Trident's Guranteeing Message processing issue

[churly lin]

Hi all:

     Does a trident topology support a method to inform the user that a batch of tuples has been 'fully processed"?

    My situation is:

    I read data from kafka then process the input data. And after the batch of input tuples has been fully processed, I write the processed kafka message offset into Zookeeper. So that I can read from kafka using this stored offset next time when I restart the topology. To realize this strategy, I need to be informed when a batch has been fully processed, am I right?.

    So, Does trident support a way to inform user if a batch of tuples has been fully processed or not?

    Thank you.