Chatbot Deployment Error on Databricks

[Tee_O]

Hi All,

I am trying to deploy a chatbot on databricks and I get this error in the build log:

Build could not start due to an internal error. If you are serving a model from UC and Azure storage firewall or Private Link is configured on your storage account, please verify your network connectivity is configured properly according to https://learn.microsoft.com/en-us/azure/databricks/security/network/serverless-network-security/serv.... For additional assistance, contact Databricks Support.

Any Ideas what the issue is and how I can fix it? I have discussed with my IT Team but they are not sure what it is.

[Vidhi_Khaitan]

Hi team,
The underlying cause of this issue is an incorrect Network Connectivity Configuration (NCC) for Azure Storage in the Databricks environment. The NCC determines which resources are accessible from within the Databricks environment. If the NCC is not configured correctly, Databricks may not be able to access the required Azure Storage resources, leading to connectivity issues with Model Serving.

There are two common causes for an incorrect NCC:

1. Azure Storage firewall is enabled, but the required IP addresses or Azure services are not whitelisted in the firewall rules.

2. Azure Private Link is used, but the required private endpoints are not configured for Azure Storage. This second option is the usual problem related to this error, as the Azure blob storage endpoint is not present in the Azure NCC configuration.

The Model serving endpoint “build could not start due to an internal error” occurs because Model Serving uses the Azure blob storage path to download model artifacts. If this private endpoint is not present, Model Serving won’t be able to download the model artifacts needed for the deployment.

To solve this issue - 

1. Follow the steps mentioned in the NCC configuration Azure documentation.
2. Ensure to follow step 4 and add the  blob endpoints needed along with the dfs ones for model serving.

Hope this helps!