Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[SPARK-37675][CORE][SHUFFLE] Return PushMergedRemoteMetaFailedFetchResult if no available push-merged block #34934

Closed
wants to merge 3 commits into from
Closed

[SPARK-37675][CORE][SHUFFLE] Return PushMergedRemoteMetaFailedFetchResult if no available push-merged block #34934

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
d1e23d9
Fix calculate approxChunkSize when bitmaps isEmpty
pan3793 Dec 17, 2021
fd7d84b
Revert "Fix calculate approxChunkSize when bitmaps isEmpty"
pan3793 Dec 17, 2021
82da1c2
Return PushMergedRemoteMetaFailedFetchResult if no available push-mer…
pan3793 Dec 17, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 12 additions & 5 deletions core/src/main/scala/org/apache/spark/storage/PushBasedFetchHelper.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,7 +119,7 @@ private class PushBasedFetchHelper(
for (i <- bitmaps.indices) {
val blockChunkId = ShuffleBlockChunkId(shuffleId, shuffleMergeId, reduceId, i)
chunksMetaMap.put(blockChunkId, bitmaps(i))
logDebug(s"adding block chunk $blockChunkId of size $approxChunkSize")
logDebug(s"Adding block chunk $blockChunkId of size $approxChunkSize")
blocksToFetch += ((blockChunkId, approxChunkSize, SHUFFLE_PUSH_MAP_ID))
}
blocksToFetch
Expand All @@ -145,16 +145,23 @@ private class PushBasedFetchHelper(
logDebug(s"Received the meta of push-merged block for ($shuffleId, $shuffleMergeId," +
s" $reduceId) from ${req.address.host}:${req.address.port}")
try {
iterator.addToResultsQueue(PushMergedRemoteMetaFetchResult(shuffleId, shuffleMergeId,
reduceId, sizeMap((shuffleId, reduceId)), meta.readChunkBitmaps(), address))
val bitmaps = meta.readChunkBitmaps()
if (bitmaps.nonEmpty) {
iterator.addToResultsQueue(PushMergedRemoteMetaFetchResult(shuffleId, shuffleMergeId,
reduceId, sizeMap((shuffleId, reduceId)), bitmaps, address))
} else {
logInfo(s"No available push-merged block for ($shuffleId, $shuffleMergeId," +
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hmm...I thought we only raise the merged status request when there are merged blocks. Because we ensured this when fetching merge status:

spark/core/src/main/scala/org/apache/spark/MapOutputTracker.scala

Line 1537 in f1532a2

val remainingMapStatuses = if (mergeStatus != null && mergeStatus.totalSize > 0) {

But why bitmaps could be empty here? cc @mridulm @otterc

s" $reduceId) from ${req.address.host}:${req.address.port}")
iterator.addToResultsQueue(
PushMergedRemoteMetaFailedFetchResult(shuffleId, shuffleMergeId, reduceId, address))
}
} catch {
case exception: Exception =>
logError(s"Failed to parse the meta of push-merged block for ($shuffleId, " +
s"$shuffleMergeId, $reduceId) from" +
s" ${req.address.host}:${req.address.port}", exception)
iterator.addToResultsQueue(
PushMergedRemoteMetaFailedFetchResult(shuffleId, shuffleMergeId, reduceId,
address))
PushMergedRemoteMetaFailedFetchResult(shuffleId, shuffleMergeId, reduceId, address))
}
}

Expand Down
3 changes: 3 additions & 0 deletions core/src/main/scala/org/apache/spark/storage/ShuffleBlockFetcherIterator.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -972,6 +972,9 @@ final class ShuffleBlockFetcherIterator(

case PushMergedRemoteMetaFetchResult(
shuffleId, shuffleMergeId, reduceId, blockSize, bitmaps, address) =>
// PushMergedRemoteMetaFailedFetchResult will be returned instead of
// PushMergedRemoteMetaFetchResult if no available push-merged block.
assert(bitmaps.length > 0)
// The original meta request is processed so we decrease numBlocksToFetch and
// numBlocksInFlightPerAddress by 1. We will collect new shuffle chunks request and the
// count of this is added to numBlocksToFetch in collectFetchReqsFromMergedBlocks.
Expand Down