spark1.x升级spark2如何升级及需要考虑的问题

问题导读

1.spark2升级哪些内容变化？
2.升级中spark哪些没有发生变化？
3.cloudera中，spark1和spark2能否并存？
4.升级后，可能会遇到什么问题？







spark2出来已经很长时间了，但是由于spark1.6比较稳定，很多依然在使用。如果想使用spark2，那么该如何升级。我们window升级一般为直接点击升级即可，剩下的事情，不用我们管。但是spark的升级确实有点出乎意料。相当于我们直接安装，但是可以借用以前的配置，比如配置文件基本是不变的，如果目录相同，环境变量变化也不大。
如果只是单纯的学习，升级是没有问题的。但是如果我们生产环境，升级就需要注意了，因为升级后会带来不少的负作用。
spark安装参考http://www.aboutyun.com/forum.php?mod=viewthread&tid=20620

下面介绍如何升级：

1.spark升级
首先停止所有服务[mw_shl_code=bash,true]./stop-all.sh[/mw_shl_code]
这里额外补充一些内容：
spark有stop-all.sh,


hadoop也有同样的命令，只不过hadoop在准备弃用下面两个命令。那么如果想使用这两个命令，我们最好到对应的目录里面sbin，然后执行[mw_shl_code=bash,true]./stop-all.sh[/mw_shl_code]

既然手工配置，升级我们需要考虑的问题：

1.配置文件是否变化
参考官网spark1.x和2.x所幸应该是没有变化的，配置文件还是那些。
http://spark.apache.org/docs/latest/spark-standalone.html，这样升级就放心了，因为我们可以使用原先的配置文件，不能再麻烦了。

2.变化的有哪些

我们停止集群后，后面开始相关的配置。
我这里的spark版本为1.6,这里要升级为2.2
首先重命名spark文件夹
[mw_shl_code=bash,true]sudo mv spark spark1.6[/mw_shl_code]
解压spark2.2包
[mw_shl_code=bash,true] sudo tar zxvf spark-2.2.0-bin-hadoop2.7.tgz -C /data
[/mw_shl_code]
查看权限为500


为了防止出现问题，因此改变下权限：
[mw_shl_code=bash,true]sudo chown -R aboutyun:aboutyun spark-2.2.0-bin-hadoop2.7/
[/mw_shl_code]
[mw_shl_code=bash,true]sudo chmod -R 777 spark-2.2.0-bin-hadoop2.7/[/mw_shl_code]
我们队这个文件夹重命名
[mw_shl_code=bash,true]sudo mv spark-2.2.0-bin-hadoop2.7/ spark[/mw_shl_code]

将spark1.6的文件spark-env.sh、slaves、spark-defaults.conf复制到spark
对于三个文件，如果都比较完善的话，是不需要修改的
slaves
机器不变化，是不需要修改的。
spark-env.sh
JAVA_HOME=/data/jdk1.8
SCALA_HOME=/data/scala2
SPARK_MASTER_HOST=192.168.1.10
HADOOP_CONF_DIR=/data/hadoop/etc/hadoop
SPARK_LOCAL_DIR=/data/spark_data
SPARK_WORKER_DIR=/data/spark_data/spark_works

说明：SPARK_MASTER_IP在spark1.x中，spark2中使用的是SPARK_MASTER_HOST

spark-defaults.conf
spark.master                        spark://master:7077
spark.eventLog.enabled                true
spark.eventLog.dir                file:///data/spark_data/history/event-log
spark.serializer                org.apache.spark.serializer.KryoSerializer
spark.history.fs.LogDirectory        file:///data/spark_data/history/spark-events        

