Github user seyfe commented on a diff in the pull request:
https://github.com/apache/spark/pull/16165#discussion_r91369539
--- Diff:
core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala ---
@@ -461,7 +461,7 @@ private[history] class FsHistoryProvider(conf:
SparkConf, clock: Clock)
appListener.appAttemptId,
appListener.startTime.getOrElse(-1L),
appListener.endTime.getOrElse(-1L),
- fileStatus.getModificationTime(),
+ clock.getTimeMillis(),
--- End diff --
Hi @vanzin, we can achieve this by simply doing `if (appCompleted)
fileStatus.getModificationTime else clock.getTimeMillis()`.
But I am worried about this scenario: If we assume that modtime will not
change when job is completed, the last updated time will move forward while job
is in progress and it will go back when job is completed.
But then I found [this
code](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/scheduler/EventLoggingListener.scala#L230)
which is executed when we complete the job. We call `fileSystem.setTimes`
after `fileSystem.rename`.
So seems like my assumption is not a valid case. Let me think about it
more/add more tests etc..
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
