Skip to content

[VL] Fix shuffle spill triggered by evicting buffers during stop#6422

Merged
zhouyuan merged 1 commit intoapache:mainfrom
marin-ma:fix-writer-stop-spill
Jul 12, 2024
Merged

[VL] Fix shuffle spill triggered by evicting buffers during stop#6422
zhouyuan merged 1 commit intoapache:mainfrom
marin-ma:fix-writer-stop-spill

Conversation

@marin-ma
Copy link
Contributor

In VeloxHashBasedShuffleWriter::stop, evict partition buffers can allocating extra memory and triggers spill. In this case the target partition buffer should not get shrunk.

Can be reproduced by the added UT.


Java stack:

org.apache.gluten.exception.GlutenException: Error during calling Java code from native code: org.apache.gluten.exception.GlutenException: (shuffle) nativeEvict: evict failed - Invalid: Value buffer of fixed-width array is null.
	at org.apache.gluten.vectorized.ShuffleWriterJniWrapper.nativeEvict(Native Method)
	at org.apache.spark.shuffle.ColumnarShuffleWriter$$anon$1.spill(ColumnarShuffleWriter.scala:170)
	at org.apache.gluten.memory.memtarget.Spillers$AppendableSpillerList.spill(Spillers.java:86)
	at org.apache.gluten.memory.memtarget.Spillers$WithMinSpillSize.spill(Spillers.java:66)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets.spillTree(TreeMemoryTargets.java:80)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets.spillTree(TreeMemoryTargets.java:55)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets.spillTree(TreeMemoryTargets.java:73)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets.spillTree(TreeMemoryTargets.java:55)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets.spillTree(TreeMemoryTargets.java:73)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets.spillTree(TreeMemoryTargets.java:55)
	at org.apache.gluten.memory.memtarget.spark.TreeMemoryConsumer.spill(TreeMemoryConsumer.java:115)
	at org.apache.spark.memory.TaskMemoryManager.trySpillAndAcquire(TaskMemoryManager.java:227)
	at org.apache.spark.memory.TaskMemoryManager.acquireExecutionMemory(TaskMemoryManager.java:190)
	at org.apache.spark.memory.MemoryConsumer.acquireMemory(MemoryConsumer.java:137)
	at org.apache.gluten.memory.memtarget.spark.TreeMemoryConsumer.borrow(TreeMemoryConsumer.java:65)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets$Node.borrow0(TreeMemoryTargets.java:126)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets$Node.borrow(TreeMemoryTargets.java:118)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets$Node.borrow0(TreeMemoryTargets.java:126)
	at org.apache.gluten.memory.memtarget.TreeMemoryTargets$Node.borrow(TreeMemoryTargets.java:118)
	at org.apache.gluten.memory.memtarget.OverAcquire.borrow(OverAcquire.java:56)
	at org.apache.gluten.memory.memtarget.ThrowOnOomMemoryTarget.borrow(ThrowOnOomMemoryTarget.java:35)
	at org.apache.gluten.memory.listener.ManagedReservationListener.reserve(ManagedReservationListener.java:43)
	at org.apache.gluten.vectorized.ShuffleWriterJniWrapper.stop(Native Method)
	at org.apache.spark.shuffle.ColumnarShuffleWriter.internalWrite(ColumnarShuffleWriter.scala:189)
	at org.apache.spark.shuffle.ColumnarShuffleWriter.write(ColumnarShuffleWriter.scala:231)
	at org.apache.spark.shuffle.ShuffleWriteProcessor.write(ShuffleWriteProcessor.scala:59)
	at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:104)
	at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:54)
	at org.apache.spark.TaskContext.runTaskWithListeners(TaskContext.scala:166)
	at org.apache.spark.scheduler.Task.run(Task.scala:141)
	at org.apache.spark.executor.Executor$TaskRunner.$anonfun$run$4(Executor.scala:620)
	at org.apache.spark.util.SparkErrorUtils.tryWithSafeFinally(SparkErrorUtils.scala:64)
	at org.apache.spark.util.SparkErrorUtils.tryWithSafeFinally$(SparkErrorUtils.scala:61)
	at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:94)
	at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:623)
	at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136)
	at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635)
	at java.base/java.lang.Thread.run(Thread.java:833)

@github-actions
Copy link

Thanks for opening a pull request!

Could you open an issue for this pull request on Github Issues?

https://github.com/apache/incubator-gluten/issues

Then could you also rename commit message and pull request title in the following format?

[GLUTEN-${ISSUES_ID}][COMPONENT]feat/fix: ${detailed message}

See also:

@zhouyuan zhouyuan merged commit e95f32a into apache:main Jul 12, 2024
yma11 added a commit to yma11/gluten that referenced this pull request Jul 15, 2024
yma11 added a commit to yma11/gluten that referenced this pull request Jul 15, 2024
yma11 added a commit to yma11/gluten that referenced this pull request Jul 15, 2024
yma11 added a commit to yma11/gluten that referenced this pull request Jul 15, 2024
yma11 added a commit to yma11/gluten that referenced this pull request Jul 15, 2024
kecookier pushed a commit to kecookier/gluten that referenced this pull request Nov 19, 2024
… stop (apache#6422)

In VeloxHashBasedShuffleWriter::stop, evict partition buffers can allocating extra memory and triggers spill. In this case the target partition buffer should not get shrunk.

Added UT for this case.
@weiting-chen weiting-chen mentioned this pull request Nov 20, 2024
37 tasks
weiting-chen pushed a commit that referenced this pull request Nov 22, 2024
… stop (#6422) (#7991)

In VeloxHashBasedShuffleWriter::stop, evict partition buffers can allocating extra memory and triggers spill. In this case the target partition buffer should not get shrunk.

Added UT for this case.

Co-authored-by: Rong Ma <rong.ma@intel.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants