Details
-
Improvement
-
Status: Triage Needed
-
P2
-
Resolution: Fixed
-
None
-
None
Description
At present, the `WriteRecords` support for the KafkaIO does not recognize the `partition` property defined on `ProducerRecord` instances consumed by the transform. This ticket would added support so that any explicit partitioning that was defined would be acknowledged accordingly while still respecting the default behavior if it was not explicitly included.
This can be easily identified within the `KafkaWriter` class used behind the scenes in the `WriteRecords` transform:
producer.send(
// The null property in the following constructor represents partition
new ProducerRecord<>(
topicName, null, timestampMillis, record.key(), record.value(), record.headers()),
new SendCallback());
Because of this limitation, in a scenario where a user may desire an explicitly defined partitioning strategy as opposed to round-robin, they would have to create their own custom DoFn that defines a KafkaProducer (preferably within a @StartBundle) similar to the following approach (in Kotlin):
private class ExampleProducerDoFn(...): DoFn<...>() { private lateinit var producer: KafkaProducer<...> @StartBundle fun startBundle(context: StartBundleContext) { val options = context.pipelineOptions.`as`(YourPipelineOptions::class.java) producer = getKafkaProducer(options) } @ProcessElement fun processElement(context: ProcessContext){ // Omitted for brevity // Produce the record to a specific topic at a specific partition producer.send(ProducerRecord( "your_topic_here", your_partition_here, context.element().kv.key, context.element().kv.value )) } }
Attachments
Issue Links
- links to
