KafkaReader 64x64

    Short Description



    KafkaReader Attributes



    See also

    Short Description

    KafkaReader reads events (messages) from a Kafka cluster.

    Data source Input ports Output ports Each to all outputs Different to different outputs Transformation Transf. req. Java CTL Auto-propagated metadata

    Kafka cluster




    Port type Number Required Description Metadata



    Consumed Kafka events



    KafkaReader does not propagate metadata.

    KafkaReader has a metadata template on its output port available.

    Table 36. KafkaMetadata - Output port 0
    Field number Field name Data type Description




    Event topic




    Topic partition of the event




    Event key




    Event value




    Event timestamp



    map[string, byte]

    Optional event metadata headers




    Event offset (order) in its respective partition

    KafkaReader Attributes

    Attribute Req Description Possible values




    A Kafka connection to be used. See Kafka Connections.

    e.g. MyKafkaConnection


    A Kafka topic to consume events from. A single topic name or a semicolon-separated list of topic names can be used.

    Either this attribute or Topic pattern attribute must be specified. If both are specified, Topic pattern takes precedence.

    e.g. my-events | topic1;topic2

    Topic pattern

    A regular expression pattern to match Kafka topic names to be consumed from.

    Either this attribute or Topic attribute must be specified. If both are specified, Topic pattern takes precedence.

    e.g. my-topics\d+



    A consumer group this consumer belongs to. It is required for parition auto-assignment among multiple consumers.

    Corresponds to Kafka consumer property.

    e.g. my-group

    Key type

    Event key data type to use.

    Corresponds to Kafka key.deserializer consumer property.

    String (default) | Bytes

    Value type

    Event value data type to use. Corresponds to Kafka value.deserializer consumer property.

    String (default) | Bytes


    When enabled, the consumer’s offset will be periodically committed in the background.

    Corresponds to Kafka consumer property.

    true (default) | false

    Maximum time for new event

    The maximum waiting time for new Kafka events to appear.

    e.g. 500ms | 3s | 1m

    Requested number of events

    The maximum number of events to be consumed. The actual number of consumed events might be higher because the events are consumed in batches. The size of this batch can be changed by setting the Kafka max.poll.records consumer property.

    e.g. 1000

    Maximum time of reading

    The total maximum time of reading.

    e.g. 10s | 1m

    Output mapping

    Defines the mapping of results to a standard output port.


    Assigned topic partition

    Topic partition to consume events from. If not specified it will be assigned automatically.

    natural numbers

    Seek to offset

    A specific event offset to start consuming from.

    natural numbers

    Seek to boundary

    Set to start consuming from the first or last event offset.

    Seek to beginning | Seek to end

    Seek to timestamp

    Set to start consuming from a specific event offset based on specified timestamp.

    e.g. 2020-12-01 06:00:00

    Commit position after seek

    Set to commit position after seeking. Practically only affects behavior with auto-commit disabled.

    true (default) | false

    Auto-commit interval

    Time interval in which consumed offsets are committed when auto-commit is enabled.

    Corresponds to Kafka consumer property.

    e.g. 500ms | 3s | 1m

    Consumer configuration

    Additional Kafka consumer properties. See Apache Kafka documentation for details.



    KafkaReader allows you to subscribe to (read and process) events from Kafka using the Kafka Consumer API.

    Using the Group attribute, readers can be organized into consumer groups - each reader within the group reads from a unique partition and the group as a whole consumes all messages from the entire topic.

    Besides the configuration available through the component attributes, any other Kafka consumer property can be configured in the Consumer configuration attribute.

    Managing committed offsets
    1. Auto-commit - with auto-commit option enabled (default), the offset is committed periodically in the background (with configurable time interval). The drawback of this option is that it does not allow to rollback in case of further event processing failure.

    2. Manual commit - to commit processed events manually you need to disable the auto-commit option and design your job to use offset of the last successfully processed event (e.g. stored as a job parameter) in the Seek to offset attribute or use a dedicated

      KafkaCommit component. You can find more details in the

      KafkaCommit component section.


    Basic reading from a Kafka topic

    Reading from a Kafka topic in parallel

    Basic reading from a Kafka topic
    1. Create a new Kafka connection to an existing Kafka Cluster by following

    2. Place a new

      KafkaReader component, in its

      Connection attribute, select the previously created connection.

    3. Specify the topic name to consume from in the

      Topic attribute, e.g. example-topic .

    4. Specify an arbitrary consumer group ID in the

      Group attribute,e.g. example-group .

    5. When the job is executed, the events in the specified Topic will be consumed.

    Reading from a Kafka topic in parallel
    1. Have a Kafka topic with multiple, let’s say 3 partitions and some events produced in them.

    2. Add multiple KafkaReader components to your job, up to the number of partitions in the topic. Having more readers than the number of partitions has no effect, the additional readers will not read any records.

    3. Configure the reader components, set the

      Connection ,

      Topic and

      Group to be the same in each of the components.

    4. Connect the reader output edges.

    5. When the job is executed, the readers will get partition assigned automatically and events will get consumed evenly.