Spark 1.6.1 SASL

问题:

I wonder if anyone gotten SASL to work with Spark 1.6.1 on YARN?

Basically Spark documentation states that you only require 3 parameters enabled:

spark.authenticate.enableSaslEncryption=true    
spark.network.sasl.serverAlwaysEncrypt=true
spark.authenticate=true

http://spark.apache.org/docs/latest/security.html

However, upon launching my spark job with --master yarn and --deploy-mode client, I see the following in my spark executors logs:

6/05/17 06:50:51 ERROR client.TransportClientFactory: Exception while bootstrapping client after 29 ms

java.lang.RuntimeException: java.lang.IllegalArgumentException: Unknown message type: -22
        at org.apache.spark.network.shuffle.protocol.BlockTransferMessage$Decoder.fromByteBuffer(BlockTransferMessage.java:67)
        at org.apache.spark.network.shuffle.ExternalShuffleBlockHandler.receive(ExternalShuffleBlockHandler.java:71)
        at org.apache.spark.network.server.TransportRequestHandler.processRpcRequest(TransportRequestHandler.java:149)
        at org.apache.spark.network.server.TransportRequestHandler.handle(TransportRequestHandler.java:102)
        at org.apache.spark.network.server.TransportChannelHandler.channelRead0(TransportChannelHandler.java:104)
        at org.apache.spark.network.server.TransportChannelHandler.channelRead0(TransportChannelHandler.java:51)
        at io.netty.channel.SimpleChannelInboundHandler.channelRead(SimpleChannelInboundHandler.java:105)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:333)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:319)
        at io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:254)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:333)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:319)
        at io.netty.handler.codec.MessageToMessageDecoder.channelRead(MessageToMessageDecoder.java:103)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:333)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:319)
        at org.apache.spark.network.util.TransportFrameDecoder.channelRead(TransportFrameDecoder.java:86)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:333)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:319)
        at io.netty.channel.DefaultChannelPipeline.fireChannelRead(DefaultChannelPipeline.java:787)
        at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:130)
        at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:511)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:468)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:382)
        at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:354)
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:116)
        at java.lang.Thread.run(Thread.java:745)

I'm still troubleshooting this. However, it will be awesome if someone has seen this before.

回答1:

You will need to set the spark.authenticate=true in YARN as well.

Excerpted from YarnShuffleService.java in the Spark code base:

 * The service also optionally supports authentication. This ensures that executors from one
 * application cannot read the shuffle files written by those from another. This feature can be
 * enabled by setting `spark.authenticate` in the Yarn configuration before starting the NM.
 * Note that the Spark application must also set `spark.authenticate` manually and, unlike in
 * the case of the service port, will not inherit this setting from the Yarn configuration. This
 * is because an application running on the same Yarn cluster may choose to not use the external
 * shuffle service, in which case its setting of `spark.authenticate` should be independent of
 * the service's.

You can do this by adding the following to core-site.xml in your hadoop config.

<property>
    <name>spark.authenticate</name><value>true</value>
</property>