[
https://issues.apache.org/jira/browse/BEAM-4783?focusedWorklogId=144402&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-144402
]
ASF GitHub Bot logged work on BEAM-4783:
----------------------------------------
Author: ASF GitHub Bot
Created on: 14/Sep/18 18:43
Start Date: 14/Sep/18 18:43
Worklog Time Spent: 10m
Work Description: kyle-winkelman commented on issue #6181: [BEAM-4783]
Add bundleSize for splitting BoundedSources.
URL: https://github.com/apache/beam/pull/6181#issuecomment-421449558
After looking further into
[BEAM-1815](https://jira.apache.org/jira/browse/BEAM-1815) my understanding is
that the double shuffle happened in streaming mode only because there is the
groupByKey in GroupCombineFunctions.groupByKeyOnly followed by updateStateByKey
in SparkGroupAlsoByWindowViaWindowSet.groupAlsoByWindow. This will not happen
in batch mode so I believe we should always use groupByKey() without the
hashpartitioner when in batch mode. That way we don't squash a large amount of
data into a few partitions.
If you agree I will change (in TransformTranslator):
```java
JavaRDD<WindowedValue<KV<K, Iterable<WindowedValue<V>>>>>
groupedByKey;
if
(context.getSerializableOptions().get().as(SparkPipelineOptions.class).getBundleSize()
> 0) {
groupedByKey =
GroupCombineFunctions.groupByKeyOnlyDefaultPartitioner(inRDD,
keyCoder, wvCoder);
} else {
groupedByKey = GroupCombineFunctions.groupByKeyOnly(inRDD,
keyCoder, wvCoder);
}
```
to:
```java
JavaRDD<WindowedValue<KV<K, Iterable<WindowedValue<V>>>>>
groupedByKey =
GroupCombineFunctions.groupByKeyOnlyDefaultPartitioner(inRDD,
keyCoder, wvCoder);
```
----------------------------------------------------------------
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: 144402)
Time Spent: 2h 50m (was: 2h 40m)
> 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
> Affects Versions: 2.5.0
> Reporter: Kyle Winkelman
> Assignee: Jean-Baptiste Onofré
> Priority: Major
> Labels: newbie
> Time Spent: 2h 50m
> 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)
