[
https://issues.apache.org/jira/browse/MAPREDUCE-7370?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17585574#comment-17585574
]
ASF GitHub Bot commented on MAPREDUCE-7370:
-------------------------------------------
cnauroth commented on code in PR #4248:
URL: https://github.com/apache/hadoop/pull/4248#discussion_r956291996
##########
hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapred/lib/MultipleOutputs.java:
##########
@@ -527,9 +558,41 @@ public void collect(Object key, Object value) throws
IOException {
* @throws java.io.IOException thrown if any of the MultipleOutput files
* could not be closed properly.
*/
- public void close() throws IOException {
+ public void close() throws IOException, InterruptedException {
+ int nThreads = conf.getInt(MRConfig.MULTIPLE_OUTPUTS_CLOSE_THREAD_COUNT,
+ MRConfig.DEFAULT_MULTIPLE_OUTPUTS_CLOSE_THREAD_COUNT);
+ AtomicBoolean encounteredException = new AtomicBoolean(false);
+ ThreadFactory threadFactory = new
ThreadFactoryBuilder().setNameFormat("MultipleOutputs-close")
+ .setUncaughtExceptionHandler(
Review Comment:
Yes, makes sense. Thanks!
> Parallelize MultipleOutputs#close call
> --------------------------------------
>
> Key: MAPREDUCE-7370
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-7370
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Affects Versions: 3.3.0
> Reporter: Prabhu Joseph
> Assignee: groot
> Priority: Major
> Labels: pull-request-available
> Time Spent: 2h 20m
> Remaining Estimate: 0h
>
> This call takes more time when there are lot of files to close and there is a
> high latency to close. Parallelize MultipleOutputs#close call to improve the
> speed.
> {code}
> public void close() throws IOException {
> for (RecordWriter writer : recordWriters.values()) {
> writer.close(null);
> }
> }
> {code}
> Idea is from [[email protected]]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