上面都不需要修改，当然如果需要调整的自行修改即可。

修改环境变量
~/.bashrc
[mw_shl_code=bash,true]export HADOOP_HOME=/data/hadoop
export SPARK_HOME=/data/spark
export ZOOKEEPER_HOME=/data/zookeeper-3.4.6
export KAFKA_HOME=/data/kafka_2.11
export HIVE_HOME=/data/hive-1.2.1
export PATH=$HIVE_HOME/bin:$KAFKA_HOME/bin:$ZOOKEEPER_HOME/bin:$HADOOP_HOME/bin:$HADOOP_HOME/sbin:$SPARK_HOME/bin:$SPARK_HOME/sbin:$PATH

export FLUME_HOME=/data/flume-1.6.0
export PATH=$FLUME_HOME/bin:$PATH
[/mw_shl_code]

[mw_shl_code=applescript,true]source ~/.bashrc
[/mw_shl_code]
这一步很重要，否则可能还是原先的版本


上面由于我们文件名为spark，因此不需要修改。

接着我们复制到其它客户端：
[mw_shl_code=bash,true]scp -r spark aboutyun@slave1:/data[/mw_shl_code]

[mw_shl_code=bash,true]scp -r spark aboutyun@slave2:/data[/mw_shl_code]

在远程复制的时候，需要记得将slave1和slave2的hadoop文件夹删除，否则会将hadoop2.7.4和hadoop2.6.5包混合
说明：
一般来讲我们是不能直接复制到非home目录的，所以我们需要将data文件夹授权为777，这样我们才能远程复制成功。
接着我们启动spark，进入spark的sbin目录
[mw_shl_code=bash,true]./start-all.sh
[/mw_shl_code]

对于spark的升级，注意如果使用的是hadoop，需要对应hadoop版本，否则可能会出错。对于Scala版本同样需要注意，Scala支持版本为2.11


#########################cloudera升级
除了spark原生态升级，对于cloudera升级就比较简单了，cloudera中，spark1.6和spark2是可以并存的，直接安装spark2即可。
#########################
spark升级带来哪些副作用

如果我们已经线上使用，那么需要谨慎升级，否则可能会发生预料之外的事情。下面内容仅供大家参考
计算准确性
SELECT '0.1' = 0返回的是true!Spark 2.2中，0.1会被转换为int，如果你的数据类型全部是文本类型，做数值计算时，结果极有可能不正确。之前的版本中0.1会被转换为double类型绝大多数场景下这样的处理是正确的。目前为止，社区还没有很好的处理这个问题，针对这个问题，我给社区提交过一个PR，想要自己解决这个问题的同学，可以手动合并下：https://github.com/apache/spark/pull/18986
过于复杂的SQL语句执行可能会出现64KB字节码编译限制的问题，这算是个老问题了，Spark自从上了Tungsten基本上一直存在这个问题，也算是受到了JVM的限制，遇到此类问题，建议大家找找PR：https://github.com/apache/spark/search?utf8=%E2%9C%93&q=64KB&type=Issues
数据计算精度有问题，SELECT 1 > 0.0001会报错，这个问题已在2.1.2及2.2.0中修复：https://issues.apache.org/jira/browse/SPARK-20211
2.1.0版本中INNER JOIN涉及到常量计算结果不正确，后续版本已修复：https://issues.apache.org/jira/browse/SPARK-19766
2.1.0中，执行GROUPING SET(col)，如果col列数据为null，会报空指针异常，后续版本已修复：https://issues.apache.org/jira/browse/SPARK-19509
2.1.0中，嵌套的CASE WHEN语句执行有可能出错，后续版本已修复：https://issues.apache.org/jira/browse/SPARK-19472


