版本
spark 3.1.x
hive 1.2.x
hadoop 2.6.0-cdh-5.13.1
背景
由于好多公司的yarn集群用的是cdh版本的，用Cloudera Manager管理的。而截止到目前为止，spark的最新版本已经到了3.1.1。而对于cdh 2.6.0-5.13.1来说目前支持的hive版本为1.2.1，所以我们做一下集中尝试：
- 直接修改pom文件中的hive对应的版本
- 直接修改编译的hadoop版本
- 在spark运行的时候，动态加载hive对应的版本包
直接修改pom文件中的hive对应的版本
直接在spark的父pom文件增加如下proflie信息：
<profile>
<id>hive-1.2</id>
<properties>
<hive.version>1.2.1</hive.version>
<!-- Version used for internal directory structure -->
<hive.version.short>1.2</hive.version.short>
<hive.storage.version>2.6.0</hive.storage.version>
<datanucleus-core.version>3.2.10</datanucleus-core.version>
</properties>
</profile>
运行
./dev/make-distribution.sh --name 2.6.0-cdh5.13.1 --pip --tgz -Phive-1.2 -Phive-thriftserver -Pyarn
报错：
[INFO] --- scala-maven-plugin:4.3.0:compile (scala-compile-first) @ spark-hive_2.12 ---
[INFO] Using incremental compilation using Mixed compile order
[INFO] Compiler bridge file: .sbt/1.0/zinc/org.scala-sbt/org.scala-sbt-compiler-bridge_2.12-1.3.1-bin_2.12.10__52.0-1.3.1_20191012T045515.jar
[INFO] compiler plugin: BasicArtifact(com.github.ghik,silencer-plugin_2.12.10,1.6.0,null)
[INFO] Compiling 29 Scala sources and 2 Java sources to spark/sql/hive/target/scala-2.12/classes ...
[ERROR] [Error] spark/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveShim.scala:29: object SerializationUtilities is not a msmber of package org.apache.hadoop.hive.ql.exec
[ERROR] [Error] spark/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveShim.scala:150: not found: value SerializationUtilities
[ERROR] [Error] spark/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveShim.scala:154: not found: value SerializationUtilities
[ERROR] [Error] spark/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveUDFs.scala:350: too many arguments (4) for constructor SimpleGenericUDAFParameterInfo: (x$1: Array[org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector], x$2: Boolean, x$3: Boolean)org.apache.hadoop.hive.ql.udf.generic.SimpleGenericUDAFParameterInfo
[ERROR] four errors found
说明hive1.2.1版本的不兼容
直接修改编译的hadoop版本
直接修改hadoop的版本为2.6.0-cdh5.13.1
运行如下命令：
./dev/make-distribution.sh --name 2.6.0-cdh5.13.1 --pip --tgz -Phive-1.2 -Phive-thriftserver -Pyarn -Dhadoop.version=2.6.0-cdh5.13.1
报错：
[INFO] --- scala-maven-plugin:4.3.0:compile (scala-compile-first) @ spark-core_2.12 ---
[INFO] Using incremental compilation using Mixed compile order
[INFO] Compiler bridge file: .sbt/1.0/zinc/org.scala-sbt/org.scala-sbt-compiler-bridge_2.12-1.3.1-bin_2.12.10__52.0-1.3.1_20191012T045515.jar
[INFO] compiler plugin: BasicArtifact(com.github.ghik,silencer-plugin_2.12.10,1.6.0,null)
[INFO] Compiling 560 Scala sources and 99 Java sources to spark/core/target/scala-2.12/classes ...
[ERROR] [Error] spark/core/src/main/scala/org/apache/spark/ui/HttpSecurityFilter.scala:107: type mismatch;
found : K where type K
required: String
[ERROR] [Error] spark/core/src/main/scala/org/apache/spark/ui/HttpSecurityFilter.scala:107: value map is not a member of V
[ERROR] [Error] spark/core/src/main/scala/org/apache/spark/ui/HttpSecurityFilter.scala:107: missing argument list for method stripXSS in class XssSafeRequest
Unapplied methods are only converted to functions when a function type is expected.
You can make this conversion explicit by writing `stripXSS _` or `stripXSS(_)` instead of `stripXSS`.
[ERROR] [Error] spark/core/src/main/scala/org/apache/spark/ui/PagedTable.scala:307: value startsWith is not a member of K
[ERROR] [Error] spark/core/src/main/scala/org/apache/spark/util/Utils.scala:580: value toLowerCase is not a member of object org.apache.hadoop.util.StringUtils
[ERROR] 5 errors found
说明对2.6.0-cdh5.13.1版本的不兼容
在spark运行的时候，动态加载hive对应的版本包
根据官网的说明 ,spark从1.4.0 开始就能和不同的hive元数据进行交互，也就是说spark编译的hive内部版本和spark访问hive的元数据是独立的，可以配置不同的hive版本进行对应元数据的访问。具体的配置可以参考以上官网配置。
网友评论