What Should I Do If a Real-Time Service Remains in the Deploying/Modifying State?

Symptom

After deploying the service, it remains stuck in the deploying state. On the Events tab page of the service details page, an error message is displayed: "Failed to start the pod container in deployment [%s]."

Possible Causes

Certain abnormal events related to Pods are reported in the service events, but only a summary of the issue is provided. When a Pod is not ready, the service does not recognize the instance as ready and available, causing it to remain in the deploying state.

Solution

1. Log in to the ModelArts console and choose Model Inference > Real-Time Inference.
2. Click the real-time inference service name. On the service details page that is displayed, click the Deploy tab.
3. Locate the deployment of the abnormal instance and click View Details next to the number of deployed replicas. The instance details page is displayed.
   Figure 1 Viewing details
   
4. On the service details page, click the number of Pods. The Pod details page is displayed on the right.
5. Click Events in the Operation column to view the events of the abnormal pod.

   The pod event list provides detailed event logs, consistent with those shown in Kubernetes, including event names and specific descriptions. Use this information to further troubleshoot the issue.