Github user ConeyLiu commented on a diff in the pull request:
https://github.com/apache/spark/pull/19316#discussion_r140408246
--- Diff:
core/src/main/scala/org/apache/spark/storage/memory/MemoryStore.scala ---
@@ -387,11 +387,18 @@ private[spark] class MemoryStore(
// the block's actual memory usage has exceeded the unroll memory by a
small amount, so we
// perform one final call to attempt to allocate additional memory if
necessary.
if (keepUnrolling) {
- serializationStream.close()
- reserveAdditionalMemoryIfNecessary()
+ serializationStream.flush()
+ if (bbos.size > unrollMemoryUsedByThisBlock) {
+ val amountToRequest = bbos.size - unrollMemoryUsedByThisBlock
+ keepUnrolling = reserveUnrollMemoryForThisTask(blockId,
amountToRequest, memoryMode)
+ if (keepUnrolling) {
+ unrollMemoryUsedByThisBlock += amountToRequest
+ }
+ }
}
if (keepUnrolling) {
+ serializationStream.close()
--- End diff --
Here, we should close the `serializationStream` after we check it again.
Previous code we close it first, and then request the exceed memory. So there
is a potential problem that we can't request enought memory, while the
`serializationStream` is closeed.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
