All Products
Search

This Product

    Platform For AI:Configure custom monitoring metrics

    Document Center

    Platform For AI:Configure custom monitoring metrics

    Last Updated:Apr 01, 2026

    Report custom metrics from your service code and use them in EAS monitoring dashboards and auto scaling rules.

    How it works

    1. Add a reporting call to your service code that periodically POSTs metric data to http://localhost:8080/api/builtin/realtime_metrics.

    2. Declare the metric in your service JSON configuration. EAS reads this declaration at deployment time and creates a monitoring dashboard.

    3. After deployment, view the metric on the Service Monitoring tab. Data appears within one minute and is retained for one week.

    4. (Optional) Use the custom metric in an auto scaling rule.

    Limitations

    • The metrics field is supported only for services deployed with a custom image or custom processor.

    • Auto scaling supports only QPS (queries per second) and CPU utilization as built-in scaling triggers. To scale on other signals, use custom metrics.

    Prerequisites

    Ensure the following conditions are met:

    Step 1: Report metrics from your service code

    Add a periodic call in your service code to POST metric data to the following endpoint:

    POST http://localhost:8080/api/builtin/realtime_metrics

    The request body is a JSON array. Each element represents one metric sample:

    [
    {
        "name": "qps",
        "tags": {
            "status": "200"
        },
        "value": 20
    },
    {
        "name": "qps",
        "tags": {
            "status": "400"
        },
        "value": 13
    }
]
    Field Required Description
    name Yes Metric name. Must match the name declared in the service JSON configuration.
    value Yes Numeric value of the metric sample.
    tags No Key-value pairs that add dimensions to the metric. Tags split one metric into multiple series. For example, track QPS separately for each HTTP status code. In auto scaling rules, reference a tagged metric as custom[<name>]@<tag-key>[<tag-value>].

    To report a metric without dimensions, omit tags:

    [
    {
        "name": "qps",
        "value": 20
    }
]

    Step 2: Declare the metric in your service configuration

    Add a metrics field to your service JSON configuration. EAS uses this declaration to create the monitoring dashboard.

    Example: deploy a custom image service with a qps metric and a status tag:

    {
  "name": "metrics_test",
  "containers": [
    {
      "image": "registry-vpc.cn-chengdu.aliyuncs.com/eas/eas-image-****:metrics",
      "command": "python3 -u /image.py",
      "port": 5000
    }
  ],
  "metrics": [
    {
      "name": "qps",
      "tags": "status"
    }
  ],
  "metadata": {
    "instance": 1,
    "cpu": 2,
    "memory": 1000
  }
}

    Replace the image value with your own image URL. The metrics field supports the following parameters:

    Parameter Required Description
    name Yes Metric name. EAS creates a dashboard for this metric. Auto scaling rules reference it as custom[<name>].
    tags No Tag key name that splits the metric into series. For example, "tags": "status" separates values for status=200 and status=400. Reference individual series in auto scaling rules as custom[<name>]@status[<value>].

    For other configuration parameters, see JSON deployment parameters.

    Deploy the service:

    1. Go to the Deploy Service page. For more information, see Model service deployment by using the PAI console.

    2. In the Configuration Editor section, click JSON Deployment and paste your JSON configuration.

    3. Click Deploy.

    Step 3: View custom metrics

    After deployment, metric data appears in the dashboard within one minute. Data is retained for one week.

    To view custom metrics:

    1. On the EAS-Online Model Services page, click the service name to open Service Details.

    2. On the Service Monitoring tab, select the custom metric from the upper-left drop-down list.

    The dashboard shows:

    • Average metric value across all instances

    • Metric value for each individual instance

    image.png

    Step 4: Configure auto scaling

    Custom metrics work the same as built-in metrics in auto scaling rules. The metric name format depends on whether tags are configured.

    For more information about auto scaling, see Horizontal auto scaling.

    Auto scaling metric name syntax

    Configuration Format Example
    No tags custom[<metric-name>] custom[qps]
    With a tag custom[<metric-name>]@<tag-key>[<tag-value>] custom[qps]@status[200]

    Syntax breakdown: custom[qps]@status[200]=3

    • custom[qps] -- custom metric named qps

    • @status[200] -- filter by tag status = 200

    • =3 -- scaling threshold. EAS scales out when the average exceeds 3 and scales in when it drops below 3.

    Configure with eascmd

    For more information, see the "Method 2: Manage the horizontal auto scaling feature by using a client" section in Enable or disable the horizontal auto-scaling feature.

    No tags -- scale when average QPS exceeds 3:

    eascmd autoscale service_name -Dmin=1 -Dmax=10 -Dstrategies.custom[qps]=3

    With a tag -- scale based on QPS where status=200:

    eascmd autoscale service_name -Dmin=1 -Dmax=10 -Dstrategies.custom[qps]@status[200]=3

    Configure in the console

    For more information, see the "Method 1: Manage the horizontal auto scaling feature in the console" section in Enable or disable the horizontal auto-scaling feature.

    In the Auto Scaling Settings dialog box, in the Custom Scaling Metric section, enter:

    • No tags: enter custom[qps] as the metric name and 3 as the metric value. image.png

    • With a tag: enter custom[qps]@status[200] as the metric name and 3 as the metric value. image.png

    What's next