[
https://issues.apache.org/jira/browse/FLINK-1139?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14220779#comment-14220779
]
ASF GitHub Bot commented on FLINK-1139:
---------------------------------------
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/incubator-flink/pull/173#discussion_r20709391
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/executiongraph/ExecutionJobVertex.java
---
@@ -282,6 +282,14 @@ private void subtaskInFinalState(int subtask) {
numSubtasksInFinalState++;
if (numSubtasksInFinalState == parallelism) {
+
--- End diff --
BLOCKER: This here introduces a critical bug. This function is not meant to
abort with an exception. It should be
```
try {
getJobVertex().finalizeOnMaster(getGraph().getUserClassLoader());
}
catch (Throwable t) {
getGraph().fail(t);
```
> HadoopOutputFormat fails for DOP > 1
> ------------------------------------
>
> Key: FLINK-1139
> URL: https://issues.apache.org/jira/browse/FLINK-1139
> Project: Flink
> Issue Type: Bug
> Components: Hadoop Compatibility
> Reporter: Fabian Hueske
> Assignee: Fabian Hueske
>
> HadoopOutputFormat fails for DOP > 1. The reason is that job finalization
> (removal of the temp directory) is done after the first task finishes.
> All other tasks will fail in that moment, since they cannot continue to write
> their data to the temp directory.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
