背景

  今天在spark-shell临时性导出300万数数据生成csv文件的时候，其中我启动spark-shell的语句如下；

# 用yarn模式运行spark，client模式提交任务，运行在etl队列，分配4个executors 每个1核 2G内存
spark-shell --master yarn --deploy-mode client --executor-memory 2G --executor-cores 1 --num-executors 4 --queue etl

  生成临时性csv文件如下；

# coalesce(1)表示最后生成一个文件输出
spark.sql("select * from dw.ods_rs_media_tbi_media_location_kkf where event_day='20191111'").coalesce(1).write.option("header", true).csv("/source/kkf_temp.csv")

  报错如下；

2020-05-29 17:13:08 WARN  TaskSetManager:66 - Lost task 1.2 in stage 2.0 (TID 19, shucang-02.szanba.ren, executor 2): org.apache.spark.SparkException: Kryo serialization failed: Buffer overflow. Available: 0, required: 30076801. To avoid this, increase spark.kryoserializer.buffer.max value.
        at org.apache.spark.serializer.KryoSerializerInstance.serialize(KryoSerializer.scala:350)
        at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:393)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
        at java.lang.Thread.run(Thread.java:748)
Caused by: com.esotericsoftware.kryo.KryoException: Buffer overflow. Available: 0, required: 30076801
        at com.esotericsoftware.kryo.io.Output.require(Output.java:163)
        at com.esotericsoftware.kryo.io.Output.writeBytes(Output.java:246)
        at com.esotericsoftware.kryo.io.Output.writeBytes(Output.java:232)
        at com.esotericsoftware.kryo.serializers.DefaultArraySerializers$ByteArraySerializer.write(DefaultArraySerializers.java:54)
        at com.esotericsoftware.kryo.serializers.DefaultArraySerializers$ByteArraySerializer.write(DefaultArraySerializers.java:43)
        at com.esotericsoftware.kryo.Kryo.writeClassAndObject(Kryo.java:628)
        at com.twitter.chill.Tuple2Serializer.write(TupleSerializers.scala:37)
        at com.twitter.chill.Tuple2Serializer.write(TupleSerializers.scala:33)
        at com.esotericsoftware.kryo.Kryo.writeClassAndObject(Kryo.java:628)
        at com.esotericsoftware.kryo.serializers.DefaultArraySerializers$ObjectArraySerializer.write(DefaultArraySerializers.java:366)
        at com.esotericsoftware.kryo.serializers.DefaultArraySerializers$ObjectArraySerializer.write(DefaultArraySerializers.java:307)
        at com.esotericsoftware.kryo.Kryo.writeClassAndObject(Kryo.java:628)
        at org.apache.spark.serializer.KryoSerializerInstance.serialize(KryoSerializer.scala:347)
        ... 4 more

  这个bug的意思是说你spark序列化的缓存不够，撑爆了，那其实解决的思路就是把序列缓存加大，注意一定要配置在启动spark-shell环境的参数里，则在原来启动spark-shell的基础上加上--conf spark.kryoserializer.buffer.max=256m --conf spark.kryoserializer.buffer=64m,问题得到解决，这个值可以根据自己的需要来加，具体启动如下，sparkConf 和SparkSession略有区别；

spark-shell --master yarn --deploy-mode client --executor-memory 2G --executor-cores 1 --num-executors 4 --queue etl --conf spark.kryoserializer.buffer.max=256m --conf spark.kryoserializer.buffer=64m

  如果你是写scala/java/py-spark代码遇到这个bug，那么也是一样在启动时配上即可，scala为例子如下，sparkConf 和SparkSession略有区别；
  sparkConf 版本；

// sparkConf 方式
sparkConf.set("spark.kryoserializer.buffer", "64m")
sparkConf.set("spark.kryoserializer.buffer.max", "128m")

//可以打印你设置的值
println( sparkConf.get("spark.kryoserializer.buffer.max") )

  SparkSession 版本；

//SparkSession方式
 def createSpark(): SparkSession = {
    val spark = SparkSession
      .builder()
      .appName("kryoserializer")
      .enableHiveSupport()
      .config("spark.kryoserializer.buffer", "64m")
      .config("spark.kryoserializer.buffer.max", "256m")
      .master("local")
      .getOrCreate()
    spark
  }
  
//可以打印你设置的值
println( spark.conf.get("spark.kryoserializer.buffer.max") )

  题外话：我测试了下在启动的spark-sql环境里面，利用set spark.kryoserializer.buffer=64m和set spark.kryoserializer.buffer.max=256m运行不报错，但是依旧bug不解决问题，思考了下，应该是在启动spark环境的时候才是配置是优先级最高的，如果已经进入了spark环境，那么这种类似配置类的参数就已经定型了，无法修改，还是会选择spark-conf.default的配置，这纯属个人想法，没看源码验证，知道的大佬可以评论区验证下。

set spark.kryoserializer.buffer=64m
set spark.kryoserializer.buffer.max=256m


select * from dw.ods_rs_media_tbi_media_location_kkf where event_day='20191111';

--运行不报错，但是不起效果，依然报错