集成kerberos后，很多使用服务的程序代码需要改写，例如java通过jdbc链接impala；java扫描hbase表；java的kafka客户端.......，除了 spark程序以外。
下面介绍下集成kerberos后，要做什么准备才能让spark程序正常跑起来
在实施方案前，假设读者已经基本熟悉以下技术 (不细说)
- 熟悉spark程序，spark-submit脚本
- cdh集成kerberos没有问题
- hbase没问题
- 这里不提供spark读hbase代码
方案实施
- cdh集成spark2，参考以下地址，不细说
https://blog.csdn.net/u010936936/article/details/73650417
-
安装完成之后能看到这个玩意
spark2.png -
spark集成kerberos
spark集成kerberos.png 重启spark2并部署客户端
添加软链接
[root@node1 spark2]# vi /etc/extra-lib/hbase/classpath.txt
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/htrace-core-3.0.4.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/htrace-core-3.2.0-incubating.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/htrace-core4-4.0.1-incubating.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/commons-pool2-2.4.3.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/jedis-2.9.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/druid-1.1.6.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-annotations-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-annotations-1.2.0-cdh5.11.0-tests.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-client-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-client.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-common-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-common-1.2.0-cdh5.11.0-tests.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-examples-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-examples.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-external-blockcache-1.2.0-cdh5.11.0.jar
"/etc/extra-lib/hbase/classpath.txt" 41L, 3584C
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/htrace-core-3.0.4.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/htrace-core-3.2.0-incubating.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/htrace-core4-4.0.1-incubating.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/commons-pool2-2.4.3.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/jedis-2.9.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/jars/druid-1.1.6.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-annotations-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-annotations-1.2.0-cdh5.11.0-tests.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-client-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-client.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-common-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-common-1.2.0-cdh5.11.0-tests.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-examples-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-examples.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-external-blockcache-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-external-blockcache.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-hadoop2-compat-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-hadoop2-compat-1.2.0-cdh5.11.0-tests.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-hadoop-compat-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-hadoop-compat-1.2.0-cdh5.11.0-tests.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-it-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-it-1.2.0-cdh5.11.0-tests.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-prefix-tree-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-procedure-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-protocol-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-resource-bundle-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-rest-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-rsgroup-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-rsgroup-1.2.0-cdh5.11.0-tests.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-server-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-server-1.2.0-cdh5.11.0-tests.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-shell-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-spark-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH-5.11.0-1.cdh5.11.0.p0.34/lib/hbase/hbase-thrift-1.2.0-cdh5.11.0.jar
/opt/cloudera/parcels/CDH/jars/ojdbc6.jar
/etc/extra-lib/hbase/spring/spring-beans-5.0.2.RELEASE.jar
/etc/extra-lib/hbase/spring/spring-core-5.0.2.RELEASE.jar
/etc/extra-lib/hbase/spring/spring-expression-5.0.2.RELEASE.jar
/etc/extra-lib/hbase/spring/spring-jcl-5.0.2.RELEASE.jar
:wq
#每个节点都要有
scp /etc/extra-lib/hbase/classpath.txt root@bi-slave1:/etc/extra-lib/hbase/
...
修改Spark2在CM内的配置：
“spark2-conf/spark-env.sh 的 Spark 2 服务高级配置代码段（安全阀）”
“spark2-conf/spark-env.sh 的 Spark 2 客户端高级配置代码段（安全阀）”
在以上两个配置框内加入：
export SPARK_DIST_CLASSPATH="(paste -sd: "/etc/extra-lib/hbase/classpath.txt")"
spark2_server.png
spark2_client.png
重启spark，并部署客户端
-
添加hbase_site.xml到$SPARK_HOME/conf
-
下载客户端代码
客户端配置_hbase.png -
解压获得hbase_site.xml文件, 并添加到$SPARK_HOME/conf里面
hbase_site.png
yarn-conf里面也需要有该xml
-
拷贝到其他节点
拷贝8.png 到这里大功告成！
spark测试
- 因为写的spark程序需要读取hbase的表，先给用户赋权限，这里以deng_yb账号为例子
- 用hbase账号登录hbase shell
[root@node1 ~]# kinit -kt /opt/cm-5.11.0/run/cloudera-scm-agent/process/1247-hbase-HBASERESTSERVER/hbase.keytab hbase/[email protected]
[root@node1 ~]# klist
Ticket cache: FILE:/tmp/krb5cc_0
Default principal: hbase/[email protected]
Valid starting Expires Service principal
06/13/18 21:20:03 06/14/18 21:20:03 krbtgt/[email protected]
renew until 06/18/18 21:20:03
[root@node1 ~]# hbase shell
.....
为deng_yb账号赋权
这里赋权U:DAY_ORG_CMP_OSI和U:DAY_ORG_PRO_CATE_SPARK读写
hbase(main):001:0> list
TABLE
U:DAY_ORG_CMP_OSI
U:DAY_ORG_PRO_CATE_SPARK
test
test_user
user_info
5 row(s) in 0.5330 seconds
=> ["U:DAY_ORG_CMP_OSI", "U:DAY_ORG_PRO_CATE_SPARK", "test", "test_user", "user_info"]
hbase(main):002:0> grant 'deng_yb','RW','U:DAY_ORG_CMP_OSI'
0 row(s) in 0.6180 seconds
hbase(main):003:0> grant 'deng_yb','RW','U:DAY_ORG_PRO_CATE_SPARK'
0 row(s) in 0.1320 seconds
- 登录deng_yb账号
[root@node1 ~]# kinit deng_yb
Password for [email protected]:
[root@node1 ~]# klist
Ticket cache: FILE:/tmp/krb5cc_0
Default principal: [email protected]
Valid starting Expires Service principal
06/13/18 21:24:18 06/14/18 21:24:18 krbtgt/[email protected]
renew until 06/20/18 21:24:18
- 查看hbase权限下的表
hbase(main):001:0> list
TABLE
U:DAY_ORG_CMP_OSI
U:DAY_ORG_PRO_CATE_SPARK
2 row(s) in 0.5380 seconds
=> ["U:DAY_ORG_CMP_OSI", "U:DAY_ORG_PRO_CATE_SPARK"]
hbase(main):002:0>
#赋权成功
- 执行spark程序 （执行的账号必须能够访问hive元数据）
spark2-submit --master yarn --deploy-mode cluster --executor-memory 4G --total-executor-cores 4 --driver-memory 4g --class com.W.Main /usr/local/W/bi-bdap-0.1.0-SNAPSHOT.jar
-
执行完之后，记得去查看AM日志
am_log.png -
AM日志没问题，再去看executors日志
executor_1.png
executor_2.png 到这里，假如日志没显示异常代表spark运行成功
问题解惑
-
部分开发遇到认证失败
异常.png
这种情况一般是因为$SPARK_CLASSPATH里面没有hbase的包
以下几个jar是必须要有的
spark_必备hbase包.png
-
连接hbase一直报超时
链接hbase超时.png
这种情况下一般是因为$SPARK_HOME/conf下缺少hbase_site.xml文件。
-
结论
- spark在集群环境跑，必须要满足
1). SPARK_HOME/conf必须要有hbase_site.xml或者程序Configuration必须addResource（hbase_site.xml）
参考 : https://issues.apache.org/jira/browse/SPARK-12279?jql=project%20%3D%20SPARK%20AND%20text%20~%20%22kerberos%22