Served model endpoint creation failed and not able to see build log

[hardik]

Dear Databricks Support Team,

I am currently encountering an issue while attempting to serve a model using your platform. Whenever I initiate the model serving process, it fails, and I am unable to successfully deploy my model.

Additionally, I am facing difficulties in accessing the build logs, which are crucial for diagnosing the issue. Despite my attempts to view these logs, as evidenced in the attached screenshot, I find myself unable to locate the necessary information that could shed light on the cause of the failure.

[Kaniz · Accepted solution]

Hi @hardik, I understand that you’re experiencing difficulties with model serving and deployment. Let’s troubleshoot this issue step by step.

Model Serving with Databricks:

• Databricks offers Model Serving, which exposes MLflow machine learning models as scalable REST API endpoints. Here are some key points:
  • Advantages: It provides a production-ready environment, serverless compute, high availability, scalability (up to 25,000+ queries per second), and integration with MLflow Model Registry.
  • Limitations: Ensure that your model is registered in the MLflow Model Registry, and be aware of pack....
• If you’re using Azure Databricks, similar features are available, but make sure to create a workspac....

Troubleshooting Tips:

• Verify that your model artifacts are correctly located.
• Check if your model has any init scripts or package dependencies that might cause issues.
• Monitor endpoint health metrics (QPS, latency, error rates) to diagnose any anomalies.
• If you’re using Azure DevOps Pipelines, ensure proper firewall rules for your Azure database.

Accessing Build Logs:

• If you’re unable to access build logs, consider the following:
  • Check permissions and ensure you have the necessary access.
  • Look for specific log files related to the model serving process.
  • If using a specific tool or platform, consult its documentation for detailed steps on accessing logs.

Remember that debugging model deployment can be complex, but with patience and systematic troubleshooting, we can identify the root cause.

Feel free to provide additional details or ask further questions! 🛠️🔍