Configuring the Sentry Service

Enabling the Sentry Service Using Cloudera Manager

Before Enabling the Sentry Service

  • Verify the prerequisites for the Sentry service: Before You Install Sentry
  • Setting Hive Warehouse Directory Permissions

    The Hive warehouse directory (/user/hive/warehouse or any path you specify as hive.metastore.warehouse.dir in your hive-site.xml) must be owned by the Hive user and group.
    • Using the default Hive warehouse directory - Permissions on the warehouse directory must be set as follows (see following Note for caveats):
      • 771 on the directory itself (by default, /user/hive/warehouse)
      • 771 on all subdirectories (for example, /user/hive/warehouse/mysubdir)
      • All files and subdirectories should be owned by hive:hive
      For example:
      $ sudo -u hdfs hdfs dfs -chmod -R 771 /user/hive/warehouse
      $ sudo -u hdfs hdfs dfs -chown -R hive:hive /user/hive/warehouse
      If you have enabled Kerberos on your cluster, you must kinit as the hdfs user before you set permissions. For example:
      sudo -u hdfs kinit -kt <hdfs.keytab> hdfs
      sudo -u hdfs hdfs dfs -chmod -R 771 /user/hive/warehouse
      $ sudo -u hdfs hdfs dfs -chown -R hive:hive /user/hive/warehouse
    • Using a non-default Hive warehouse: If you would like to use a different directory as the Hive warehouse, update the hive.metastore.warehouse.dir property, and make sure you set the required permissions on the new directory. For example, if the new warehouse directory is /data, set the permissions as follows:
      $ hdfs dfs -chown hive:hive /data
      $ hdfs dfs -chmod 771 /data

      Note that when you update the default Hive warehouse, previously created tables will not be moved over automatically. Therefore, tables created before the update will remain at /user/hive/warehouse/<old_table>. However, after the update, any new tables created in the default location will be found at /data/<new_table>.

      For Sentry/HDFS sync to work as expected, add the new warehouse URL to the list of Sentry Synchronization Path Prefixes.

  • Disable impersonation for HiveServer2 in the Cloudera Manager Admin Console. Impersonation in HiveServer2 allows users to execute queries and access HDFS files as the connected user rather than the super user who started the HiveServer2 daemon. This enforces an access control policy at the file level using HDFS file permissions or ACLs. Keeping impersonation enabled means Sentry does not have end-to-end control over the authorization process. While Sentry can enforce access control policies on tables and views in the Hive warehouse, it has no control over permissions on the underlying table files in HDFS. Hence, even if users do not have the Sentry privileges required to access a table in the warehouse, as long as they have permission to access the corresponding table file in HDFS, any jobs or queries submitted will bypass Sentry authorization checks and execute successfully. Use the following instructions to disable impersonation:
    1. Go to the Hive service.
    2. Click the Configuration tab.
    3. Select Scope > HiveServer2.
    4. Select Category > Main.
    5. Uncheck the HiveServer2 Enable Impersonation checkbox.
    6. Click Save Changes to commit the changes.
  • If you are using MapReduce, enable the Hive user to submit MapReduce jobs.
    1. Open the Cloudera Manager Admin Console and go to the MapReduce service.
    2. Click the Configuration tab.
    3. Select Scope > TaskTracker.
    4. Select Category > Security.
    5. Set the Minimum User ID for Job Submission property to zero (the default is 1000).
    6. Click Save Changes to commit the changes.
    7. Repeat steps 1-6 for every TaskTracker role group for the MapReduce service that is associated with Hive.
    8. Restart the MapReduce service.
  • If you are using YARN, enable the Hive user to submit YARN jobs.
    1. Open the Cloudera Manager Admin Console and go to the YARN service.
    2. Click the Configuration tab.
    3. Select Scope > NodeManager.
    4. Select Category > Security.
    5. Ensure the Allowed System Users property includes the hive user. If not, add hive.
    6. Click Save Changes to commit the changes.
    7. Repeat steps 1-6 for every NodeManager role group for the YARN service that is associated with Hive.
    8. Restart the YARN service.
  • Block the Hive CLI user from accessing the Hive metastore:

    1. In the Cloudera Manager Admin Console, select the Hive service.
    2. On the Hive service page, click the Configuration tab.
    3. In the search field, search for Hive Metastore Access Control and Proxy User Groups Override to locate the hadoop.proxyuser.hive.groups setting.
    4. Click the plus sign three times to add the following groups:
      • hive
      • hue
      • sentry
    5. Click Save Changes.

