Spark SQL Thrift Server 配置 Kerberos身份认证和权限管理

　转载请注明出处：http://www.cnblogs.com/xiaodf/

　　之前的博客介绍了通过Kerberos + Sentry的方式实现了hive server2的身份认证和权限管理功能，本文主要介绍Spark SQL JDBC方式操作Hive库时的身份认证和权限管理实现。

　ThriftServer是一个JDBC/ODBC接口，用户可以通过JDBC/ODBC连接ThriftServer来访问SparkSQL的数据。ThriftServer在启动的时候，会启动了一个sparkSQL的应用程序，而通过JDBC/ODBC连接进来的客户端共同分享这个sparkSQL应用程序的资源，也就是说不同的用户之间可以共享数据；ThriftServer启动时还开启一个侦听器，等待JDBC客户端的连接和提交查询。所以，在配置ThriftServer的时候，至少要配置ThriftServer的主机名和端口，如果要使用hive数据的话，还要提供hive metastore的uris。

前提：

　　本文是在以下几个部署前提下进行的实验：

　　（1）CDH 开启了Kerberos身份认证，并安装了Sentry；

　　（2）Hive权限通过Sentry服务控制；

　　（3）HDFS开启了HDFS ACL与Sentry的权限同步功能，通过sql语句更改Hive表的权限，会同步到相应的HDFS文件。

　　以上各项配置可参考我之前博客：http://www.cnblogs.com/xiaodf/p/5968248.html

订阅我的微信公众号《大数据技术进阶》，及时查看更多大数据技术实践分享

1、Thrift Server 安装配置

1.1. 下载spark安装包

CDH自带的spark不支持thrift server，所以需要自行下载spark编译好的安装包，下载地址如下：http://spark.apache.org/downloads.html

本文下载的spark版本为1.5.2，

1.2. 添加配置文件

　　将集群hive-site.xml文件拷贝到spark目录的conf下

[root@t162 spark-1.5.2-bin-hadoop2.6]# cd conf/
[root@t162 conf]# ll
total 52
-rw-r--r-- 1 root root  202 Oct 25 13:05 docker.properties.template
-rw-r--r-- 1 root root  303 Oct 25 13:05 fairscheduler.xml.template
-rw-r--r-- 1 root root 5708 Oct 25 13:08 hive-site.xml
-rw-r--r-- 1 root root  949 Oct 25 13:05 log4j.properties.template
-rw-r--r-- 1 root root 5886 Oct 25 13:05 metrics.properties.template
-rw-r--r-- 1 root root   80 Oct 25 13:05 slaves.template
-rw-r--r-- 1 root root  507 Oct 25 13:05 spark-defaults.conf.template
-rwxr-xr-x 1 root root 4299 Oct 25 13:08 spark-env.sh
-rw-r--r-- 1 root root 3418 Oct 25 13:05 spark-env.sh.template
-rwxr-xr-x 1 root root  119 Oct 25 13:09 stopjdbc.sh

　　修改hive-site.xml参数hive.server2.enable.doAs为true，注意doAs务必是true，否则spark jdbc用户权限控制会失效。

  <property>
    <name>hive.server2.enable.doAs</name>
    <value>true</value>
  </property>

　　生成spark-env.sh文件，并添加参数

export  LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/cloudera/parcels/CDH/lib/hadoop/lib/native

　　此处HADOOP_HOME=/opt/cloudera/parcels/CDH/lib/hadoop

1.3. 生成启动thrift服务脚本startjdbc.sh

　　调用start-thriftserver.sh脚本启动thrift server

#!/bin/sh
#start Spark-thriftserver
export YARN_CONF_DIR=/etc/hadoop/conf
file="hive-site.xml"
dir=$(pwd)
cd conf/
if [ ! -e "$file" ]
then
 cp /etc/hive/conf/hive-site.xml $dir/conf/
fi
cd ../sbin
./start-thriftserver.sh --name SparkJDBC --master yarn-client  --num-executors 10 --executor-memory 2g --executor-cores 4 --driver-memory 10g 
--driver-cores 2  --conf spark.storage.memoryFraction=0.2 --conf spark.shuffle.memoryFraction=0.6 --hiveconf hive.server2.thrift.port=10001 
--hiveconf hive.server2.logging.operation.enabled=true --hiveconf hive.server2.authentication.kerberos.principal=hive/t162@HADOOP.COM 
--hiveconf hive.server2.authentication.kerberos.keytab /home/hive.keytab　　

