databricks-demos / dbdemos

Demos to implement your Databricks Lakehouse
Other
255 stars 80 forks source link

customer-360 demo: error resulting in hidden system catalog [__databricks_internal] #56

Closed evogelpohl closed 11 months ago

evogelpohl commented 11 months ago

I attempted to install the c360 customer_360 demo on an Azure premium workspace w/ unity enabled. However, my Azure Quota for the clusters the demo wanted to create was insufficient to perform the demo install. It errored out w/ an azure-quota-error.

However, immediately after the error a system catalog appeared in Unity, [__databricks_internal], owned by system.

(correlated, perhaps not causal)

It contains a schema for a DLT project I created quite a long time ago & tables with the prefix [__dlt_materializationschema{guid}].

Screenshot 2023-08-03 at 9 35 46 AM

I can't re-hide this table. Thoughts?

QuentinAmbard commented 11 months ago

Hi, Non-metastore-admin users will not see it as long as permissions have not been explicitly granted. These catalogs/schemas are used to store internal tables for DLT, and will appear if DLT pipelines publishing to UC have been created in the workspace. if you drop the DLT pipeline I'd assume the schema should disappear by itself no?

QuentinAmbard commented 11 months ago

for your install you can run it with use_current_cluster = True, it'll prevent from creating a new cluster and avoid the quota issue

evogelpohl commented 11 months ago

right, but the catalog wasn't there prior to the installation of c360. Not sure why it appear other than that the c360 added DLT & exposed it. I had other DLT jobs running and it was hidden to me. I am a metastore owner. I'll drop the pipeline & see if it vanishes.

QuentinAmbard commented 11 months ago

I think it's just because it created a new DLT pipeline and the new one are using the internal catalog. If you create one your self it'll appear too.

evogelpohl commented 11 months ago

Thanks. Reminder on how those hidden tables tie to DLT.