    Setting this parameter blocks access to the Hive metastore for the user running the Hive CLI if they are not part of the hive, hue, or sentry groups. The Hive CLI can still run, but after setting this parameter as described here, the hive user can impersonate only members of the hive, sentry, or hue groups. If you are using Sqoop, the Sqoop user must also have access to the Hive metastore.

Enabling the Sentry Service for Hive

  1. Go to the Hive service.
  2. Click the Configuration tab.
  3. Select Scope > Hive (Service-Wide).
  4. Select Category > Main.
  5. Locate the Sentry Service property and select Sentry.
  6. Locate the Enable Stored Notifications in Database property and select it.
  7. Click Save Changes to commit the changes.
  8. Restart the Hive service.
Enabling Sentry on Hive service places several HiveServer2 properties on a restricted list properties that cannot be modified at runtime by clients. See HiveServer2 Restricted Properties.

Enabling the Sentry Service for Impala

  1. Enable the Sentry service for Hive (as instructed above).
  2. Go to the Impala service.
  3. Click the Configuration tab.
  4. Select Scope > Impala (Service-Wide).
  5. Select Category > Main.
  6. Locate the Sentry Service property and select Sentry.
  7. Click Save Changes to commit the changes.
  8. Restart Impala.

Enabling the Sentry Service for Solr

Enable the Sentry service as follows:
  1. Go to the Solr service.
  2. Click the Configuration tab.
  3. Select Scope > Solr (Service-Wide).
  4. Select Category > Main.
  5. Locate the Sentry Service property and select Sentry.
  6. Click Save Changes to commit the changes.
  7. Restart Solr.

After enabling Sentry for Solr, you may want to configure authorization as described in Configuring Sentry Authorization for Cloudera Search.

Enabling the Sentry Service for Hue

Hue uses a Security app to make it easier to interact with Sentry. When you set up Hue to manage Sentry permissions, make sure that users and groups are set up correctly. Every Hue user connecting to Sentry must have an equivalent OS-level user account on all hosts so that Sentry can authenticate Hue users. Each OS-level user should also be part of an OS-level group with the same name as the corresponding user's group in Hue.

For more information on using the Security app, see the related blog post.

Enable the Sentry service as follows:
  1. Enable the Sentry service for Hive and Impala (as instructed above).
  2. Go to the Hue service.
  3. Click the Configuration tab.
  4. Select Scope > Hue (Service-Wide).
  5. Select Category > Main.
  6. Locate the Sentry Service property and select Sentry.
  7. Click Save Changes to commit the changes.
  8. Restart Hue.

Add the Hive, Impala and Hue Groups to Sentry's Admin Groups

  1. Go to the Sentry service.
  2. Click the Configuration tab.
  3. Select Scope > Sentry (Service-Wide).
  4. Select Category > Main.
  5. Locate the Admin Groups property and add the hive, impala and hue groups to the list. If an end user is in one of these admin groups, that user has administrative privileges on the Sentry Server.
  6. Click Save Changes to commit the changes.

