Hive SQL Syntax for Use with Sentry

Sentry permissions can be configured through GRANT and REVOKE statements issued either interactively or programmatically through the HiveServer2 SQL command line interface, Beeline (documentation available here). The syntax described below is very similar to the GRANT and REVOKE commands that are available in well-established relational database systems.

In HUE, the Sentry Admin that creates roles and grants privileges must belong to a group that has ALL privileges on the server. For example, you can create a role for the group that contains the hive or impala user, and grant ALL ON SERVER .. WITH GRANT OPTION to that role:
CREATE ROLE <admin role>;
GRANT ALL ON SERVER <server1> TO ROLE <admin role> WITH GRANT OPTION;
GRANT ROLE <admin role> TO GROUP <hive>;

Sentry supports column-level authorization with the SELECT privilege. Information about column-level authorization is in the Column-Level Authorization section of this page.

CREATE ROLE Statement

The CREATE ROLE statement creates a role to which privileges can be granted. Privileges can be granted to roles, which can then be assigned to users. A user that has been assigned a role will only be able to exercise the privileges of that role.

Only users that have administrative privileges can create or drop roles. By default, the hive, impala and hue users have admin privileges in Sentry.

CREATE ROLE <role name>;

Note that role names are case-insensitive.

DROP ROLE Statement

The DROP ROLE statement can be used to remove a role from the database. Once dropped, the role will be revoked for all users to whom it was previously assigned. Queries that are already executing will not be affected. However, since Hive checks user privileges before executing each query, active user sessions in which the role has already been enabled will be affected.

DROP ROLE <role name>;

GRANT ROLE Statement

The GRANT ROLE statement can be used to grant roles to groups. Only Sentry admin users can grant roles to a group.

GRANT ROLE <role name> [, <role name>]
    TO GROUP <group name> [,GROUP <group name>]
Sentry only allows you to grant roles to groups that have alphanumeric characters and underscores (_) in the group name. If the group name contains a non-alphanumeric character that is not an underscore, you can put the group name in backticks (`) to execute the command. For example, Sentry will return an error for the following command:
GRANT ROLE test TO GROUP test-group;
To grant a role to this group, put the group name in backticks:
GRANT ROLE test TO GROUP `test-group`;
The following command, which contains an underscore, is also acceptable:
GRANT ROLE test TO GROUP test_group;

Operating system group names must be in lowercase letters. Although group names are case-insensitive to Sentry, Sentry modifies capital letters within group names to be lowercase. For example, Sentry will change TestGroup to testgroup. It is not possible to disable this normalization. Therefore, group information within the environment must be in lowercase letters.

REVOKE ROLE Statement

The REVOKE ROLE statement can be used to revoke roles from groups. Only Sentry admin users can revoke the role from a group.

REVOKE ROLE <role name> [, <role name>]
    FROM GROUP <group name> [,GROUP <group name>]

GRANT <Privilege> Statement

Use the GRANT <Privilege> statement to grant privileges on an object to a role. The statement uses the following syntax:

GRANT
    <privilege> [, <privilege> ]
    ON <object type> <object name>
    TO ROLE <role name> [,ROLE <role name>]
You can grant the SELECT privilege on specific columns of a table. For example:
GRANT SELECT <column name> ON TABLE <table name> TO ROLE <role name>;

GRANT <Privilege> ON URIs (HDFS and S3A)

If the GRANT for Sentry URI does not specify the complete scheme, or the URI mentioned in Hive DDL statements does not have a scheme, Sentry automatically completes the URI by applying the default scheme based on the HDFS configuration provided in the fs.defaultFS property. Using the same HDFS configuration, Sentry can also auto-complete URIs in case the URI is missing a scheme and an authority component.

When a user attempts to access a URI, Sentry will check to see if the user has the required privileges. During the authorization check, if the URI is incomplete, Sentry will complete the URI using the default HDFS scheme. Note that Sentry does not check URI schemes for completion when they are being used to grant privileges. This is because users can GRANT privileges on URIs that do not have a complete scheme or do not already exist on the filesystem.

For example, in CDH 5.8 and later, the following CREATE EXTERNAL TABLE statement works even though the statement does not include the URI scheme.

