Based on my own experiments (can't find this documented anywhere), if 2 messages, having the same correlation ID and sequence number, aggregator will only take the 1st message and discard/ignore the other message.
Is there a way to make aggregator use the last message received instead?
The aggregation will merge the payload into 1 string.
Simple scenario: 3 messages with same correlation ID and sequence size of 2, ordered by time received
Current output: abcghi
Expected output: defghi
This scenario happens when sequence# 2 is missing which is meant for the 1st message. And the correlation ID (obtained from decoded payload) is very limited, hence it will be used multiple times.
Original message
The raw messages came in this format:
Sample raw message payload:
The aggregator basically combine the text
You need to use a custom release strategy (it can have the same logic as the default SequenceSizeReleaseStrategy, but it can't be that class). With the default strategy duplicate sequences are skipped.
However, you will also need a custom message group store to enact your desired behavior; otherwise the output will be abcdefghi.
However, the discarded message will not be sent to the discard channel in that case.
It's generally not a good idea to reuse correlation id; if you must, then use the group-timeout or a reaper to discard the partial group before any possibility of reuse of the correlation id.
BTW, you might find it easier to use a custom CorrelationStrategy and ReleaseStrategy rather than using the default and manipulating the headers.