Enabling the Sentry Service Using the Command Line

Before Enabling the Sentry Service

  • Setting Hive Warehouse Directory Permissions

    The Hive warehouse directory (/user/hive/warehouse or any path you specify as hive.metastore.warehouse.dir in your hive-site.xml) must be owned by the Hive user and group.
    • Using the default Hive warehouse directory - Permissions on the warehouse directory must be set as follows (see following Note for caveats):
      • 771 on the directory itself (by default, /user/hive/warehouse)
      • 771 on all subdirectories (for example, /user/hive/warehouse/mysubdir)
      • All files and subdirectories should be owned by hive:hive
      For example:
      $ sudo -u hdfs hdfs dfs -chmod -R 771 /user/hive/warehouse
      $ sudo -u hdfs hdfs dfs -chown -R hive:hive /user/hive/warehouse
      If you have enabled Kerberos on your cluster, you must kinit as the hdfs user before you set permissions. For example:
      sudo -u hdfs kinit -kt <hdfs.keytab> hdfs
      sudo -u hdfs hdfs dfs -chmod -R 771 /user/hive/warehouse
      $ sudo -u hdfs hdfs dfs -chown -R hive:hive /user/hive/warehouse
    • Using a non-default Hive warehouse: If you would like to use a different directory as the Hive warehouse, update the hive.metastore.warehouse.dir property, and make sure you set the required permissions on the new directory. For example, if the new warehouse directory is /data, set the permissions as follows:
      $ hdfs dfs -chown hive:hive /data
      $ hdfs dfs -chmod 771 /data

      Note that when you update the default Hive warehouse, previously created tables will not be moved over automatically. Therefore, tables created before the update will remain at /user/hive/warehouse/<old_table>. However, after the update, any new tables created in the default location will be found at /data/<new_table>.

      For Sentry/HDFS sync to work as expected, add the new warehouse URL to the list of Sentry Synchronization Path Prefixes.

  • HiveServer2 impersonation must be turned off. Impersonation in HiveServer2 allows users to execute queries and access HDFS files as the connected user rather than the super user who started the HiveServer2 daemon. This enforces an access control policy at the file level using HDFS file permissions or ACLs. Keeping impersonation enabled means Sentry does not have end-to-end control over the authorization process. While Sentry can enforce access control policies on tables and views in the Hive warehouse, it has no control over permissions on the underlying table files in HDFS. Hence, even if users do not have the Sentry privileges required to access a table in the warehouse, as long as they have permission to access the corresponding table file in HDFS, any jobs or queries submitted will bypass Sentry authorization checks and execute successfully.
  • If you are using MapReduce, you must enable the Hive user to submit MapReduce jobs. You can ensure that this is true by setting the minimum user ID for job submission to 0. Edit the taskcontroller.cfg file and set min.user.id=0.
    If you are using YARN, you must enable the Hive user to submit YARN jobs, add the user hive to the allowed.system.users configuration property. Edit the container-executor.cfg file and add hive to the allowed.system.users property. For example,
    allowed.system.users = nobody,impala,hive,hbase
  • Block the Hive CLI user from accessing the Hive metastore by setting the following property in the cluster's core-site.xml file:

    <property>
         <name>hadoop.proxyuser.hive.groups</name>
         <value>hive,hue</value>
         <description>Sets groups from which the hive user can impersonate other users.</description>
    </property>
                

    Setting this parameter blocks access to the Hive metastore for the user running the Hive CLI if they are not part of the hive or the hue groups. The Hive CLI can still run, but after setting this parameter as described here, the hive user can impersonate only members of the hive or the hue groups. If you are using Sqoop, the Sqoop user must also have access to the Hive metastore.

  • Add the hive, impala and hue groups to Sentry's sentry.service.admin.group in the sentry-site.xml file. If an end user is in one of these admin groups, that user has administrative privileges on the Sentry Server.
    <property>
      <name>sentry.service.admin.group</name>
      <value>hive,impala,hue</value>
     </property>