　　上面脚本实际上就是提交了一个spark job，其中主要参数如下：

　　master ：指定spark提交模式为yarn-client

　　hive.server2.thrift.port : 指定thrift server的端口

　　hive.server2.authentication.kerberos.principal：指定启动thrift server的超级管理员principal，此处超级管理员为hive

　　hive.server2.authentication.kerberos.keytab : 超级管理员对应的keytab　

　执行startjdbc.sh需要kinit到hive库的超管来执行，hive库的超管需要在开启sentry与hdfs权限同步基础上，被赋予整个hive库的权限，即对hive库的hdfs整个目录也有所有权限。

1.4. 生成关闭thrift服务脚本stopjdbc.sh

#!/bin/sh
# Stop SparkJDBC
cd sbin
./spark-daemon.sh stop org.apache.spark.sql.hive.thriftserver.HiveThriftServer2 1

2、认证测试

　　Spark SQL Thriftserver认证，目的是让不同的用户，使用不同的身份来登录beeline。使用Kerberos，的确可以解决服务互相认证、用户认证的功能。

2.1. 启动thrift server

　　使用使用管理员账户启动，已配置在启动脚本中。thriftserver实际是个spark Job，通过spark-submit提交到YARN上去，需要这个账户用来访问YARN和HDFS；如果使用一些普通账户，由于HDFS权限不足，可能启动不了，因为需要往HDFS写一些东西。

[root@t162 spark-1.5.2-bin-hadoop2.6]# ./startjdbc.sh
starting org.apache.spark.sql.hive.thriftserver.HiveThriftServer2, logging to /home/iie/spark-1.5.2-bin-hadoop2/spark-1.5.2-bin-hadoop2.6/sbin/../logs/
spark-root-org.apache.spark.sql.hive.thriftserver.HiveThriftServer2-1-t162.out
......
16/10/25 16:56:07 INFO thrift.ThriftCLIService: Starting ThriftBinaryCLIService on port 10001 with 5...500 worker threads

　　可以通过输出日志查看服务启动情况

[root@t162 spark-1.5.2-bin-hadoop2.6]# tailf /home/iie/spark-1.5.2-bin-hadoop2/spark-1.5.2-bin-hadoop2.6/sbin/../logs/
spark-root-org.apache.spark.sql.hive.thriftserver.HiveThriftServer2-1-t162.out

2.2. 通过beeline连接thrift server

　　因为服务启动了kerberos身份认证，没有认证时连接服务会报错,如下所示：

[root@t161 ~]# beeline -u "jdbc:hive2://t162:10001/;principal=hive/t162@HADOOP.COM"
16/10/25 16:59:04 WARN mapreduce.TableMapReduceUtil: The hbase-prefix-tree module jar containing PrefixTreeCodec is not present.  Continuing without it.
scan complete in 2ms
Connecting to jdbc:hive2://t162:10001/;principal=hive/t162@HADOOP.COM
16/10/25 16:59:06 [main]: ERROR transport.TSaslTransport: SASL negotiation failure
javax.security.sasl.SaslException: GSS initiate failed [Caused by GSSException: No valid credentials provided (Mechanism level: Failed to find any Kerberos tgt)]

　　我们用user1用户进行认证，就可以连接了，用户事先已创建，创建方式见http://www.cnblogs.com/xiaodf/p/5968282.html

[root@t161 ~]# kinit user1
Password for user1@HADOOP.COM:
[root@t161 ~]# beeline -u "jdbc:hive2://t162:10001/;principal=hive/t162@HADOOP.COM"
16/10/25 17:01:46 WARN mapreduce.TableMapReduceUtil: The hbase-prefix-tree module jar containing PrefixTreeCodec is not present.  Continuing without it.
scan complete in 3ms
Connecting to jdbc:hive2://t162:10001/;principal=hive/t162@HADOOP.COM
Connected to: Spark SQL (version 1.5.2)
Driver: Hive JDBC (version 1.1.0-cdh5.7.2)
Transaction isolation: TRANSACTION_REPEATABLE_READ
Beeline version 1.1.0-cdh5.7.2 by Apache Hive
0: jdbc:hive2://t162:10001/> 　

