Hi, I've noticed something a bit surprising (to me at least) when a Kafka 8 producer writes to a Kafka 10 Cluster where the messages are subsequently processed by a Kafka Connect sink. The messages are Avro encoded (a suitable Avro key/value converter is specified via worker.properties properties), which makes it a little difficult to process but creating a String from the byte[] gives a reasonable idea of the message contents.
The general pattern seems to be: The toConnectData() method in the org.apache.kafka.connect.storage.Converter interface takes 2 parameters: the topic and a byte[]. On the first invocation, the byte[] only contains part of the message, specifically the id attribute - this causes the Avro decoder to fail with an EOFException (not surprisingly). This is followed by a second invocation of the toConnectData method and this time the byte[] contains an entire and parseable message - in this case the message id component is prefixed by the character 'H' (i.e. the first element in the byte[]). Is this expected behaviour? Is there any way to suppress the first invocation? Thanks, David