Configuring the Sentry Server

Configure the following properties in sentry-site.xml on the Sentry Server host.
<property>
    <name>sentry.service.server.rpc-address</name>
    <value>nightly54-1.gce.cloudera.com</value>
</property>
  <property>
    <name>sentry.service.server.rpc-port</name>
    <value>8038</value>
</property>
<property>
    <name>sentry.service.admin.group</name>
    <value>hive,impala,hue</value>
</property>
<property>
    <name>sentry.service.allow.connect</name>
    <value>hive,impala,hue,hdfs</value>
</property>
<property>
    <name>sentry.store.group.mapping</name>
    <value>org.apache.sentry.provider.common.HadoopGroupMappingService</value>
</property>
<property>
    <name>sentry.service.server.principal</name>
    <value>sentry/_HOST@GCE.CLOUDERA.COM</value>
</property>
<property>
    <name>sentry.service.security.mode</name>
    <value>kerberos</value>
</property>
<property>
    <name>sentry.service.server.keytab</name>
    <value>sentry.keytab</value>
</property>
<property>
    <name>sentry.store.jdbc.url</name>
    <value>jdbc:<JDBC connection URL for backend database></value>
</property>
<property>
    <name>sentry.store.jdbc.driver</name>
    <value><JDBC Driver class for backend database></value>
</property>
<property>
    <name>sentry.store.jdbc.user</name>
    <value><User ID for backend database user></value>
</property>
<property>
    <name>sentry.store.jdbc.password</name>
    <value><Password for backend database user></value>
</property>
<property>
    <name>sentry.service.processor.factories</name>
    <value>org.apache.sentry.provider.db.service.thrift.SentryPolicyStoreProcessorFactory, 
             org.apache.sentry.hdfs.SentryHDFSServiceProcessorFactory</value>
</property>
<property>
    <name>sentry.policy.store.plugins</name>
    <value>org.apache.sentry.hdfs.SentryPlugin</value>
</property>
<property>
    <name>sentry.hdfs.integration.path.prefixes</name>
    <value>/user/hive/warehouse</value>
</property>

Configuring HiveServer2 for the Sentry Service

Configure the following properties in sentry-site.xml on the HiveServer2 host.
<property>
    <name>hive.sentry.server</name>
    <value>server1</value>
</property>
<property>
    <name>sentry.service.server.principal</name>
    <value>sentry/_HOST@EXAMPLE.COM</value>
</property>
<property>
    <name>sentry.service.security.mode</name>
    <value>kerberos</value>
</property>
<property>
    <name>sentry.hive.provider.backend</name>
    <value>org.apache.sentry.provider.db.SimpleDBProviderBackend</value>
</property>
<property>
    <name>sentry.service.client.server.rpc-address</name>
    <value>example.cloudera.com</value>
</property>
<property>
    <name>sentry.service.client.server.rpc-port</name>
    <value>8038</value>
</property>
<property>
    <name>hive.sentry.provider</name>
    <value>org.apache.sentry.provider.file.HadoopGroupResourceAuthorizationProvider</value>
</property>
<property>
    <name>hive.sentry.failure.hooks</name>
    <value>com.cloudera.navigator.audit.hive.HiveSentryOnFailureHook</value>
</property>
Add the following properties to hive-site.xml to allow the Hive service to communicate with the Sentry service.
<property>
   <name>hive.security.authorization.task.factory</name>
   <value>org.apache.sentry.binding.hive.SentryHiveAuthorizationTaskFactoryImpl</value>
</property>
<property>
   <name>hive.server2.session.hook</name>
   <value>org.apache.sentry.binding.hive.HiveAuthzBindingSessionHook</value>
</property>
<property>
   <name>hive.sentry.conf.url</name>
   <value>file:///{{PATH/TO/DIR}}/sentry-site.xml</value>
</property>
<property>
   <name>hive.security.authorization.task.factory</name>
   <value>org.apache.sentry.binding.hive.SentryHiveAuthorizationTaskFactoryImpl</value>
</property>
Enabling Sentry on Hive service places several HiveServer2 properties on a restricted list properties that cannot be modified at runtime by clients. See HiveServer2 Restricted Properties.

Configuring the Hive Metastore for the Sentry Service

Add the following properties to hive-site.xml to allow the Hive metastore to communicate with the Sentry service.
<property>
<name>hive.metastore.filter.hook</name>
<value>org.apache.sentry.binding.metastore.SentryMetaStoreFilterHook</value>
</property>

<property>  
    <name>hive.metastore.pre.event.listeners</name>  
    <value>org.apache.sentry.binding.metastore.MetastoreAuthzBinding</value>  
    <description>list of comma separated listeners for metastore events.</description>
</property>

<property>
    <name>hive.metastore.event.listeners</name>  
    <value>org.apache.sentry.binding.metastore.SentryMetastorePostEventListener</value>  
    <description>list of comma separated listeners for metastore, post events.</description>
</property>

