Error ModelArts.4302 Occurred in Real-Time Service Prediction

Symptom

After a real-time service is deployed and running, an inference request is sent to the service, but error ModelArts.4302 occurred.

Cause Analysis and Solution

Error ModelArts.4302 may occur in multiple scenarios. The following describes two typical scenarios:

1. "error_msg": "Gateway forwarding error. Failed to invoke backend service due to connection refused. "

   This error occurs in either of the following cases:

   • The traffic exceeded the threshold that can be processed by the model. In this case, reduce the traffic or increase the number of model instances.
   • The image is faulty. In this case, separately run the image and check whether it is functional.
2. "error_msg": "Due to self protection, the backend service is disconnected, please wait moment."

   This error occurs due to excessive number of model errors. A large number of model errors trigger dispatcher circuit breaker, leading to a prediction failure. In this case, check the result returned by the model and handle these errors. Adjust request parameters or reduce the request traffic for higher model calling success rate.