add User to RLS in workspace, slows down the app to a crawl

[Anonymous]

We are experiencing an interesting issue with RLS.  

When we add (or delete) a user to one of our RLS groups and hit save we get a spinning when next to the save.  And it spins forever.  

IN the meantime if someone tries to go into the related application it takes forever to load.  So a normal app that would open in a few seconds takes 10+ minutes to open or filter on anything.  

We have several hundred users in our various RLS groups. Is there a limitation to how RLS works?  I found this article where a person mentions it caching for each user but the document they linked to no longer exists.

https://community.powerbi.com/t5/Desktop/RLS-slowness/m-p/702763 

Would a dynamic RLS be something that could optimize this issue better?

[parry2k]

@Anonymous Sorry my question was more on what DAX expression is used for RLS

[Anonymous]

@parry2k  data model is less than a gig big.  

It runs fine when users login and if you do a performance analyzer on the data model on your desktop while viewing as a group.  

However, it basically locks up if we try to add or delete a user from the security group online.  

[v-lionel-msft]

Hi @Anonymous ,

Is it slow when adding members? How long does it take when you log in to Power BI Service? What is the response speed when you perform other operations on Power BI Service? It sounds more like the slow response speed of the cloud server of Power BI Service.

The impact of RLS on performance is mainly reflected in the following aspects, which seem to be inconsistent with your situation.

Optimize RLS 

Performance Issues when to many RLSs are defined 

Best regards,
Lionel Chen

If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

[Anonymous · Accepted solution]

@v-lionel-msft  it appears to be a known issue (bug) by Microsoft.  They stated on their support page that if you turn your workspace off of premium you can then add people and then turn it back on.  

The support page says they expect a fix by 2/20/2020

[mai_2020]

Seems to still be a problem. But will try turning it off and on again..

[parry2k]

@Anonymous depends on the model, yes RLS can have a performance impact.

How big is your data mode? What logic you have put to implement RLS?

[Anonymous]

@parry2k 

Our dax is very simple.  Something like "SalesGroupID = "39383"  In the dimension for sales group.  Thus restricting that group to that salesgroupID