Configuring Impala as a Client for the Sentry Service

Set the following configuration properties in ../impala-conf/sentry-site.xml on the Catalog Server.
<property>
   <name>sentry.service.client.server.rpc-port</name>
   <value>8038</value>
</property>
<property>
   <name>sentry.service.client.server.rpc-address</name>
   <value>hostname</value>
</property>
<property>
   <name>sentry.service.client.server.rpc-connection-timeout</name>
   <value>200000</value>
</property>
<property>
   <name>sentry.service.security.mode</name>
   <value>kerberos</value>
</property>
You must also add the following configuration properties to Impala's /etc/default/impala file. For more information , see Configuring Impala Startup Options through the Command Line.
  • On the catalogd and the impalad.
    --sentry_config=<absolute path to sentry service configuration file>
  • On the impalad.
    --server_name=<server name>
    If the --authorization_policy_file flag is set, Impala will use the policy file-based approach. Otherwise, the database-backed approach will be used to implement authorization.

Enabling Solr as a Client for the Sentry Service Using the Command Line

You can enable Sentry using Cloudera Manager or by manually modifying files. For more information on enabling Sentry using Cloudera Manager, see Configuring Sentry Policy File Authorization Using Cloudera Manager and Enabling Sentry Policy File Authorization for Solr.

Sentry is enabled with addition of two properties to /etc/default/solr or /opt/cloudera/parcels/CDH-*/etc/default/solr.
  • In a Cloudera Manager deployment, required properties are added automatically when you click Enable Sentry Authorization in the Solr configuration page in Cloudera Manager.
  • If you are using configs, you must configure the proper config=myConfig permissions as described in Using Roles and Privileges with Sentry.
  • In a deployment not managed by Cloudera Manager, you must make these changes yourself. The variable SOLR_AUTHORIZATION_SENTRY_SITE specifies the path to sentry-site.xml. The variable SOLR_AUTHORIZATION_SUPERUSER specifies the first part of SOLR_KERBEROS_PRINCIPAL. This is solr for the majority of users, as solr is the default. Settings are of the form:
    SOLR_AUTHORIZATION_SENTRY_SITE=/location/to/sentry-site.xml
    SOLR_AUTHORIZATION_SUPERUSER=solr

To enable Sentry collection-level authorization checking on a new collection, the instancedir for the collection must use a modified version of solrconfig.xml with Sentry integration. Each collection has a separate solrconfig.xml file, meaning you can define different behavior for each collection. The command solrctl instancedir --generate generates two versions of solrconfig.xml: the standard solrconfig.xml without sentry integration and the sentry-integrated version called solrconfig.xml.secure. To use the sentry-integrated version, replace solrconfig.xml with solrconfig.xml.secure before creating the instancedir.

You can enable Sentry on an existing collection. The process varies depending on whether you are using a config or instancedir.

Enabling Sentry on Collections using configs

If you have a collection that is using a non-secured config, you can enable Sentry security on that collection by modifying the collection to use a secure config. The config in use must not be immutable, otherwise it cannot be changed. To update an existing non-immutable config:

  1. Delete the existing config using the solrctl config --delete command. For example:
    solrctl config --delete myManaged
  2. Create a new non-immutable config using the solrctl config --create command. Use a sentry-enabled template such as managedTemplateSecure. The new config must have the same name as the config being replaced. For example:
    solrctl config --create myManaged managedTemplateSecure -p immutable=false
  3. Reload the collection using to solrctl collection --reload command.
    solrctl collection --reload myCollection

For a list of all available config templates, see Config Templates.

Enabling Sentry on Collections using instancedirs

If you have a collection that is using a non-secured instancedir configuration, you can enable Sentry security on that collection by modifying the settings that are stored in instancedir. For example, you might have an existing collection named foo and a standard solrconfig.xml. By default, collections are stored in instancedirs that use the collection's name, which is foo in this case.

If your collection uses an unmodified solrconfig.xml file, you can enable Sentry by replacing the existing solrconfig.xml file. If your collection uses a solrconfig.xml that contains modifications you want to preserve, you can attempt to use a difftool to find an integrate changes in to the secure template.

