Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/4155#discussion_r23575889
--- Diff: core/src/main/scala/org/apache/spark/SparkHadoopWriter.scala ---
@@ -106,18 +107,25 @@ class SparkHadoopWriter(@transient jobConf: JobConf)
val taCtxt = getTaskContext()
val cmtr = getOutputCommitter()
if (cmtr.needsTaskCommit(taCtxt)) {
- try {
- cmtr.commitTask(taCtxt)
- logInfo (taID + ": Committed")
- } catch {
- case e: IOException => {
- logError("Error committing the output of task: " + taID.value, e)
- cmtr.abortTask(taCtxt)
- throw e
+ val outputCommitCoordinator = SparkEnv.get.outputCommitCoordinator
+ val conf = SparkEnv.get.conf
+ val canCommit: Boolean = outputCommitCoordinator.canCommit(jobID,
splitID, attemptID)
+ if (canCommit) {
--- End diff --
> Otherwise, we probably need some new state (e.g. "task gave up") that
doesn't count as an error.
We have the `TaskEndReason` mechanism for special-handling of certain kinds
of task failures, so maybe we can add a new reason and update the corresponding
[handlers in
TaskSetManager](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/scheduler/TaskSetManager.scala#L605)
and
[DAGScheduler](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala#L941).
We might still end up having to throw an exception, but we can throw a more
specific one that we can catch and transform into the right TaskEndReason.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
