Hi Folks,

I would love to talk about expanding scope of Ranger!
Features:

  1.  Handling fine-grained access (FGA) on the attributes within arbitrary 
nestedstructures (eg JSON responses to API calls).  I just submitted 
RANGER-3828<https://issues.apache.org/jira/browse/RANGER-3828>, which does just 
that for JSON structures (Review request 
74057<https://reviews.apache.org/r/74057/>).  It could be expanded to include 
avro, etc.  We use it for FGA on API JSON response objects, but it or something 
like it could be used in a kafka or kinesis client, perhaps.  The same modeling 
of nestedstructures could be used within, say, the hive service to handle 
struct-valued attributes.  (I know some folks are skeptical, but I hope to 
convince them with working code soon!!!)
  2.  For custom ABAC policies, which are increasingly important to us, Context 
Enrichers/Policy Evaluators and the new RangerUserStoreRetriever are totally 
awesome, but a tad inflexible.  For example, I’d love for it to be easy to make 
the data from a context enricher available in tag policy conditions. I have a 
context enricher that represents a mapping from user to the set of business 
partners whose data they are permitted to view (“userPartners”).   Assuming a 
partner tag on, say, a hive table, I wish I could write something like 
“userPartners.includes(tagAttr.get(‘value’))”.  Or even better, 
UserPartners.intersects(tagAttr.get(‘value’)), where tagAttr.get(‘value’) is a 
set of partners.  (The inability of Ranger to handle set-valued tag attributes 
when Atlas has long handled them perplexes me.). In effect, I’d like there to 
be a single context that is enriched, and from which any policy type can access 
data.  Tag enrichment shouldn’t be separate from other enrichments.
  3.  A single Tagsync resource mapper service should support multiple 
“components”, as long as their resources are the same.  This is most obvious in 
a development environment:  since new contextenrichers require change to 
servicedefs, it’s crazy to have 5 hive tagsync servers, one for each 
servicedef, while one tries them out.  I can also imagine including different 
contextextenders (and therefore servicedefs) for different user populations in 
a production context (for performance if nothing else), and that should not 
require a separate tagsync service for each.

We use Ranger primarily on databricks in AWS and Starburst/Trino.  We’re 
expanding to include Athena on AWS, and OS Spark on K8S on MinIO.

Barbara


On 2022/06/30 19:27:15 Madhan Neethiraj wrote:
> (added [email protected]<mailto:[email protected]> back in this 
> thread)
>
> cdmikechen - thank you for your response.
>
> Selva - thank you for starting this thread. Responses from the community will 
> help prioritize next set of enhancements in Apache Ranger. Looking forward to 
> more responses from Apache Ranger community.
>
> I think enabling Apache Ranger authorization to services that don't support 
> pluggable authorization will be of immense value. This is critical especially 
> as enterprises use various cloud services where the service provider may not 
> support a pluggable authorizer model. Such enhancements to authorization 
> framework will help extend Apache Ranger authorization to services like 
> Databricks SQL, Snowflake, Google BigQuery, AWS Redshift, MS-SQL, Postgres, 
> and more.
>
> Thanks,
> Madhan
>
>     On 6/17/22, 12:09 AM, "陈 翔" 
> <[email protected]<mailto:[email protected]>> wrote:
>
>         1.     My company has used ranger from 1.2.0 and updated to 2.3.0 
> recently.
>
>         2.     At present, my company mainly use ranger with hive and atlas.
>
>         3.     I personally hope to support Drill or Druid.
>
>         4.     I  hope that the installation of ranger can be easier (for 
> example, remove Python dependency when installing), and I also hope that 
> other authentication methods such as oauth2/oidc can be supported. At the 
> same time, my company runs the Ranger service in k8s. We have developed an 
> k8s operator to simplify deployment and using. If possible in the future, I 
> will put relevant projects to Github.
>
>         5.     Not yet.
>
>         Hopefully this will help the community~
>         Thanks
>
>
>         发件人: [email protected]<mailto:[email protected]> 
> <[email protected]<mailto:[email protected]>>
>         日期: 星期五, 2022年6月17日 上午10:43
>         收件人: [email protected]<mailto:[email protected]> 
> <[email protected]<mailto:[email protected]>>, 
> [email protected]<mailto:[email protected]> 
> <[email protected]<mailto:[email protected]>>, 
> [email protected]<mailto:[email protected]> 
> <[email protected]<mailto:[email protected]>>
>         主题: Need community feedback on expanding scope of Apache Ranger
>         As Apache Ranger has taken a critical role in Security & Governance 
> of the big data/Hadoop technologies, I would like to see if the scope of 
> Apache Ranger can be extended to cover other areas such as CLOUD data 
> sources. To identify the industry needs, can you please provide some feedback 
> on your current Apache Ranger use and your future needs:
>
>
>           1.  Do you currently use Apache Ranger? If so, What Version?
>           2.  Apache Ranger supports authorization for 20+ data 
> sources/services. For which data sources/services do you use Apache Ranger 
> authorization?
>           3.  What additional data sources/services would you like to have 
> Apache Ranger authorization?
>           4.  What new features/enhancements would you like to see in Apache 
> Ranger?
>           5.  Have you developed Apache Ranger plugins for any data 
> sources/services? If yes, please list them.
>
>         As a community, we can review this (feedback) information to create 
> more backlog tasks for us to scope the next major release.
>
>         Thanks,
>         Selva-
>
>

--
Barbara Eckman, Ph.D.
she/her/hers
Distinguished Architect
Enterprise Metadata, Lineage and Access Control
Comcast
 [Logo  Description automatically generated]
[signature_951927206]


Reply via email to