3、权限测试

　　不同的用户通过kinit使用自己的Principal+密码通过Kerberos的AS认证拿到TGT，就可以登录到spark sql thriftserver上去查看库、表；

　　不过由于sts还不支持sqlbased authorization，所以还只能做到底层hdfs的权限隔离，比较可惜；相对来说hive的完整度高一些，支持SQLstandard authorization。

　　因为事先我们已经开启了HDFS ACL与Sentry的权限同步功能，所以spark sql jdbc 的用户权限通过hive2的权限设置来实现。即先jdbc登录hive2 ，再利用hive sql语句进行用户权限设置，然后表和数据库的权限会同步到对应的HDFS目录和文件，从而实现spark sql thriftserver基于底层hdfs的用户权限隔离。

　　如下所示,user1对test库的table1表有权限，对test库的table2表无权限，读table2表时显示无hdfs权限，即权限设置成功！

0: jdbc:hive2://node1:10000/> select * from test.table1 limit 1;
+--------------+-------------+---------------------+----------+-----------+----------+---------------------------+-----------+-----------+------------------------+------------+---------------+-------------+--+
|     cint     |   cbigint   |       cfloat        | cdouble  | cdecimal  | cstring  |         cvarchar          | cboolean  | ctinyint  |       ctimestamp       | csmallint  |     cipv4     |    cdate    |
+--------------+-------------+---------------------+----------+-----------+----------+---------------------------+-----------+-----------+------------------------+------------+---------------+-------------+--+
| 15000000001  | 1459107060  | 1.8990000486373901  | 1.7884   | 1.92482   | 中文测试1    | /browser/addBasicInfo.do  | true      | -127      | 2014-05-14 00:53:21.0  | -63        | 0  | 2014-05-14  |
+--------------+-------------+---------------------+----------+-----------+----------+---------------------------+-----------+-----------+------------------------+------------+---------------+-------------+--+
1 row selected (3.165 seconds)
0: jdbc:hive2://node1:10000/> select * from test.table2 limit 10;
Error: org.apache.hadoop.security.AccessControlException: Permission denied: user=user1, access=READ_EXECUTE, inode="/user/hive/warehouse/test.db/table2":hive:hive:drwxrwx--x
	at org.apache.hadoop.hdfs.server.namenode.DefaultAuthorizationProvider.checkAccessAcl(DefaultAuthorizationProvider.java:365)
	at org.apache.hadoop.hdfs.server.namenode.DefaultAuthorizationProvider.check(DefaultAuthorizationProvider.java:258)
	at org.apache.hadoop.hdfs.server.namenode.DefaultAuthorizationProvider.checkPermission(DefaultAuthorizationProvider.java:175)
	at org.apache.sentry.hdfs.SentryAuthorizationProvider.checkPermission(SentryAuthorizationProvider.java:178)
	at org.apache.hadoop.hdfs.server.namenode.FSPermissionChecker.checkPermission(FSPermissionChecker.java:152)
	at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.checkPermission(FSNamesystem.java:6617)
	at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.checkPermission(FSNamesystem.java:6599)
	at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.checkPathAccess(FSNamesystem.java:6524)
	at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.getListingInt(FSNamesystem.java:5061)
	at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.getListing(FSNamesystem.java:5022)
	at org.apache.hadoop.hdfs.server.namenode.NameNodeRpcServer.getListing(NameNodeRpcServer.java:882)
	at org.apache.hadoop.hdfs.server.namenode.AuthorizationProviderProxyClientProtocol.getListing(AuthorizationProviderProxyClientProtocol.java:335)
	at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolServerSideTranslatorPB.getListing(ClientNamenodeProtocolServerSideTranslatorPB.java:615)
	at org.apache.hadoop.hdfs.protocol.proto.ClientNamenodeProtocolProtos$ClientNamenodeProtocol$2.callBlockingMethod(ClientNamenodeProtocolProtos.java)
	at org.apache.hadoop.ipc.ProtobufRpcEngine$Server$ProtoBufRpcInvoker.call(ProtobufRpcEngine.java:617)
	at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:1073)
	at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2086)
	at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2082)
	at java.security.AccessController.doPrivileged(Native Method)
	at javax.security.auth.Subject.doAs(Subject.java:415)
	at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1693)
	at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2080) (state=,code=0)

　

