Github user steveloughran commented on a diff in the pull request:
https://github.com/apache/spark/pull/6935#discussion_r47018390
--- Diff:
core/src/main/scala/org/apache/spark/scheduler/EventLoggingListener.scala ---
@@ -230,6 +230,13 @@ private[spark] class EventLoggingListener(
}
}
fileSystem.rename(new Path(logPath + IN_PROGRESS), target)
+ // touch file to ensure modtime is current across those filesystems
where rename()
+ // does not set it, -and which support setTimes(); it's a no-op on
most object stores
+ try {
+ fileSystem.setTimes(target, System.currentTimeMillis(), -1)
+ } catch {
+ case e: Exception => logDebug(s"failed to set time of $target", e)
+ }
--- End diff --
It doesn't impact the new incomplete caching work but it does ensure that
all filesystems have the same outcome on the `stop()` call: the final file has
the modtime of when the log was closed. Without that, Posix and HDFS would be
different. Because there's other timestamp checking going on in the original
scan, it may be safest to leave it in.
---
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]
