Package org.apache.storm.kafka.spout.subscription

Class RoundRobinManualPartitioner

java.lang.Object
org.apache.storm.kafka.spout.subscription.RoundRobinManualPartitioner
All Implemented Interfaces:
Serializable, ManualPartitioner
public class RoundRobinManualPartitioner extends Object implements ManualPartitioner
Assign partitions in a round robin fashion for all spouts, not just the ones that are alive. Because the parallelism of the spouts does not typically change while running this makes the assignments more stable in the face of crashing spouts.

Round Robin means that first spout of N spouts will get the first partition, and the N+1th partition... The second spout will get the second partition and N+2th partition etc.

See Also:

  • Constructor Details

    • RoundRobinManualPartitioner

      public RoundRobinManualPartitioner()

  • Method Details

    • getPartitionsForThisTask

      public Set<org.apache.kafka.common.TopicPartition> getPartitionsForThisTask(List<org.apache.kafka.common.TopicPartition> allPartitionsSorted, TopologyContext context)
      Description copied from interface: ManualPartitioner
      Filter the list of all partitions handled by this set of spouts to get only the partitions assigned to this task.
      Specified by:
      getPartitionsForThisTask in interface ManualPartitioner
      Parameters:
      allPartitionsSorted - all of the partitions that the set of spouts want to subscribe to in a strict ordering that is consistent across tasks
      context - the context of the topology
      Returns:
      the subset of the partitions that this spout task should handle.