行为变化
那些不算太致命，改改代码或配置就可以兼容的问题。
Spark 2.2的UDAF实现有所变动，如果你的Hive UDAF没有严格按照标准实现，有可能会计算报错或数据不正确，建议将逻辑迁移到Spark AF，同时也能获得更好的性能
Spark 2.1开始全表读取分区表采用FilePartition的方式，单个Partition内可以读取多个文件，如果对文件做了压缩，这种方式有可能导致查询性能变差，可以适当降低spark.sql.files.maxPartitionBytes的值，默认是128MB(对于大部分的Parquet压缩表来说，这个默认设置其实会导致性能问题)
Spark 2.x限制了Hive表中spark.sql.*相关属性的操作，明明存在的属性，使用SHOW TBLPROPERTIES tb("spark.sql.sources.schema.numParts")无法获取到，同理也无法执行ALTER TABLE tb SET TBLPROPERTIES ('spark.sql.test' = 'test')进行修改
无法修改外部表的属性ALTER TABLE tb SET TBLPROPERTIES ('test' = 'test')这里假设tb是EXTERNAL类型的表
DROP VIEW IF EXISTS tb，如果这里的tb是个TABLE而非VIEW，执行会报错AnalysisException: Cannot drop a table with DROP VIEW，在2.x以下不会报错，由于我们指定了IF EXISTS关键字，这里的报错显然不合理，需要做异常处理。
如果你访问的表不存在，异常信息在Spark2.x里由之前的Table not found变成了Table or view not found，如果你的代码里依赖这个异常信息，就需要注意调整了。
EXPLAIN语句的返回格式变掉了，在1.6里是多行文本，2.x中是一行，而且内容格式也有稍微的变化，相比Spark1.6，少了Tungsten关键字;EXPLAIN中显示的HDFS路径过长的话，在Spark 2.x中会被省略为...

2.x中默认不支持笛卡尔积操作，需要通过参数spark.sql.crossJoin.enabled开启
OLAP分析中常用的GROUPING__ID函数在2.x变成了GROUPING_ID()
如果你有一个基于Hive的UDF名为abc，有3个参数，然后又基于Spark的UDF实现了一个2个参数的abc，在2.x中，2个参数的abc会覆盖掉Hive中3个参数的abc函数，1.6则不会有这个问题
执行类似SELECT 1 FROM tb GROUP BY 1的语句会报错，需要单独设置spark.sql.groupByOrdinal false类似的参数还有spark.sql.orderByOrdinal false
CREATE DATABASE默认路径发生了变化，不在从hive-site.xml读取hive.metastore.warehouse.dir，需要通过Spark的spark.sql.warehouse.dir配置指定数据库的默认存储路径。
CAST一个不存在的日期返回null，如：year('2015-03-40')，在1.6中返回2015
Spark 2.x不允许在VIEW中使用临时函数(temp function)https://issues.apache.org/jira/browse/SPARK-18209
Spark 2.1以后，窗口函数ROW_NUMBER()必须要在OVER内添加ORDER BY，以前的ROW_NUMBER() OVER()执行会报错
Spark 2.1以后，SIZE(null)返回-1，之前的版本返回null
Parquet文件的默认压缩算法由gzip变成了snappy，据官方说法是snappy有更好的查询性能，大家需要自己验证性能的变化
DESC FORMATTED tb返回的内容有所变化，1.6的格式和Hive比较贴近，2.x中分两列显示
异常信息的变化，未定义的函数，Spark 2.x: org.apache.spark.sql.AnalysisException: Undefined function: 'xxx’., Spark 1.6: AnalysisException: undefined function xxx，参数格式错误：Spark 2.x:Invalid number of arguments, Spark 1.6: No handler for Hive udf class org.apache.hadoop.hive.ql.udf.generic.GenericUDAFXXX because: Exactly one argument is expected..
Spark Standalone的WebUI中已经没有这个API了：/api/v1/applications：https://issues.apache.org/jira/browse/SPARK-12299，https://issues.apache.org/jira/browse/SPARK-18683

