Skip to content

[WIP][SPARK-35917][SHUFFLE][CORE][3.2] Disable push-based shuffle feature to prevent it from being used #33329

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

Closed
otterc wants to merge 3 commits into from
Closed

[WIP][SPARK-35917][SHUFFLE][CORE][3.2] Disable push-based shuffle feature to prevent it from being used #33329

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
0a00987
SPARK-35917 Disable push-based shuffle feature to prevent it from bei…
otterc Jun 28, 2021
acd4b88
Updated the config documentation for both the client and the server
otterc Jun 28, 2021
c6cb54e
Removed the TODOs
otterc Jul 14, 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
2 changes: 2 additions & 0 deletions common/network-common/src/main/java/org/apache/spark/network/util/TransportConf.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -381,6 +381,8 @@ public boolean useOldFetchProtocol() {
* 'org.apache.spark.network.shuffle.ExternalBlockHandler$NoOpMergedShuffleFileManager'.
* To turn on push-based shuffle at a cluster level, set the configuration to
* 'org.apache.spark.network.shuffle.RemoteBlockPushResolver'.
*
* Push-based shuffle is not yet supported.
*/
public String mergedShuffleFileManagerImpl() {
return conf.get("spark.shuffle.server.mergedShuffleFileManagerImpl",
Expand Down
3 changes: 3 additions & 0 deletions common/network-yarn/src/main/java/org/apache/spark/network/yarn/YarnShuffleService.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -220,6 +220,9 @@ protected void serviceInit(Configuration externalConf) throws Exception {
TransportConf transportConf = new TransportConf("shuffle", new HadoopConfigProvider(_conf));
MergedShuffleFileManager shuffleMergeManager = newMergedShuffleFileManagerInstance(
transportConf);
if (!(shuffleMergeManager instanceof ExternalBlockHandler.NoOpMergedShuffleFileManager)) {
throw new UnsupportedOperationException("Push-based shuffle is not yet supported.");
}
blockHandler = new ExternalBlockHandler(
transportConf, registeredExecutorFile, shuffleMergeManager);

Expand Down
2 changes: 1 addition & 1 deletion core/src/main/scala/org/apache/spark/internal/config/package.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2098,7 +2098,7 @@ package object config {
"conjunction with the server side flag spark.shuffle.server.mergedShuffleFileManagerImpl " +
"which needs to be set with the appropriate " +
"org.apache.spark.network.shuffle.MergedShuffleFileManager implementation for push-based " +
"shuffle to be enabled")
"shuffle to be enabled. Push-based shuffle is not yet supported.")
.version("3.1.0")
Copy link
Member

@dongjoon-hyun dongjoon-hyun Jul 13, 2021

Choose a reason for hiding this comment

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

This configuration is added at 3.1.0 and do we want to add comments like Push-based shuffle is not yet supported at 3.1.3?

Copy link
Contributor Author

@otterc otterc Jul 14, 2021 •
edited
Loading

Choose a reason for hiding this comment

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

In 3.1, multiple critical pieces of push-based shuffle were not there, for example SPARK-32922, SPARK-32921, and SPARK-32920. So, even if anyone even tried to turn it on, it would just not work.
However, in 3.2 we are just missing 2 correctness fixes which involve modifying push-based shuffle protocols: #33078 and #33034. Without these 2 push-based shuffle still works but this is what we want to prevent.

Given this I don't think it makes any difference if we add these comments to branches of version < 3.2. WDYT?

.booleanConf
.createWithDefault(false)
Expand Down
6 changes: 5 additions & 1 deletion core/src/main/scala/org/apache/spark/util/Utils.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2595,10 +2595,14 @@ private[spark] object Utils extends Logging {
* to run in YARN mode, with external shuffle service enabled
*/
def isPushBasedShuffleEnabled(conf: SparkConf): Boolean = {
conf.get(PUSH_BASED_SHUFFLE_ENABLED) &&
val isPushBasedShuffleEnabled = conf.get(PUSH_BASED_SHUFFLE_ENABLED) &&
(conf.get(IS_TESTING).getOrElse(false) ||
(conf.get(SHUFFLE_SERVICE_ENABLED) &&
conf.get(SparkLauncher.SPARK_MASTER, null) == "yarn"))
if (isPushBasedShuffleEnabled && !conf.get(IS_TESTING).getOrElse(false)) {
throw new UnsupportedOperationException("Push-based shuffle is not yet supported.")
}
isPushBasedShuffleEnabled
}

/**
Expand Down
2 changes: 2 additions & 0 deletions core/src/test/scala/org/apache/spark/shuffle/HostLocalShuffleReadingSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,7 @@ import org.scalatest.matchers.should.Matchers._

import org.apache.spark._
import org.apache.spark.internal.config._
import org.apache.spark.internal.config.Tests.IS_TESTING
import org.apache.spark.network.TransportContext
import org.apache.spark.network.netty.{NettyBlockTransferService, SparkTransportConf}
import org.apache.spark.network.server.TransportServer
Expand Down Expand Up @@ -136,6 +137,7 @@ class HostLocalShuffleReadingSuite extends SparkFunSuite with Matchers with Loca

test("Enable host local shuffle reading when push based shuffle is enabled") {
val conf = new SparkConf()
.set(IS_TESTING, true)
.set(SHUFFLE_SERVICE_ENABLED, true)
.set("spark.yarn.maxAttempts", "1")
.set(PUSH_BASED_SHUFFLE_ENABLED, true)
Expand Down
2 changes: 1 addition & 1 deletion core/src/test/scala/org/apache/spark/util/UtilsSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1448,7 +1448,7 @@ class UtilsSuite extends SparkFunSuite with ResetSystemProperties with Logging {
conf.set(SHUFFLE_SERVICE_ENABLED, true)
conf.set(SparkLauncher.SPARK_MASTER, "yarn")
conf.set("spark.yarn.maxAttempts", "1")
assert(Utils.isPushBasedShuffleEnabled(conf) === true)
assertThrows[UnsupportedOperationException](Utils.isPushBasedShuffleEnabled(conf))
conf.set("spark.yarn.maxAttempts", "2")
assert(Utils.isPushBasedShuffleEnabled(conf) === true)
}
Expand Down