Hi Team, This is Kiran. Recently I started using Storm's Trident in one of my projects. Below are the issues i am encountering with, it will be BENEFICIAL & GREAT if you help me in resolving this.
Issue #1: (for Approach 1) 1. Partitioning the tuples on specified fields (Using partitionBy API) from the input stream 2. Batching the tuples from step 1 as Arraylist<TridentTuple> (Using Aggregator API) 3. Step 2 is taking 28 seconds of the ETA, as per my findings its taking that much of time to validate against the input stream that there are no more tuples to batch and emit. 4. However, i managed it to 100ms using TimerTask that invokes compete() API of the aggregator. But this time i am loosing tuples which are not getting partitioned as per the fields on multiple requests. Issue #2: (while choosing an alternative to approach 2) 1. Grouping the tuples on specified fields (Using groupBy API) from the input stream 2. Batching the tuples from step 1 as Arraylist<TridentTuple> (Using ReducerAggregator API) 3. Step 2 is taking 28 seconds of the ETA, as per my findings its taking that much of time to validate against the input stream that there are no more tuples to batch and emit. --- In this approach i am not loosing any tuples, but i am not even able to resolve the issue of 28 seconds ETA. Requesting you to do the needy ASAP. *:) happy Regards, Kiran Kumar Dasari.
