Kafka-batchsource.md

Kafka Batch Source

Description

Kafka batch source. Emits the record from kafka. It will emit a record based on the schema and format you use, or if no schema or format is specified, the message payload will be emitted. The source will remember the offset it read last run and continue from that offset for the next run. The Kafka batch source supports providing additional kafka properties for the kafka consumer, reading from kerberos-enabled kafka and limiting the number of records read. This plugin uses kafka 2.6 java apis.

Use Case

This source is used whenever you want to read from Kafka. For example, you may want to read messages from Kafka and write them to a Table.

Properties

Use Connection Whether to use a connection, if a connection is used, the brokers do not need to be provided.

Connection Name of the connection to use, should have the macro function ${conn:(connection-name)} to provide.

referenceName: This will be used to uniquely identify this source for lineage, annotating metadata, etc.

kafkaBrokers: List of Kafka brokers specified in host1:port1,host2:port2 form. (Macro-enabled)

topic: The Kafka topic to read from. (Macro-enabled)

offsetDir: Optional directory path to track the latest offset we read from kafka. It is useful for incrementally processing data from Kafka across subsequent runs. (Macro-enabled)

partitions: List of topic partitions to read from. If not specified, all partitions will be read. (Macro-enabled)

initialPartitionOffsets: The initial offset for each topic partition. This offset will only be used for the first run of the pipeline. Any subsequent run will read from the latest offset from previous run. Offsets are inclusive. If an offset of 5 is used, the message at offset 5 will be read. (Macro-enabled)

schema: Output schema of the source. If you would like the output records to contain a field with the Kafka message key, the schema must include a field of type bytes or nullable bytes, and you must set the keyField property to that field's name. Similarly, if you would like the output records to contain a field with the timestamp of when the record was read, the schema must include a field of type long or nullable long, and you must set the timeField property to that field's name. Any field that is not the keyField, partitionField and keyField will be used in conjuction with the format to parse Kafka message payloads.

maxNumberRecords The maximum of messages the source will read from each topic partition. If the current topic partition does not have this number of messages, the source will read to the latest offset. Note that this is an estimation, the acutal number of messages the source read may be smaller than this number.

principal The kerberos principal used for the source when kerberos security is enabled for kafka.

keytabLocation The keytab location for the kerberos principal when kerberos security is enabled for kafka.

kafkaProperties Additional kafka consumer properties to set.

format: Optional format of the Kafka event message. Any format supported by CDAP is supported. For example, a value of 'csv' will attempt to parse Kafka payloads as comma-separated values. If no format is given, Kafka message payloads will be treated as bytes.

keyField: Optional name of the field containing the message key. If this is not set, no key field will be added to output records. If set, this field must be present in the schema property and must be bytes.

partitionField: Optional name of the field containing the partition the message was read from. If this is not set, no partition field will be added to output records. If set, this field must be present in the schema property and must be an int.

offsetField: Optional name of the field containing the partition offset the message was read from. If this is not set, no offset field will be added to output records. If set, this field must be present in the schema property and must be a long.

Example

This example reads from the 'purchases' topic of a Kafka instance running on brokers host1.example.com:9092 and host2.example.com:9092. The source will add a field named 'key' which will have the message key in it. It parses the Kafka messages using the 'csv' format with 'user', 'item', 'count', and 'price' as the message schema.

{
    "name": "Kafka",
    "type": "streamingsource",
    "properties": {
        "topics": "purchases",
        "brokers": "host1.example.com:9092,host2.example.com:9092",
        "format": "csv",
        "keyField": "key",
        "schema": "{
            \"type\":\"record\",
            \"name\":\"purchase\",
            \"fields\":[
                {\"name\":\"key\",\"type\":\"bytes\"},
                {\"name\":\"user\",\"type\":\"string\"},
                {\"name\":\"item\",\"type\":\"string\"},
                {\"name\":\"count\",\"type\":\"int\"},
                {\"name\":\"price\",\"type\":\"double\"}
            ]
        }"
    }
}

For each Kafka message read, it will output a record with the schema:

field name	type
key	bytes
user	string
item	string
count	int
price	double