Viewing Details About a Real-Time Service

Service

Displays the topology, basic information, network configuration, traffic weight, and advanced configuration of the service.

For details about the service topology, see Viewing the Service Deployment Topology.

Click to copy the API URL for calling the service.

Deployment

Displays the service deployment information. You can switch between cards on the left to view the information of different deployments.

Key operations available on the deployment details page:

• In the left-hand deployment card, you can perform operations such as upgrading deployments, stopping service deployments, and scaling real-time service deployments.
• Click View Details next to the number of deployment replicas to view the details.
  • You can delete instances as needed or scale deployed instances. For details, see Scaling a Real-Time Service Deployment.
  • In the deployment replica list, you can click the number of pods to view pod details. You can view pod events and logs in the pod operation column and delete pods as required.

  If the resource pool is a public resource pool, you cannot delete and recreate instances or delete pods.

• Click Modify Configuration next to Traffic Weight to modify the traffic weight of the service deployment. For details, see Viewing/Modifying Traffic Weights.
• Click the version count to see all deployed versions. You can switch between or delete these versions. The current version cannot be deleted.
• In the basic information area of the deployment, you can copy the deployment name and ID, modify the deployment description, and view the creation time, modification time, and creator of the deployment.

• In the resource configuration area of the deployment, click the resource pool name to go to the resource pool details page and view the resource pool configuration. For details, see Viewing the Dedicated Resource Pool List and Details.

Prediction

Performs real-time prediction on this page. For details, see Using the Prediction Feature.

Monitoring

Shows monitoring data for the service. For details, see Viewing Performance Metrics of a Real-Time Service on ModelArts.

Cloud Shell

You can use Cloud Shell provided by the ModelArts console to log in to the instance container of a running real-time service. For details, see Using Cloud Shell to Debug a Real-Time Service Instance Container.

Events

Service events: Records key activities like starting, stopping, updating, or recovering services. Data is retained for one month and automatically cleared thereafter.

Pod events: Records lifecycle events and exceptions for pods within the Kubernetes cluster. Data is retained for one hour and automatically cleared thereafter.

For details about how to view events of a service, see Viewing Events of a Real-Time Service.

Logs

Runtime logs are exported to Log Tank Service (LTS). LTS automatically creates log groups and streams and caches logs for seven days by default. You can search for and analyze runtime logs.

• Log search: Search logs using specific keywords or phrases. Narrow your results by selecting a specific time range to find events and issues during that period. For details, see Searching for Logs.
• Statistical charts: After sending logs to LTS, use Using SQL Analysis Syntax to find important log data and view the results as statistical charts. For details, see Visualizing Logs in Statistical Charts.
• Log analysis: Before searching for analyzing logs, set up structured data and indexing for them. For details, see Setting Cloud Structuring Parsing.
• Real-time logs: Once you connect your real-time service logs to LTS, they will be sent every minute. You can view these updates from the Real-Time Logs tab, where you can also easily search and analyze the data. For details, see Viewing Real-Time Logs.

Intelligent O&M

As foundation model deployment scales and the complexities of cross-node deployment and load balancing grow, traditional resource scaling strategies based on native Kubernetes Horizontal Pod Autoscaler (HPA) can no longer satisfy the need for refined adjustments to the prefill (P) and decode (D) instance ratio. In real-world operations, users often face low resource utilization or performance bottlenecks due to sub-optimal P/D ratios, yet existing tools lack the ability to provide dynamic ratio recommendations based on simulation algorithms. To resolve this, ModelArts resource pools now support the installation of the HRA plugin. By utilizing simulation algorithms to calculate and display the optimal P/D ratio recommendation, users can manually adjust instance ratios based on real-time metric analysis, without relying on auto scaling, to achieve highly efficient inference service deployment within their resource capacity.

Enable Monitoring and click the edit button to view the optimal inference unit ratio calculated based on the intelligent algorithm policy.

When a real-time service uses a physical resource pool with the HRA plugin installed, and the model assets of the real-time service contain the "dynamic ratio recommendation" label, inference unit ratio detection is supported.

Private Network Connections

Displays the private network connection requests that need to be approved by the current account.

ModelArts offers private network connection. When you create a private network connection request, it automatically sets up a VPCEP to connect your VPC with the real-time inference service securely. For details, see Accessing a Real-Time Service Through a Private Network.

Tags

Displays tags that have been added to the service. You can add, modify, and delete tags.

For details about how to use tags, see Using TMS Tags to Manage Resources by Group.