GRANT ALL ON URI 'hdfs://namenode:XXX/path/to/table' TO ROLE <role name>;
CREATE EXTERNAL TABLE foo LOCATION 'namenode:XXX/path/to/table' TO ROLE <role name>;
Similarly, the following CREATE EXTERNAL TABLE statement works even though it is missing scheme and authority components.
GRANT ALL ON URI 'hdfs://namenode:XXX/path/to/table' TO ROLE <role name>;
CREATE EXTERNAL TABLE foo LOCATION
  '/path/to/table'
Since Sentry supports both HDFS and Amazon S3, in CDH 5.8 and later, Cloudera recommends that you specify the fully qualified URI in GRANT statements to avoid confusion. If the underlying storage is a mix of S3 and HDFS, the risk of granting the wrong privileges increases. The following are examples of fully qualified URIs:
  • HDFS: hdfs://host:port/path/to/hdfs/table
  • S3: s3a://host:port/path/to/s3/table

REVOKE <Privilege> Statement

You can use the REVOKE <Privilege> statement to revoke previously-granted privileges that a role has on an object.

REVOKE
    <privilege> [, <privilege> ]
    ON <object type> <object name>
    FROM ROLE <role name> [,ROLE <role name>]
For example, you can revoke previously-granted SELECT privileges on specific columns of a table with the following statement:
REVOKE SELECT <column name> ON TABLE <table name> FROM ROLE <role name>;

GRANT <Privilege> ... WITH GRANT OPTION

You can delegate granting and revoking privileges to other roles. For example, a role that is granted a privilege WITH GRANT OPTION can GRANT/REVOKE the same privilege to/from other roles. Hence, if a role has the ALL privilege on a database and the WITH GRANT OPTION set, users granted that role can execute GRANT/REVOKE statements only for that database or child tables of the database.
GRANT
    <privilege>
    ON <object type> <object name>
    TO ROLE <role name>
    WITH GRANT OPTION
Only a role with GRANT option on a specific privilege or its parent privilege can revoke that privilege from other roles. Once the following statement is executed, all privileges with and without grant option are revoked.
REVOKE
    <privilege>
    ON <object type> <object name>
    FROM ROLE <role name>
Hive does not currently support revoking only the WITH GRANT OPTION from a privilege previously granted to a role. To remove the WITH GRANT OPTION, revoke the privilege and grant it again without the WITH GRANT OPTION flag.

SET ROLE Statement

Sentry enforces restrictions on queries based on the roles and privileges that the user has. A user can have multiple roles and a role can have multiple privileges.

The SET ROLE command enforces restrictions at the role level, not at the user level. When you use the SET ROLE command to make a role active, the role becomes current for the session. If a role is not current for the session, it is inactive and the user does not have the privileges assigned to that role. A user can only use the SET ROLE command for roles that have been granted to the user.

To list the roles that are current for the user, use the SHOW CURRENT ROLES command. By default, all roles that are assigned to the user are current.

You can use the following SET ROLE commands:

SET ROLE NONE
Makes all roles for the user inactive. When no role is current, the user does not have any privileges and cannot execute a query.
SET ROLE ALL
Makes all roles that have been granted to the user active. All privileges assigned to those roles are applied. When the user executes a query, the query is filtered based on those privileges.
SET ROLE role name
Makes a single role active. The privileges assigned to that role are applied. When the user executes a query, the query is filtered based on the privileges assigned to that role.

SHOW Statement

  • To list the database(s) for which the current user has database, table, or column-level access:
    SHOW DATABASES;
  • To list the table(s) for which the current user has table or column-level access:
    SHOW TABLES;
  • To list the column(s) to which the current user has SELECT access:
    SHOW COLUMNS (FROM|IN) <table name> [(FROM|IN) <database name>];
  • To list all the roles in the system (only for sentry admin users):
    SHOW ROLES;
  • To list all the roles in effect for the current user session:
    SHOW CURRENT ROLES;
  • To list all the roles assigned to the given group name (only allowed for Sentry admin users and others users that are part of the group specified by group name):
    SHOW ROLE GRANT GROUP group name;
  • The SHOW statement can also be used to list the privileges that have been granted to a role or all the grants given to a role for a particular object.

    To list all the grants for the given <role name> (only allowed for Sentry admin users and other users that have been granted the role specified by <role name>). The following command will also list any column-level privileges:
    SHOW GRANT ROLE <role name>;
  • To list all the grants for a role on the given <object name> (only allowed for Sentry admin users and other users that have been granted the role specified by <role name>). The following command will also list any column-level privileges:
    SHOW GRANT ROLE <role name> on <object type> <object name>;

SELECT

The SELECT privilege allows a user to view table data and metadata. In additon, you can use the SELECT privilige to provide column-level authorization. See Column-Level Authorization below for details.

