[
https://issues.apache.org/jira/browse/HADOOP-4000?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12672848#action_12672848
]
Billy Pearson commented on HADOOP-4000:
---------------------------------------
I see this also when using streaming. the reduce map disappears because over
100% task
> Reduce progress in the Reducer crosses 1
> ----------------------------------------
>
> Key: HADOOP-4000
> URL: https://issues.apache.org/jira/browse/HADOOP-4000
> Project: Hadoop Core
> Issue Type: Bug
> Components: mapred
> Reporter: Amar Kamat
>
> I printed the progress in the {{Progress.java}} for every progress made and I
> found this scenario
> {noformat}
> update called for reduce with progress 0.99998355
> update called for reduce with progress 0.99998355
> update called for reduce with progress 1.0000043
> update called for reduce with progress 1.0000043
> update called for reduce with progress 1.0000244
> {noformat}
> Following are the lines I added to get this
> {code:title=Progress.java|borderStyle=solid}
> /** Called during execution on a leaf node to set its progress. */
> public synchronized void set(float progress) {
> System.out.println("update called for " + status + " with progress " +
> progress);
> this.progress = progress;
> }
> {code}
> This needs investigation and will also explain why occasionally users see
> jobs that are complete and their progress less that 100%.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
