how do you disable serverless interactive compute for all users

[DataCurious]

I don't want users using serverless interactive compute for their jobs. how do i disable it for everyone or for specific users

[Walter_C]

You can disable Serverless from the account console. By following steps in doc https://docs.databricks.com/en/admin/workspace-settings/serverless.html#enable-serverless-compute 

[DataCurious]

There is no such option in my account setting

[Walter_C]

Ok on this case seems that your workspace was enabled by default with serverless function, on this case you wont be able to disable the feature, you can use budget policies for control https://docs.databricks.com/en/admin/usage/budget-policies.html 

[oskaribu]

But how do you restrict users from using serverless? The documentation at least says "Attribute serverless usage" not restrict using it...

[Walter_C]

Correct, as of now there is no way to restrict users the usage of serverless, if it is enabled it will become available for anyone, the idea with budgets is that you can control it a little bit more. I am aware of some feature requests to include compute policies that applies to serverless, but nothing yet available

[oskaribu]

Alright, thanks for the clarification! Hopefully there will be some features around this in the future.

[Walter_C]

Sure, This is the id of the idea for reference DB-13439

[israelst]

This is a blocker! We have a strict compute policies with specific tags and limitations... What is it worth if any developer can pick the serverless and bypass everything?! We also lack the ability to disable serverless in our account! this requires an urgent fix!

[prakharcode]

We are also waiting for the resolution of this, as we want to limit serverless compute for some groups of our users.

[uno_1]

How can we track the id DB-13439? What's the point of creating compute policies if any developer can use serverless compute?

[AugustinLourdhu]

We are also waiting for this feature, as we need to limit serverless compute access for certain users.

[Judith]

Same here, we want to control who uses serverless, not just everybody.. Shitty thing to enable it for everyone and not be able to disable or regulate it Databricks!!

[Staab]

Same here, we want to control who uses serverless, not just everybody.
Also in regard of network security, it seems like by default serverless compute enable enduser unlimted access to the internet.

[Staab]

In addition the provided budget policies act currently more like a tool to track and warn about costs that got generated by serverless compute. But budget policies  (in compare to cluster policies) do not limit the computing costs of certain Databricks User Groups.

What we - and probably also many other companies - want to achieve is, to define budgets inside Databricks on the level of IT Projects or for each Data Science Use Case for certain timeframes: Developers that got assigned to the Use-Case-Budget can e.g. train and fine-tune a ML-Pipeline only within the budget limits.