Skip to main content
paluskapter
Silver ⭐️
January 7, 2019
Solved

Cloudera RecordService dead?

Hi,

 

I'm investigating the approaches for row level security in Impala, and I have just found multiple references for RecordService. However, it's homepage is removed, the source code is gone from GitHub, and it didn't get a new release for 2+ years on Maven, so I assume that it's dead.

Does anybody knows what happened to it, why it was cancelled?

 

These are the ones which I found as possible workarounds:

  • Store the data on HDFS in separate files splitted by the filter expression, and use HDFS ACLs for the files
  • Use HBase as a storage engine and query the data directly from it, as it supports row-level security
  • Disable access to the original Impala table, create views from it with the appropriate filters and use regular Impala ACLs for the views

Is there any alternative for row level security in Impala? RecordService sounded like the solution what I needed.

 

Thanks,
Peter

Robert Justice
Gold ⭐️⭐️
January 7, 2019

You are correct that Record Service has been no longer supported as of CDH 5.13 and the Record Service beta has ended (please see here).   Cloudera is actively working on a generally available solution. That solution takes advantage of a new streamlined architecture, while solving the same problems that RecordService set out to solve, both dynamic column and and row-level permissions that work across multiple data access methods (including Spark).    For now, the workarounds you mentioned would suffice, depending on your specific use case and needs.

Robert Justice, Technical Resolution Manager

3 replies

Robert Justice
Gold ⭐️⭐️
Gold ⭐️⭐️
January 7, 2019

You are correct that Record Service has been no longer supported as of CDH 5.13 and the Record Service beta has ended (please see here).   Cloudera is actively working on a generally available solution. That solution takes advantage of a new streamlined architecture, while solving the same problems that RecordService set out to solve, both dynamic column and and row-level permissions that work across multiple data access methods (including Spark).    For now, the workarounds you mentioned would suffice, depending on your specific use case and needs.

Translate
Robert Justice, Technical Resolution Manager
paluskapter
Silver ⭐️
January 8, 2019

Thank you very much for the confirmation!

What is the name of the new solution, and do you have a timeline for a release?

Translate
Robert Justice
Gold ⭐️⭐️
Gold ⭐️⭐️
January 8, 2019

There is no new product name.  This will be a new feature within Sentry called Attribute Based Access Control using Navigator Lineage tags for metadata.  At present, there is no estimated release date, but we will be sure to announce that.

Translate
Robert Justice, Technical Resolution Manager