What causes MODEL_SERVING_ENDPOINT_NOT_FOUND?

The serving endpoint was deleted manually or by an automated cleanup script. The endpoint name in the application code or pipeline configuration has a typo. The endpoint was created in a different workspace or environment than where the query is being made. An infrastructure-as-code deployment renamed the endpoint without updating consumer references

How do I fix MODEL_SERVING_ENDPOINT_NOT_FOUND?

Step 1: In the Databricks workspace, navigate to Serving and confirm whether the endpoint exists and is in the Ready state.. Step 2: If the endpoint is missing, recreate it by specifying the model name, version or alias, and compute configuration.. Step 3: Update the application code or pipeline to use the correct endpoint name.. Step 4: If the endpoint exists but is still initializing, wait for it to reach the Ready state before retrying.. Step 5: For production systems, manage endpoints via Databricks Asset Bundles to prevent accidental deletion.

High severityconfiguration

Power BI Refresh Error:
MODEL_SERVING_ENDPOINT_NOT_FOUND

What does this error mean?

A request to query or manage a Databricks Model Serving endpoint failed because the endpoint with the specified name does not exist or was deleted.

Common causes

1The serving endpoint was deleted manually or by an automated cleanup script
2The endpoint name in the application code or pipeline configuration has a typo
3The endpoint was created in a different workspace or environment than where the query is being made
4An infrastructure-as-code deployment renamed the endpoint without updating consumer references

How to fix it

1Step 1: In the Databricks workspace, navigate to Serving and confirm whether the endpoint exists and is in the Ready state.
2Step 2: If the endpoint is missing, recreate it by specifying the model name, version or alias, and compute configuration.
3Step 3: Update the application code or pipeline to use the correct endpoint name.
4Step 4: If the endpoint exists but is still initializing, wait for it to reach the Ready state before retrying.
5Step 5: For production systems, manage endpoints via Databricks Asset Bundles to prevent accidental deletion.

Frequently asked questions

How long does it take for a model serving endpoint to become ready after creation?

Typically 5 to 15 minutes for CPU endpoints and 10 to 30 minutes for GPU endpoints, depending on the model size and container build time. The endpoint state transitions through Pending, Updating, and then Ready.

Can I route traffic between two model versions on the same endpoint?

Yes. Databricks Model Serving supports traffic splitting — you can configure multiple served entities on one endpoint and assign each a percentage of incoming traffic, which is useful for A/B testing or gradual rollouts.