Github user vanzin commented on a diff in the pull request:
https://github.com/apache/spark/pull/2471#discussion_r18782362
--- Diff:
core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala ---
@@ -195,22 +241,68 @@ private[history] class FsHistoryProvider(conf:
SparkConf) extends ApplicationHis
}
}
- val newIterator = logInfos.iterator.buffered
- val oldIterator = applications.values.iterator.buffered
- while (newIterator.hasNext && oldIterator.hasNext) {
- if (newIterator.head.endTime > oldIterator.head.endTime) {
- addIfAbsent(newIterator.next)
- } else {
- addIfAbsent(oldIterator.next)
+ applications.synchronized {
--- End diff --
You shouldn't ever need to synchronize on `applications` because it's a
read-only variable. It's replaced atomically with a new list (which is why it's
volatile) when there are changes. But synchronizing on it doesn't achieve
anything.
Your explanation doesn't cover a whole lot of other sources of races when
two tasks are running concurrently and looking at the current status of the
file system (and potentially modifying it). So yeah, having these tasks run
single-threaded would be much more simple to reason about.
---
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