内容摘自：
http://www.jianshu.com/p/482407c88d27

###########################################
spark升级遇到问题总结

spark的升级后，会遇到很奇怪的问题，
1.进程会有多个master
2.端口无缘无故被暂用
3.进程都正常，master连接不上

启用spark-shell报错如下
[mw_shl_code=bash,true]To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).
17/11/17 11:30:12 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
17/11/17 11:30:14 WARN client.StandaloneAppClient$ClientEndpoint: Failed to connect to master master:7077
org.apache.spark.SparkException: Exception thrown in awaitResult:
        at org.apache.spark.util.ThreadUtils$.awaitResult(ThreadUtils.scala:205)
        at org.apache.spark.rpc.RpcTimeout.awaitResult(RpcTimeout.scala:75)
        at org.apache.spark.rpc.RpcEnv.setupEndpointRefByURI(RpcEnv.scala:100)
        at org.apache.spark.rpc.RpcEnv.setupEndpointRef(RpcEnv.scala:108)
        at org.apache.spark.deploy.client.StandaloneAppClient$ClientEndpoint$$anonfun$tryRegisterAllMasters$1$$anon$1.run(StandaloneAppClient.scala:106)
        at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
        at java.util.concurrent.FutureTask.run(FutureTask.java:266)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
        at java.lang.Thread.run(Thread.java:745)
Caused by: java.lang.RuntimeException: java.io.StreamCorruptedException: invalid stream header: 01000C31
        at java.io.ObjectInputStream.readStreamHeader(ObjectInputStream.java:806)
        at java.io.ObjectInputStream.<init>(ObjectInputStream.java:299)
        at org.apache.spark.serializer.JavaDeserializationStream$$anon$1.<init>(JavaSerializer.scala:64)
        at org.apache.spark.serializer.JavaDeserializationStream.<init>(JavaSerializer.scala:64)
        at org.apache.spark.serializer.JavaSerializerInstance.deserializeStream(JavaSerializer.scala:123)
        at org.apache.spark.serializer.JavaSerializerInstance.deserialize(JavaSerializer.scala:108)
        at org.apache.spark.rpc.netty.NettyRpcEnv$$anonfun$deserialize$1$$anonfun$apply$1.apply(NettyRpcEnv.scala:258)
        at scala.util.DynamicVariable.withValue(DynamicVariable.scala:57)
        at org.apache.spark.rpc.netty.NettyRpcEnv.deserialize(NettyRpcEnv.scala:310)
        at org.apache.spark.rpc.netty.NettyRpcEnv$$anonfun$deserialize$1.apply(NettyRpcEnv.scala:257)
        at scala.util.DynamicVariable.withValue(DynamicVariable.scala:57)
        at org.apache.spark.rpc.netty.NettyRpcEnv.deserialize(NettyRpcEnv.scala:256)
        at org.apache.spark.rpc.netty.NettyRpcHandler.internalReceive(NettyRpcEnv.scala:588)
        at org.apache.spark.rpc.netty.NettyRpcHandler.receive(NettyRpcEnv.scala:570)
        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:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:266)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at io.netty.handler.codec.MessageToMessageDecoder.channelRead(MessageToMessageDecoder.java:103)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at org.apache.spark.network.util.TransportFrameDecoder.channelRead(TransportFrameDecoder.java:86)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at io.netty.channel.DefaultChannelPipeline.fireChannelRead(DefaultChannelPipeline.java:846)
        at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:131)
        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:111)
        at java.lang.Thread.run(Thread.java:745)

