model_serving_endpoints in DAB updates every time

[Binnisb]

Love the model_serving_endpoints in the dab, but now it takes over 6 minutes to deploy resources when they already exist. It says (updating) in the serving tab in the side bar even if nothing has changed.

Is there a way to not update the endpoints after initial deployment when the yml for the resource hasn't changed?

[Alberto_Umana]

Hi @Binnisb,

Thanks for your question, can you provide your yml deployment file? 

[VZLA]

@Binnisb thanks for your question!
As far as I'm aware of, there is currently no built-in way to prevent the serving endpoints from updating during deployment, even if the YAML hasn't changed. Each deployment checks for updates and goes through the update process regardless; I believe the update step is not solely based on whether the yml for the resource has changed or not. You can reaching out to Databricks support for feature enhancement requests.

[binnisbj]

resources:

  model_serving_endpoints:

    embed_serving:

      name: embedding-gpu-m

      config:

        served_entities:

        - name: embedding-gpu-m

          entity_name: system.ai.bge_large_en_v1_5

          entity_version: "3"

          workload_size: Small

          scale_to_zero_enabled: true

          workload_type: GPU_MEDIUM

          min_provisioned_throughput: 0

[Alberto_Umana]

I have created an internal feature request for this behavior: DB-I-13108