Hadoop + Hive + Apache Ranger 源码编译记录

背景介绍

由于 CDH（Cloudera's Distribution Hadoop ）近几年已经开始收费并限制节点数量和版本升级，最近使用开源的 hadoop 搭了一套测试集群，其中的权限管理组件用到了Apache Ranger，所以记录一下编译打包过程。

组件版本

Hadoop 2.8.5
Hive 2.3.7
Hbase 1.4.9
Ranger 1.2.0

注意：由于 Ranger 2.0 版本之后支持的 Hadoop 是 3.x 的版本，所以选择了 2.0 之前最新的 Ranger 1.2.0 版本

编译步骤

下载源码

复制代码

# 克隆源码
git clone https://github.com/apache/ranger.git
# 切换分支
git checkout release-ranger-1.2.0

修改配置

在根目录的 pom.xml 中修改版本：

Hadoop 从 2.7.1 修改到 2.8.5

Hive 从 2.3.4 修改到 2.3.7

Hbase 从 1.3.2 修改 1.4.9

各位可以根据需要，进行修改，当然仅修改这里很简单，但不代表编译就兼容

进行编译

复制代码

mvn clean compile package assembly:assembly install -Dmaven.test.skip  -DskipTests -Drat.skip=true -Dpmd.skip=true

成功后，可见如下信息：

打包后的产物就在 target 目录中：

这里面有我们部署的 admin 包，还有各个支持 ranger 权限控制的插件

错误记录

1，Hadoop API 不兼容

https://lists.apache.org/thread/rpk62z0x1drs7oqss1rmy0rrwnp1r5j6

修改 KMSAuthenticationFilter.java 类的

复制代码

props.setProperty(DelegationTokenAuthenticationHandler.TOKEN_KIND,
                  KMSClientProvider.TOKEN_KIND);
// 调整为Text 转成 toString
props.setProperty(DelegationTokenAuthenticationHandler.TOKEN_KIND,
                  KMSClientProvider.TOKEN_KIND.toString());

2，Hbase API 不兼容

a. 修改 ranger-hbase-plugin 子模块：

RangerAuthorizationCoprocessor.java：新增实现接口

复制代码

	@Override
	public void preGetClusterStatus(ObserverContext<MasterCoprocessorEnvironment> observerContext) throws IOException {

	}

	@Override
	public void postGetClusterStatus(ObserverContext<MasterCoprocessorEnvironment> observerContext, ClusterStatus clusterStatus) throws IOException {

	}

	@Override
	public void preClearDeadServers(ObserverContext<MasterCoprocessorEnvironment> observerContext) throws IOException {

	}

	@Override
	public void postClearDeadServers(ObserverContext<MasterCoprocessorEnvironment> observerContext, List<ServerName> list, List<ServerName> list1) throws IOException {

	}

	@Override
	public void preMoveServers(ObserverContext<MasterCoprocessorEnvironment> ctx, Set<Address> servers, String targetGroup) throws IOException {

	}

	@Override
	public void postMoveServers(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> servers, String s) throws IOException {

	}


	@Override
	public void preMoveServersAndTables(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set, Set<TableName> set1, String s) throws IOException {

	}

	@Override
	public void postMoveServersAndTables(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set, Set<TableName> set1, String s) throws IOException {

	}

	@Override
	public void preRemoveServers(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set) throws IOException {

	}

	@Override
	public void postRemoveServers(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set) throws IOException {

	}

RangerAuthorizationCoprocessorBase.java：修改参数类型

复制代码

    public void preMoveServers(final ObserverContext<MasterCoprocessorEnvironment> ctx, Set<HostAndPort> servers, String targetGroup) throws IOException {}
    public void postMoveServers(ObserverContext<MasterCoprocessorEnvironment> ctx, Set<HostAndPort> servers, String targetGroup) throws IOException {}

// 调整参数类型，成如下：
    public void preMoveServers(final ObserverContext<MasterCoprocessorEnvironment> ctx, Set<Address> servers, String targetGroup) throws IOException {}
    public void postMoveServers(ObserverContext<MasterCoprocessorEnvironment> ctx, Set<Address> servers, String targetGroup) throws IOException {}

b.修改 ranger-hbase-plugin-shim 子模块

RangerAuthorizationCoprocessor.java：新增实现接口，注释掉部分接口

复制代码

// 新增
    @Override
	public void preCommitStoreFile(ObserverContext<RegionCoprocessorEnvironment> observerContext, byte[] bytes, List<Pair<Path, Path>> list) throws IOException {

	}

	@Override
	public void postCommitStoreFile(ObserverContext<RegionCoprocessorEnvironment> observerContext, byte[] bytes, Path path, Path path1) throws IOException {

	}


	@Override
	public void preMoveServersAndTables(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set, Set<TableName> set1, String s) throws IOException {

	}

	@Override
	public void postMoveServersAndTables(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set, Set<TableName> set1, String s) throws IOException {

	}

	@Override
	public void preRemoveServers(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set) throws IOException {

	}

	@Override
	public void postRemoveServers(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set) throws IOException {

	}

	public void preBalanceRSGroup(final ObserverContext<MasterCoprocessorEnvironment> ctx, String groupName) throws IOException {}
	@Override
	public void preGetClusterStatus(ObserverContext<MasterCoprocessorEnvironment> observerContext) throws IOException {

	}

	@Override
	public void postGetClusterStatus(ObserverContext<MasterCoprocessorEnvironment> observerContext, ClusterStatus clusterStatus) throws IOException {

	}

	@Override
	public void preClearDeadServers(ObserverContext<MasterCoprocessorEnvironment> observerContext) throws IOException {

	}

	@Override
	public void postClearDeadServers(ObserverContext<MasterCoprocessorEnvironment> observerContext, List<ServerName> list, List<ServerName> list1) throws IOException {

	}

	@Override
	public void preMoveServers(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set, String s) throws IOException {

	}

	@Override
	public void postMoveServers(ObserverContext<MasterCoprocessorEnvironment> observerContext, Set<Address> set, String s) throws IOException {

	}


// 注释掉
//  public void preMoveServers(final ObserverContext<MasterCoprocessorEnvironment> ctx, Set<HostAndPort> servers, String targetGroup) throws IOException {}
//  public void postMoveServers(ObserverContext<MasterCoprocessorEnvironment> ctx, Set<HostAndPort> servers, String targetGroup) throws IOException {}

3，Python 版本不兼容

$ERROR$ around Ant part ...<exec failοnerrοr="true" executable="python">... @ 4:48

修复：ranger 编译仅仅支持 python 2.x 版本，如果我们是 python 3.x 的版本需要切换为 2 的版本，如果你安装了 conda 也很简单：

复制代码

# 切换环境
conda activate py27

4，PMD 规则校验不通过

这个直接打包时禁用掉 pmd 检查即可

复制代码

 -Dpmd.skip=true

注意事项

经过上述调整过后，我们的编译终于成功了，但需要注意的是 API 的兼容仅仅是为了不报错，如果真的需要到某个组件的功能时，建议移值高版本的特性过来，避免某些功能不能正常 work 问题