        at org.apache.spark.network.client.TransportResponseHandler.handle(TransportResponseHandler.java:207)
        at org.apache.spark.network.server.TransportChannelHandler.channelRead(TransportChannelHandler.java:120)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:287)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.handler.codec.MessageToMessageDecoder.channelRead(MessageToMessageDecoder.java:102)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at org.apache.spark.network.util.TransportFrameDecoder.channelRead(TransportFrameDecoder.java:85)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.channel.DefaultChannelPipeline$HeadContext.channelRead(DefaultChannelPipeline.java:1294)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.DefaultChannelPipeline.fireChannelRead(DefaultChannelPipeline.java:911)
        at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:131)
        at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:643)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:566)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:480)
        at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:442)
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:131)
        at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:144)
        ... 1 more
17/11/17 11:30:33 WARN client.StandaloneAppClient$ClientEndpoint: Failed to connect to master master:7077
org.apache.spark.SparkException: Exception thrown in awaitResult:
        at org.apache.spark.util.ThreadUtils$.awaitResult(ThreadUtils.scala:205)
        at org.apache.spark.rpc.RpcTimeout.awaitResult(RpcTimeout.scala:75)
        at org.apache.spark.rpc.RpcEnv.setupEndpointRefByURI(RpcEnv.scala:100)
        at org.apache.spark.rpc.RpcEnv.setupEndpointRef(RpcEnv.scala:108)
        at org.apache.spark.deploy.client.StandaloneAppClient$ClientEndpoint$$anonfun$tryRegisterAllMasters$1$$anon$1.run(StandaloneAppClient.scala:106)
        at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
        at java.util.concurrent.FutureTask.run(FutureTask.java:266)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
        at java.lang.Thread.run(Thread.java:745)
Caused by: java.lang.RuntimeException: java.io.StreamCorruptedException: invalid stream header: 01000C31
        at java.io.ObjectInputStream.readStreamHeader(ObjectInputStream.java:806)
        at java.io.ObjectInputStream.<init>(ObjectInputStream.java:299)
        at org.apache.spark.serializer.JavaDeserializationStream$$anon$1.<init>(JavaSerializer.scala:64)
        at org.apache.spark.serializer.JavaDeserializationStream.<init>(JavaSerializer.scala:64)
        at org.apache.spark.serializer.JavaSerializerInstance.deserializeStream(JavaSerializer.scala:123)
        at org.apache.spark.serializer.JavaSerializerInstance.deserialize(JavaSerializer.scala:108)
        at org.apache.spark.rpc.netty.NettyRpcEnv$$anonfun$deserialize$1$$anonfun$apply$1.apply(NettyRpcEnv.scala:258)
        at scala.util.DynamicVariable.withValue(DynamicVariable.scala:57)
        at org.apache.spark.rpc.netty.NettyRpcEnv.deserialize(NettyRpcEnv.scala:310)
        at org.apache.spark.rpc.netty.NettyRpcEnv$$anonfun$deserialize$1.apply(NettyRpcEnv.scala:257)
        at scala.util.DynamicVariable.withValue(DynamicVariable.scala:57)
        at org.apache.spark.rpc.netty.NettyRpcEnv.deserialize(NettyRpcEnv.scala:256)
        at org.apache.spark.rpc.netty.NettyRpcHandler.internalReceive(NettyRpcEnv.scala:588)
        at org.apache.spark.rpc.netty.NettyRpcHandler.receive(NettyRpcEnv.scala:570)
        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:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:266)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at io.netty.handler.codec.MessageToMessageDecoder.channelRead(MessageToMessageDecoder.java:103)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at org.apache.spark.network.util.TransportFrameDecoder.channelRead(TransportFrameDecoder.java:86)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at io.netty.channel.DefaultChannelPipeline.fireChannelRead(DefaultChannelPipeline.java:846)
        at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:131)
        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:111)
        at java.lang.Thread.run(Thread.java:745)

