[SPARK-21494] Spark 2.2.0 AES encryption not working with External shuffle

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.2.0
• Fix Version/s: 2.2.1, 2.3.0
• Component/s: Block Manager, Shuffle, Spark Core

Environment

AWS EMR

• Assignee: Marcelo Masiero Vanzin
• Reporter: Udit Mehrotra
• Created: 21/Jul/17 00:16
• Updated: 17/May/20 18:21
• Resolved: 26/Jul/17 00:58

Description

Spark’s new AES based authentication mechanism does not seem to work when configured with external shuffle service on YARN.

Here is the stack trace for the error we see in the driver logs:
ERROR YarnScheduler: Lost executor 40 on ip-10-167-104-125.ec2.internal: Unable to create executor due to Unable to register with external shuffle server due to: java.lang.IllegalArgumentException: Authentication failed.
at org.apache.spark.network.crypto.AuthRpcHandler.receive(AuthRpcHandler.java:125)
at org.apache.spark.network.server.TransportRequestHandler.processRpcRequest(TransportRequestHandler.java:157)
at org.apache.spark.network.server.TransportRequestHandler.handle(TransportRequestHandler.java:105)
at org.apache.spark.network.server.TransportChannelHandler.channelRead(TransportChannelHandler.java:118)
at org.spark_project.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
at org.spark_project.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
at org.spark_project.io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
at org.spark_project.io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:287)
at org.spark_project.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
at org.spark_project.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
at org.spark_project.io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
at org.spark_project.io.netty.handler.codec.MessageToMessageDecoder.channelRead(MessageToMessageDecoder.java:102)

Here are the settings we are configuring in ‘spark-defaults’ and ‘yarn-site’:
spark.network.crypto.enabled true
spark.network.crypto.saslFallback false
spark.authenticate true

Turning on DEBUG logs for class ‘org.apache.spark.network.crypto’ on both Spark and YARN side is not giving much information either about why authentication fails. The driver and node manager logs have been attached to the JIRA.

Issue Links

• links to: [Github] Pull Request #18706 (vanzin)