Skip to content

[SPARK-36705][SHUFFLE] Disable push based shuffle when IO encryption is enabled or serializer is not relocatable #33976

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
rmcyang wants to merge 3 commits into from
Closed

[SPARK-36705][SHUFFLE] Disable push based shuffle when IO encryption is enabled or serializer is not relocatable #33976

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
aace904
[SPARK-36705] Disable push based shuffle when IO encryption is enable…
Sep 11, 2021
9f4e4f4
[SPARK-36705] Disable push based shuffle when IO encryption is enable…
Sep 11, 2021
0e5e229
[SPARK-36705] Disable push based shuffle when IO encryption is enable…
Sep 11, 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
30 changes: 23 additions & 7 deletions 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 @@ -70,7 +70,7 @@ import org.apache.spark.internal.config.UI._
import org.apache.spark.internal.config.Worker._
import org.apache.spark.launcher.SparkLauncher
import org.apache.spark.network.util.JavaUtils
import org.apache.spark.serializer.{DeserializationStream, SerializationStream, SerializerInstance}
import org.apache.spark.serializer.{DeserializationStream, SerializationStream, Serializer, SerializerInstance}
import org.apache.spark.status.api.v1.{StackTrace, ThreadStackTrace}
import org.apache.spark.util.io.ChunkedByteBufferOutputStream

Expand Down Expand Up @@ -2597,14 +2597,30 @@ private[spark] object Utils extends Logging {
}

/**
* Push based shuffle can only be enabled when the application is submitted
* to run in YARN mode, with external shuffle service enabled
* Push based shuffle can only be enabled when below conditions are met:
* - the application is submitted to run in YARN mode
* - external shuffle service enabled
* - IO encryption disabled
* - serializer(such as KryoSerializer) supports relocation of serialized objects
Copy link
Member

@dongjoon-hyun dongjoon-hyun Sep 14, 2021 •
edited
Loading

Choose a reason for hiding this comment

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

Just a question. Do we have another available relocation-supporting serializer except KryoSerializer?

Copy link
Contributor Author

@rmcyang rmcyang Sep 14, 2021

Choose a reason for hiding this comment

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

No. According to Tuning Spark - Data Serialization, KryoSerializer and JavaSerializer are the only available options for data serialization, KryoSerializer is relocation-supporting while JavaSerializer is not.
UnsafeRowSerializer is relocation-supporting, but it seems only used for serializing UnsafeRows during shuffle.

Copy link
Member

@dongjoon-hyun dongjoon-hyun Sep 14, 2021

Choose a reason for hiding this comment

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

Ya, I thought like that. Thanks.

*/
def isPushBasedShuffleEnabled(conf: SparkConf): Boolean = {
conf.get(PUSH_BASED_SHUFFLE_ENABLED) &&
(conf.get(IS_TESTING).getOrElse(false) ||
(conf.get(SHUFFLE_SERVICE_ENABLED) &&
conf.get(SparkLauncher.SPARK_MASTER, null) == "yarn"))
val serializer = Utils.classForName(conf.get(SERIALIZER)).getConstructor(classOf[SparkConf])
.newInstance(conf).asInstanceOf[Serializer]
val canDoPushBasedShuffle =
conf.get(PUSH_BASED_SHUFFLE_ENABLED) &&
(conf.get(IS_TESTING).getOrElse(false) ||
(conf.get(SHUFFLE_SERVICE_ENABLED) &&
conf.get(SparkLauncher.SPARK_MASTER, null) == "yarn" &&
// TODO: [SPARK-36744] needs to support IO encryption for push-based shuffle
!conf.get(IO_ENCRYPTION_ENABLED) &&
serializer.supportsRelocationOfSerializedObjects))

if (!canDoPushBasedShuffle) {
logWarning("Push-based shuffle can only be enabled when the application is submitted" +
Copy link
Member

@dongjoon-hyun dongjoon-hyun Sep 14, 2021

Choose a reason for hiding this comment

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

nit. We need a space at the end of the string.

Copy link
Contributor Author

@rmcyang rmcyang Sep 14, 2021

Choose a reason for hiding this comment

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

This has been fixed in the follow-up PR #33984.

"to run in YARN mode, with external shuffle service enabled, IO encryption disabled, and" +
Copy link
Member

@dongjoon-hyun dongjoon-hyun Sep 14, 2021

Choose a reason for hiding this comment

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

ditto. We need another space at the end of this line, too.

Copy link
Contributor Author

@rmcyang rmcyang Sep 14, 2021

Choose a reason for hiding this comment

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

Fixed in the follow-up PR #33984.

Copy link
Member

@dongjoon-hyun dongjoon-hyun Sep 14, 2021

Choose a reason for hiding this comment

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

Ah, got it~ :)

"relocation of serialized objects supported.")
}
canDoPushBasedShuffle
}

/**
Expand Down
11 changes: 9 additions & 2 deletions 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 @@ -1509,10 +1509,17 @@ class UtilsSuite extends SparkFunSuite with ResetSystemProperties with Logging {
assert(Utils.isPushBasedShuffleEnabled(conf) === false)
conf.set(SHUFFLE_SERVICE_ENABLED, true)
conf.set(SparkLauncher.SPARK_MASTER, "yarn")
conf.set("spark.yarn.maxAttempts", "1")
conf.set("spark.yarn.maxAppAttempts", "1")
conf.set(SERIALIZER, "org.apache.spark.serializer.KryoSerializer")
assert(Utils.isPushBasedShuffleEnabled(conf) === true)
conf.set("spark.yarn.maxAttempts", "2")
conf.set("spark.yarn.maxAppAttempts", "2")
Copy link
Contributor

@mridulm mridulm Sep 13, 2021

Choose a reason for hiding this comment

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

While this test is no longer required (since we dont disable if multiple attempts are present - like we used to earlier), give it is a minor change, it is fine to include in this PR.

assert(Utils.isPushBasedShuffleEnabled(conf) === true)
conf.set(IO_ENCRYPTION_ENABLED, true)
assert(Utils.isPushBasedShuffleEnabled(conf) === false)
conf.set(IO_ENCRYPTION_ENABLED, false)
assert(Utils.isPushBasedShuffleEnabled(conf) === true)
conf.set(SERIALIZER, "org.apache.spark.serializer.JavaSerializer")
assert(Utils.isPushBasedShuffleEnabled(conf) === false)
}
}

Expand Down