log for any events beginning OCF-KAFKA-TOPIC-CONNECTOR so that action may be taken if for example topics are found to be missing.

Raise an issue or comment below