        at org.apache.spark.network.client.TransportResponseHandler.handle(TransportResponseHandler.java:207)
        at org.apache.spark.network.server.TransportChannelHandler.channelRead(TransportChannelHandler.java:120)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:287)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.handler.codec.MessageToMessageDecoder.channelRead(MessageToMessageDecoder.java:102)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at org.apache.spark.network.util.TransportFrameDecoder.channelRead(TransportFrameDecoder.java:85)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.channel.DefaultChannelPipeline$HeadContext.channelRead(DefaultChannelPipeline.java:1294)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.DefaultChannelPipeline.fireChannelRead(DefaultChannelPipeline.java:911)
        at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:131)
        at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:643)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:566)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:480)
        at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:442)
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:131)
        at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:144)
        ... 1 more
17/11/17 11:30:53 WARN client.StandaloneAppClient$ClientEndpoint: Failed to connect to master master:7077
org.apache.spark.SparkException: Exception thrown in awaitResult:
        at org.apache.spark.util.ThreadUtils$.awaitResult(ThreadUtils.scala:205)
        at org.apache.spark.rpc.RpcTimeout.awaitResult(RpcTimeout.scala:75)
        at org.apache.spark.rpc.RpcEnv.setupEndpointRefByURI(RpcEnv.scala:100)
        at org.apache.spark.rpc.RpcEnv.setupEndpointRef(RpcEnv.scala:108)
        at org.apache.spark.deploy.client.StandaloneAppClient$ClientEndpoint$$anonfun$tryRegisterAllMasters$1$$anon$1.run(StandaloneAppClient.scala:106)
        at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
        at java.util.concurrent.FutureTask.run(FutureTask.java:266)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
        at java.lang.Thread.run(Thread.java:745)
Caused by: java.lang.RuntimeException: java.io.StreamCorruptedException: invalid stream header: 01000C31
        at java.io.ObjectInputStream.readStreamHeader(ObjectInputStream.java:806)
        at java.io.ObjectInputStream.<init>(ObjectInputStream.java:299)
        at org.apache.spark.serializer.JavaDeserializationStream$$anon$1.<init>(JavaSerializer.scala:64)
        at org.apache.spark.serializer.JavaDeserializationStream.<init>(JavaSerializer.scala:64)
        at org.apache.spark.serializer.JavaSerializerInstance.deserializeStream(JavaSerializer.scala:123)
        at org.apache.spark.serializer.JavaSerializerInstance.deserialize(JavaSerializer.scala:108)
        at org.apache.spark.rpc.netty.NettyRpcEnv$$anonfun$deserialize$1$$anonfun$apply$1.apply(NettyRpcEnv.scala:258)
        at scala.util.DynamicVariable.withValue(DynamicVariable.scala:57)
        at org.apache.spark.rpc.netty.NettyRpcEnv.deserialize(NettyRpcEnv.scala:310)
        at org.apache.spark.rpc.netty.NettyRpcEnv$$anonfun$deserialize$1.apply(NettyRpcEnv.scala:257)
        at scala.util.DynamicVariable.withValue(DynamicVariable.scala:57)
        at org.apache.spark.rpc.netty.NettyRpcEnv.deserialize(NettyRpcEnv.scala:256)
        at org.apache.spark.rpc.netty.NettyRpcHandler.internalReceive(NettyRpcEnv.scala:588)
        at org.apache.spark.rpc.netty.NettyRpcHandler.receive(NettyRpcEnv.scala:570)
        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:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:266)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at io.netty.handler.codec.MessageToMessageDecoder.channelRead(MessageToMessageDecoder.java:103)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at org.apache.spark.network.util.TransportFrameDecoder.channelRead(TransportFrameDecoder.java:86)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:308)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:294)
        at io.netty.channel.DefaultChannelPipeline.fireChannelRead(DefaultChannelPipeline.java:846)
        at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:131)
        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:111)
        at java.lang.Thread.run(Thread.java:745)

        at org.apache.spark.network.client.TransportResponseHandler.handle(TransportResponseHandler.java:207)
        at org.apache.spark.network.server.TransportChannelHandler.channelRead(TransportChannelHandler.java:120)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:287)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.handler.codec.MessageToMessageDecoder.channelRead(MessageToMessageDecoder.java:102)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at org.apache.spark.network.util.TransportFrameDecoder.channelRead(TransportFrameDecoder.java:85)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.channel.DefaultChannelPipeline$HeadContext.channelRead(DefaultChannelPipeline.java:1294)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.DefaultChannelPipeline.fireChannelRead(DefaultChannelPipeline.java:911)
        at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:131)
        at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:643)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:566)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:480)
        at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:442)
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:131)
        at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:144)
        ... 1 more
17/11/17 11:31:13 ERROR cluster.StandaloneSchedulerBackend: Application has been killed. Reason: All masters are unresponsive! Giving up.
17/11/17 11:31:13 WARN cluster.StandaloneSchedulerBackend: Application ID is not initialized yet.
[/mw_shl_code]
显然是端口的问题，这时候排查7077
netstat -anp | grep 7077
发现被暂用，于是kill掉进程。但是依然不行，最后重启,进入spark sbin目录
[mw_shl_code=bash,true]./stop-all.sh
./start-all.sh[/mw_shl_code]
问题得到解决