　权限测试可参考之前博客：http://www.cnblogs.com/xiaodf/p/5968282.html，此处略

4、问题

4.1 问题1

　　使用spark1.6.0版本，启动thrift server服务后，执行“show databases”报如下错误：

FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.DDLTask. java.lang.RuntimeException: 

Unable to instantiate org.apache.hadoop.hive.ql.metadata.SessionHiveMetaStoreClient　　

　　查询资料说这可能是1.6版本的一个bug，换成1.5.2版本后，没有这个问题了。下面为此问题的查询链接：https://forums.databricks.com/questions/7207/spark-thrift-server-on-kerberos-enabled-hadoophive.html

问题2

Spark SQL ThriftServer服务启动7天后，用户在用beeline命令去连接服务报错连不上了。

服务日志报一下错误：

17/01/18 13:46:08 INFO HiveMetaStore.audit: ugi=hive/t162@HADOOP.COM	ip=unknown-ip-addr	cmd=Metastore shutdown complete.
17/01/18 13:46:08 WARN security.UserGroupInformation: Not attempting to re-login since the last re-login was attempted less than 7 days before.
17/01/18 13:46:08 WARN security.UserGroupInformation: Not attempting to re-login since the last re-login was attempted less than 7 days before.
17/01/18 13:46:08 WARN security.UserGroupInformation: Not attempting to re-login since the last re-login was attempted less than 7 days before.
17/01/18 13:46:08 WARN security.UserGroupInformation: Not attempting to re-login since the last re-login was attempted less than 7 days before.
17/01/18 13:46:09 WARN security.UserGroupInformation: Not attempting to re-login since the last re-login was attempted less than 7 days before.
17/01/18 13:46:12 WARN security.UserGroupInformation: Not attempting to re-login since the last re-login was attempted less than 7 days before.
17/01/18 13:46:17 WARN security.UserGroupInformation: Not attempting to re-login since the last re-login was attempted less than 7 days before.
17/01/18 13:46:19 WARN security.UserGroupInformation: Not attempting to re-login since the last re-login was attempted less than 7 days before.
17/01/18 13:46:19 WARN ipc.Client: Couldn't setup connection for hive/t162@HADOOP.COM to t162/t161:8020
17/01/18 13:46:19 WARN thrift.ThriftCLIService: Error opening session:
org.apache.hive.service.cli.HiveSQLException: Failed to open new session: java.lang.RuntimeException: java.lang.RuntimeException: java.io.IOException: 
Failed on local exception: java.io.IOException: Couldn't setup connection for hive/t162@HADOOP.COM to t162/t161:8020; Host Details :
 local host is: "t162/t161"; destination host is: "t162":8020;
	at org.apache.hive.service.cli.session.SessionManager.openSession(SessionManager.java:264)
	at org.apache.spark.sql.hive.thriftserver.SparkSQLSessionMa

　　

原因：创建kerberos库时我们设置了principal的认证有效期和最大renew时间，如下/etc/krb5.conf文件内容所示：

[libdefaults]
 default_realm = HADOOP.COM
 dns_lookup_realm = false
 dns_lookup_kdc = false
 ticket_lifetime = 24h
 renew_lifetime = 7d
 forwardable = true
 renewable=true

　　

7天后认证无法renew导致服务认证失败，用户连不上服务了。未解决这个问题我们需要定时重新kinit下服务principal，我们对服务启动脚本进行一下修改，添加定时认证脚本，如下所示：

#!/bin/sh
#start Spark-thriftserver
export YARN_CONF_DIR=/etc/hadoop/conf
file="hive-site.xml"
dir=$(pwd)
cd conf/
if [ ! -e "$file" ]
then
 cp /etc/hive/conf/hive-site.xml $dir/conf/
fi
cd ../sbin
./start-thriftserver.sh --name SparkJDBC --master yarn-client  --num-executors 10 --executor-memory 2g --executor-cores 4 --driver-memory 10g --driver-cores 2  --conf spark.storage.memoryFraction=0.2 --conf spark.shuffle.memoryFraction=0.6 --hiveconf hive.server2.thrift.port=10001 --hiveconf hive.server2.logging.operation.enabled=true --hiveconf hive.server2.authentication.kerberos.principal=hive/t162@HADOOP.COM --hiveconf hive.server2.authentication.kerberos.keytab=/home/hive.keytab
while(true)
do
kinit -kt /home/hive.keytab hive/t162@HADOOP.COM
sleep 6*24h
done &　　

经测试，问题解决！