You can grant the SELECT privilege on a server, table, or database with the following commands, respectively:

GRANT SELECT ON SERVER <server name> TO ROLE <role name>
GRANT SELECT ON DATABASE <database name> TO ROLE <role name>
GRANT SELECT ON TABLE <table name> TO ROLE <role name>
Scope Available Operations
Server View table data and metadata of all tables in all the databases on the server
Database View table data and metadata of all tables in the database
Table View table data and metadata
Column View table data and metadata for the granted column
View
  • In CDH 5.x, column-level permissions with the SELECT privilege are not available for views.
  • In CDH 6.x, column-level permissions with the SELECT privilege are avaialbe for views in Hive, but not in Impala.

Column-Level Authorization

Sentry provides column-level authorization with the SELECT privilege. You can grant the SELECT privilege to a role for a subset of columns in a table. If a new column is added to the table, the role will not have the SELECT privilege on that column until it is explicitly granted.

You can grant and revoke the SELECT privilege on a set of columns with the following commands, respectively:

GRANT SELECT <column name> ON TABLE <table name> TO ROLE <role name>;
REVOKE SELECT <column name> ON TABLE <table name> FROM ROLE <role name>;

Users with column-level authorization can execute the following commands on the columns that they have access to. Note that the commands will only return data and metadata for the columns that the user's role has been granted access to.

  • SELECT <column name> FROM TABLE <table name>;
  • SELECT COUNT <column name> FROM TABLE <table name>;
  • SELECT <column name> FROM TABLE <table name> WHERE <column name> <operator> GROUP BY <column name>;
  • SHOW COLUMNS (FROM|IN) <table name> [(FROM|IN) <database name>];
As a rule, a user with select access to columns in a table cannot perform table-level operations, however, if a user has SELECT access to all the columns in a table, that user can also execute the following command:
SELECT * FROM TABLE <table name>;

Considerations for Column-Level Authorization

When you implement column-level authorization, consider the following:

  • When a user has column-level permissions, it may be confusing that they cannot execute a select * from <tablename> statement even though they have select privileges on some of the columns in the table. Instead, the user must explicitely define the columns that they want to query.
  • Using views instead of column-level authorization requires additional administration, such as creating the view and administering the Sentry grants. In addition, a new view may be needed for a new role, and third-party applications must use a different view based on the role of the user.
  • With HDFS sync enabled, even if a user has been granted access to all columns of a table, the user will not have access ot the corresponding HDFS data files. This is because Sentry does not consider SELECT on all columns equivalent to explicitely being granted SELECT on the table.
  • Column-level access control for access from Spark SQL is not supported by the HDFS-Sentry plug-in.

Example: Using Grant/Revoke Statements to Match an Existing Policy File

Here is a sample policy file:
[groups]
# Assigns each Hadoop group to its set of roles
manager = analyst_role, junior_analyst_role
analyst = analyst_role
jranalyst = junior_analyst_role
customers_admin = customers_admin_role
admin = admin_role

[roles] # The uris below define a define a landing skid which
# the user can use to import or export data from the system.
# Since the server runs as the user "hive" files in that directory
# must either have the group hive and read/write set or
# be world read/write.
analyst_role = server=server1->db=analyst1, \
    server=server1->db=jranalyst1->table=*->action=select
    server=server1->uri=hdfs://ha-nn-uri/landing/analyst1
junior_analyst_role = server=server1->db=jranalyst1, \
    server=server1->uri=hdfs://ha-nn-uri/landing/jranalyst1

# Implies everything on server1.
admin_role = server=server1

The following sections show how you can use the new GRANT statements to assign privileges to roles (and assign roles to groups) to match the sample policy file above.

Grant privileges to analyst_role:
CREATE ROLE analyst_role;
GRANT ALL ON DATABASE analyst1 TO ROLE analyst_role;
GRANT SELECT ON DATABASE jranalyst1 TO ROLE analyst_role;
GRANT ALL ON URI 'hdfs://ha-nn-uri/landing/analyst1' \
TO ROLE analyst_role;
Grant privileges to junior_analyst_role:
CREATE ROLE junior_analyst_role;
GRANT ALL ON DATABASE jranalyst1 TO ROLE junior_analyst_role;
GRANT ALL ON URI 'hdfs://ha-nn-uri/landing/jranalyst1' \
TO ROLE junior_analyst_role;
Grant privileges to admin_role:
CREATE ROLE admin_role
GRANT ALL ON SERVER server1 TO ROLE admin_role;
Grant roles to groups:
GRANT ROLE admin_role TO GROUP admin;
GRANT ROLE analyst_role TO GROUP analyst;
GRANT ROLE jranalyst_role TO GROUP jranalyst;