Page History

...

• BatchSender should use ExecutorsService that ensures if a thread is dead a new thread will be started, eg the task is always running. DONE
• BatchSender is currently only a single thread. We should support concurrency, so you can have multiple output based on per correlation key. DONE
• BatchSender should support transactions, which you can do with .transacted(). But we should have unit test based on that if absent.
• BatchSender should have simpler logic for enqueing and locking if possible. DONE
• BatchProcessor should support persistent, so aggregated exchanges can be stored in a database using JPA. See how we do this with the Idempotent Consumer and/or Tracer. PLUGGABLE SPI
• If possible: AggregationStrategy should be invoked on the fly so you in the future can use it to determine if we are complete. DONE
• Batch timeout and completion predicate should work together so the completion is triggered by the first one. DONE
• Batch timeout should be per correlation key based, and not a single global timeout. DONE
• If possible: Add POJO based aggregation strategy so you can avoid using the low level Exchange APIs
• Add option what to do in case the correlation key cannot be evaluated, eg to ignore or throw an exception back to caller. DONE
• Add option to ignore to late exchanges if a correlation key has been closed. DONE
• Look all JIRA tickets at Apache what we have for the Aggregator
• Resequencer is also using the BatchProcessor so we should ensure it works as well and maybe it can benefit some of the new features as well.
• The persistence store should be able to be cleanup from time to time so it wont grow forever