org.apache.flink.runtime.jobmanager.accumulators

Class AccumulatorManager

java.lang.Object

org.apache.flink.runtime.jobmanager.accumulators.AccumulatorManager

public class AccumulatorManager
extends Object

This class manages the accumulators for different jobs. Either the jobs are running and new accumulator results have to be merged in, or the jobs are no longer running and the results shall be still available for the client or the web interface. Accumulators for older jobs are automatically removed when new arrive, based on a maximum number of entries. All functions are thread-safe and thus can be called directly from JobManager.

Constructor Summary

Constructors

Constructor and Description
AccumulatorManager(int maxEntries)

Method Summary

Methods

Modifier and Type	Method and Description
Map<String,Accumulator<?,?>>	getJobAccumulators(JobID jobID) Returns all collected accumulators for the job.
void	processIncomingAccumulators(JobID jobID, Map<String,Accumulator<?,?>> newAccumulators) Merges the new accumulators with the existing accumulators collected for the job.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

AccumulatorManager

public AccumulatorManager(int maxEntries)

Method Detail

processIncomingAccumulators

public void processIncomingAccumulators(JobID jobID,
                               Map<String,Accumulator<?,?>> newAccumulators)

Merges the new accumulators with the existing accumulators collected for the job.

getJobAccumulators

public Map<String,Accumulator<?,?>> getJobAccumulators(JobID jobID)

Returns all collected accumulators for the job. For efficiency the internal accumulator is returned, so please use it read-only.