To enable Sentry on an existing collection without preserving customizations

# generate a fresh instancedir
solrctl instancedir --generate foosecure
# download the existing instancedir from ZK into subdirectory foo
solrctl instancedir --get foo foo
# replace the existing solrconfig.xml with the sentry-enabled one
cp foosecure/conf/solrconfig.xml.secure foo/conf/solrconfig.xml
# update the instancedir in ZK
solrctl instancedir --update foo foo
# reload the collection
solrctl collection --reload foo

To enable Sentry on an existing collection and preserve customizations

Generate a new instancedir, compare the differences between the default solrconfig.xml and solrconfig.xml.secure files, and then add the elements that are unique to solrconfig.xml.secure to the file that your environment is using.

  1. Generate a fresh instancedir:
    solrctl instancedir --generate foo
  2. Compare the solrconfig.xml and solrconfig.xml.secure:
    diff foo/conf/solrconfig.xml foo/conf/solrconfig.xml.secure
  3. Add the elements that are unique to solrconfig.xml.secure to your existing solrconfig.xml file. You might complete this process by manually editing your existing solrconfig.xml file or by using a merge tool.
  4. Reload the collection:
    solrctl collection --reload foo

After enabling Sentry for Solr, you may want to configure authorization as described in Configuring Sentry Authorization for Cloudera Search.

HiveServer2 Restricted Properties

Enabling Sentry on Hive service places several HiveServer2 properties on a restricted list properties that cannot be modified at runtime by clients. This list is denoted by the hive.conf.restricted.list property and these properties are only configurable on the server side. The list includes:
hive.enable.spark.execution.engine
hive.semantic.analyzer.hook
hive.exec.pre.hooks
hive.exec.scratchdir
hive.exec.local.scratchdir
hive.metastore.uris,
javax.jdo.option.ConnectionURL
hadoop.bin.path
hive.session.id
hive.aux.jars.path
hive.stats.dbconnectionstring
hive.scratch.dir.permission
hive.security.command.whitelist
hive.security.authorization.task.factory
hive.entity.capture.transform
hive.access.conf.url
hive.sentry.conf.url
hive.access.subject.name
hive.sentry.subject.name
hive.sentry.active.role.set

Configuring Pig and HCatalog for the Sentry Service

Once you have the Sentry service up and running, and Hive has been configured to use the Sentry service, there are some configuration changes you must make to your cluster to allow Pig, MapReduce (using HCatLoader, HCatStorer) and WebHCat queries to access Sentry-secured data stored in Hive.

Since the Hive warehouse directory is owned by hive:hive, with its permissions set to 771, with these settings, other user requests such as commands coming through Pig jobs, WebHCat queries, and MapReduce jobs, may fail. To give these users access, perform the following configuration changes:
  • Use HDFS ACLs to define permissions on a specific directory or file of HDFS. This directory/file is generally mapped to a database, table, partition, or a data file.
  • Users running these jobs should have the required permissions in Sentry to add new metadata or read metadata from the Hive Metastore Server. For instructions on how to set up the required permissions, see Hive SQL Syntax for Use with Sentry. You can use HiveServer2's command line interface, Beeline to update the Sentry database with the user privileges.
Examples:
  • A user who is using Pig HCatLoader will require read permissions on a specific table or partition. In such a case, you can GRANT read access to the user in Sentry and set the ACL to read and run, on the file being accessed.
  • A user who is using Pig HCatStorer will require ALL permissions on a specific table. In this case, you GRANT ALL access to the user in Sentry and set the ACL to write and run on the table being used.

Securing the Hive Metastore

It's important that the Hive metastore be secured. If you want to override the Kerberos prerequisite for the Hive metastore, set the sentry.hive.testing.mode property to true to allow Sentry to work with weaker authentication mechanisms. Add the following property to the HiveServer2 and Hive metastore's sentry-site.xml:
<property>
  <name>sentry.hive.testing.mode</name>
  <value>true</value>
</property>
Impala does not require this flag to be set.

You can turn on Hive metastore security using the instructions in Cloudera Security. To secure the Hive metastore; see Hive Metastore Server Security Configuration.

Using User-Defined Functions with HiveServer2

The ADD JAR command does not work with HiveServer2 and the Beeline client when Beeline runs on a different host. As an alternative to ADD JAR, Hive's auxiliary paths functionality should be used. There are some differences in the procedures for creating permanent functions and temporary functions when Sentry is enabled. For detailed instructions, see: .