Vertex AI : how to delete a feature Store

I received today a worrying email about my GCP billing :

I tried to understand where this cost came from, and went to the billing section of the GCP console :

and Looking at the report, I tried to understand which service was costing me that much (I didn’t expect to spend so much this month) :

I digged into the demo project, to discover I had costs from Vertex AI :

Now I can digg further down in Vertex Costs :

And discovered two mains costs : (1) model deployment : about 30 € / day , and Feature Store for 10€ / day .

Removing the deployment of the model is quite straigthforward : just go to the model page, and undeploy it => you can do that with the mouse.

Here a screenshot for model deployment (it is the same for undeployment) :

But for the Feature Store it is a bit more tricky : as stated by the documentation (https://cloud.google.com/vertex-ai/docs/featurestore/managing-featurestores?cloudshell=true#delete_a_featurestore ) , you can’t undeploy a feature store by the User Interface => you need to do it programmatically.

The documentation is slightly misleading as it precises :

but the FEATURESTORE_ID is in fact the feature-store name !

To get the name you first need to list the feature-stores, and then read it :

So then you can delete it.

Here is the bash code to do so :

1. list the feature stores to get the names :

export LOCATION=your-location
export PROJECT=your-project

curl -X GET \
-H “Authorization: Bearer “$(gcloud auth application-default print-access-token) \
“https://$LOCATION-aiplatform.googleapis.com/v1/projects/$PROJECT/locations/$LOCATION/featurestores"

2. Delete the feature store :

export FEATURESTORE_NAME=name_of_the_feature_stores
export BOOLEAN=TRUE

curl -X DELETE \
-H “Authorization: Bearer “$(gcloud auth application-default print-access-token) \
“https://$LOCATION-aiplatform.googleapis.com/v1/projects/$PROJECT/locations/$LOCATION/featurestores/$FEATURESTORE_NAME?force=$BOOLEAN"

3. Verify there is no more stores on your console :

Et voilà !

Thanks.