[
https://issues.apache.org/jira/browse/BEAM-4783?focusedWorklogId=150797&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-150797
]
ASF GitHub Bot logged work on BEAM-4783:
----------------------------------------
Author: ASF GitHub Bot
Created on: 03/Oct/18 16:06
Start Date: 03/Oct/18 16:06
Worklog Time Spent: 10m
Work Description: iemejia commented on a change in pull request #6181:
[BEAM-4783] Add bundleSize for splitting BoundedSources.
URL: https://github.com/apache/beam/pull/6181#discussion_r221661790
##########
File path:
runners/spark/src/main/java/org/apache/beam/runners/spark/translation/GroupCombineFunctions.java
##########
@@ -68,6 +69,32 @@
TranslationUtils.functionToFlatMapFunction(WindowingHelpers.windowFunction()),
true);
}
+ /**
+ * An implementation of {@link
+ * org.apache.beam.runners.core.GroupByKeyViaGroupByKeyOnly.GroupByKeyOnly}
for the Spark runner.
+ * Used only when bundleSize is set in SparkPipelineOptions. Evaluating if
the default Partitioner
+ * causes a reshuffle of the data.
+ */
+ @Experimental
Review comment:
Can you please remove this annotation from here and put it in the
`SparkPipelineOptions.setBundleSize`. Internal implementation of runners do not
have so far a stable API so we have no issues here, it is the options that the
users can depend on and what I wanted to highlight is that this knob can
disappear if we find a better compromise or if we can make a good default
without it.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 150797)
Time Spent: 4h 10m (was: 4h)
> Spark SourceRDD Not Designed With Dynamic Allocation In Mind
> ------------------------------------------------------------
>
> Key: BEAM-4783
> URL: https://issues.apache.org/jira/browse/BEAM-4783
> Project: Beam
> Issue Type: Improvement
> Components: runner-spark
> Reporter: Kyle Winkelman
> Assignee: Kyle Winkelman
> Priority: Major
> Time Spent: 4h 10m
> Remaining Estimate: 0h
>
> When the spark-runner is used along with the configuration
> spark.dynamicAllocation.enabled=true the SourceRDD does not detect this. It
> then falls back to the value calculated in this description:
> // when running on YARN/SparkDeploy it's the result of max(totalCores,
> 2).
> // when running on Mesos it's 8.
> // when running local it's the total number of cores (local = 1,
> local[N] = N,
> // local[*] = estimation of the machine's cores).
> // ** the configuration "spark.default.parallelism" takes precedence
> over all of the above **
> So in most cases this default is quite small. This is an issue when using a
> very large input file as it will only get split in half.
> I believe that when Dynamic Allocation is enable the SourceRDD should use the
> DEFAULT_BUNDLE_SIZE and possibly expose a SparkPipelineOptions that allows
> you to change this DEFAULT_BUNDLE_SIZE.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
