Github user viirya commented on a diff in the pull request:
https://github.com/apache/spark/pull/16387#discussion_r96634742
--- Diff:
core/src/main/scala/org/apache/spark/util/collection/ExternalAppendOnlyMap.scala
---
@@ -192,12 +193,16 @@ class ExternalAppendOnlyMap[K, V, C](
* It will be called by TaskMemoryManager when there is not enough
memory for the task.
*/
override protected[this] def forceSpill(): Boolean = {
- assert(readingIterator != null)
- val isSpilled = readingIterator.spill()
- if (isSpilled) {
- currentMap = null
+ if (isReadingIterator) {
+ assert(readingIterator != null)
+ val isSpilled = readingIterator.spill()
+ if (isSpilled) {
+ currentMap = null
+ }
+ isSpilled
+ } else {
+ false
--- End diff --
@vanzin As @samkum tested, spilling the map could cause GC pressure. I
think that is because the map is frequently spilled.
Because the map will be checked and spilled during `insertAll`, I think it
is ok not to spill it in `forceSpill` in order to avoid frequent spilling.
What do you think?
---
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]
