Rohit Gupta created RANGER-5745:
-----------------------------------
Summary: Ranger: Support for wildcard in table name on SQL row
filter policy
Key: RANGER-5745
URL: https://issues.apache.org/jira/browse/RANGER-5745
Project: Ranger
Issue Type: Improvement
Components: Ranger
Reporter: Rohit Gupta
Assignee: Rohit Gupta
Experian has an application based on Hive in CDP Public Cloud. In a particular
Hive database, each table is partitioned by ISO 8601 date string. Users of the
database should have access to only the last five years of data.
We're currently able to achieve this requirement by applying a row-level filter
policy. However, it's necessary to create one policy per table.
Experian has the same partition specification on every table in the database,
so they'd like to be able to create a single policy, specifying criteria like
"dt_part > '2019_01_10'", the name of the database, and "*" for the table name.
Such wildcard table-name specification is not currently supported (as validated
by documentation and testing).
I provided a workaround, using a shell script to read a file containing the
table names (the CSV output from SHOW TABLES in Hue) and generate a row
filtering policy for each via the Ranger API. Experian has confirmed the
workaround does what they need, but they still would prefer to have a single
policy covering all the tables in the database.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
