Is Zeppelin in HDP 2.5 support multi-tenancy on a Kerberized Cluster

[#1]

We are looking at setting up Zeppelin on top of Livy Server. Does the following settings pass also the kerberos authentication information.

<property>
    <name>hadoop.proxyuser.livy.groups</name>
    <value>*</value>
</property>

<property>
    <name>hadoop.proxyuser.livy.hosts</name>
    <value>*</value>
</property>

https://docs.hortonworks.com/HDPDocuments/HDP2/HDP-2.5.0/bk_zeppelin-component-guide/content/install...

[#2 · Accepted solution]

Hi Deepak, see my how-to tutorial:

https://community.hortonworks.com/content/kbentry/65449/ow-to-setup-a-multi-user-active-directory-ba...

If you are using self-signed certificate, Download the SSL certificate to where zeppelin is running

1. <code>mkdir -p /etc/security/certificates

store the certificate in this directory

Import certificate for zeppelin to work with the self signed certificate.

1. <code>cd /etc/security/certificates
2. keytool -import -alias sampledcfieldcloud -file ad01.your.domain.name.cer -keystore /usr/jdk64/jdk1.8.0_77/jre/lib/security/cacerts
3. keytool -list -v -keystore /usr/jdk64/jdk1.8.0_77/jre/lib/security/cacerts | grep sampledcfieldcloud

Create home directory in hdfs for the user that you will login:

1. <code>hdfs dfs -mkdir /user/hadoopadmin
2. hdfs dfs -chown hadoopadmin:hdfs /user/hadoopadmin

Enable multi-user zeppelin use ambari -> zeppelin notebook configs

expand the Advanced zeppelin-env and look for shiro.ini entry. Below is configuration that works with our sampledcfield Cloud.

1. <code>[users]
2. # List of users with their password allowed to access Zeppelin.
3. # To use a different strategy (LDAP / Database / ...) check the shiro doc at http://shiro.apache.org/configuration.html#Configuration-INISections
4. #admin = password1
5. #user1 = password2, role1, role2
6. #user2 = password3, role3
7. #user3 = password4, role2
8. # Sample LDAP configuration, for user Authentication, currently tested for single Realm
9. [main]
10. activeDirectoryRealm = org.apache.zeppelin.server.ActiveDirectoryGroupRealm
11. #activeDirectoryRealm.systemUsername = CN=binduser,OU=ServiceUsers,DC=sampledcfield,DC=hortonworks,DC=com
12. activeDirectoryRealm.systemUsername = binduser
13. activeDirectoryRealm.systemPassword = xxxxxx
14. activeDirectoryRealm.principalSuffix = @your.domain.name
15. #activeDirectoryRealm.hadoopSecurityCredentialPath = jceks://user/zeppelin/zeppelin.jceks
16. activeDirectoryRealm.searchBase = DC=sampledcfield,DC=hortonworks,DC=com
17. activeDirectoryRealm.url = ldaps://ad01.your.domain.name:636
18. activeDirectoryRealm.groupRolesMap = "CN=hadoop-admins,OU=CorpUsers,DC=sampledcfield,DC=hortonworks,DC=com":"admin"
19. activeDirectoryRealm.authorizationCachingEnabled = true
20. sessionManager = org.apache.shiro.web.session.mgt.DefaultWebSessionManager
21. cacheManager = org.apache.shiro.cache.MemoryConstrainedCacheManager
22. securityManager.cacheManager = $cacheManager
23. securityManager.sessionManager = $sessionManager
24. securityManager.sessionManager.globalSessionTimeout = 86400000
25. #ldapRealm = org.apache.shiro.realm.ldap.JndiLdapRealm
26. #ldapRealm.userDnTemplate = uid={0},cn=users,cn=accounts,dc=example,dc=com
27. #ldapRealm.contextFactory.url = ldap://ldaphost:389
28. #ldapRealm.contextFactory.authenticationMechanism = SIMPLE
29. #sessionManager = org.apache.shiro.web.session.mgt.DefaultWebSessionManager
30. #securityManager.sessionManager = $sessionManager
31. # 86,400,000 milliseconds = 24 hour
32. #securityManager.sessionManager.globalSessionTimeout = 86400000
33. shiro.loginUrl = /api/login
34. [roles]
35. admin = *
36. [urls]
37. # anon means the access is anonymous.
38. # authcBasic means Basic Auth Security
39. # To enfore security, comment the line below and uncomment the next one
40. /api/version = anon
41. /api/interpreter/** = authc, roles[admin]
42. /api/credential/** = authc, roles[admin]
43. /api/configurations/** = authc, roles[admin]
44. #/** = anon
45. /** = authc
46. #/** = authcBasic

Grant Livy ability to impersonate

Use Ambari to update core-site.xml, restart YARN & HDFS after making this change.

1. <code><property>
2. <name>hadoop.proxyuser.livy.groups</name>
3. <value>*</value>
4. </property>
5. <property>
6. <name>hadoop.proxyuser.livy.hosts</name>
7. <value>*</value>
8. </property>

Restart hdfs and yarn after this update.

After running the livy notebook make sure the yarn logs show the logged in user as the user that is running, hadoopadmin is the user that is logged in the zeppelin notebook. You should see 2 applications running the livy-session-X and the zeppelin app running in yarn

1. <code>application_1478287338271_0003 hadoopadmin livy-session-0
2. application_1478287338271_0002 zeppelin Zeppelin

[#3]

Thanks @azeltov. To confirm. Does it also work with a Kerberized Cluster ?I am just wondering how the kerberos information is passed.

[#4]

It does work in Kerberized cluster you will need to create keytabs for zeppelin and livy service account.

[#5]

Thanks @azeltov . Even if we create a kerberos token for zeppelin, how does the kerberos tokens for individual users is passed ? All the access to HDFS, Spark and Hive is managed in Ranger for AD user or group and not for Zeppelin user.

[#6]

You must use Livy integration for the user tokens to be passed.

[#7]

The tokens aren't passed. zeppelin authenticates itself with livy and as it is a superuser (livy.superusers) livy takes the proxyUser sent by zeppelin and becomes that user.