Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/7770#discussion_r36055960
--- Diff: core/src/main/scala/org/apache/spark/ui/ToolTips.scala ---
@@ -62,6 +62,13 @@ private[spark] object ToolTips {
"""Time that the executor spent paused for Java garbage collection
while the task was
running."""
+ val PEAK_EXECUTION_MEMORY =
+ """Execution memory refers to the memory used by internal data
structures created during
+ shuffles, aggregations and joins when Tungsten is enabled. The
value of this accumulator
+ should be approximately the sum of the peak sizes across all such
data structures created
+ in this task. For SQL jobs, this only tracks all unsafe operators,
broadcast joins, and
+ external sort."""
--- End diff --
I like the idea of something like `spark.execution.memoryFraction` to
describe the fraction of memory that is reserved for internal use by Spark's
own execution operators and not by caching or user-code.
As part of this renaming, I'd also like to rename ShuffleMemoryManager and
consolidate it with Tungsten's new TaskMemoryManager and ExecutorMemoryManager
in order to clarify the fact that ShuffleMemoryManager controls memory
reservations for many different internal operators, many of which have nothing
to do with shuffle.
---
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]
