/api/configs/pod_template

Each pod template is defined by a name and a Dictionary[String, Any], where the dictionary is a JSON representation of the Kubernetes pod specification.

Default Value: Empty dictionary {} (no pod templates configured by default)

Example

The following example shows a pod template that adds GPU resources and node selectors:

{
    "gpu_template": {
        "spec": {
            "nodeSelector": {
                "accelerator": "nvidia-tesla-v100"
            },
            "containers": [{
                "name": "main",
                "resources": {
                    "limits": {
                        "nvidia.com/gpu": "1"
                    }
                }
            }]
        },
        "metadata": {
            "labels": {
                "gpu-enabled": "true"
            }
        }
    }
}

The key here is gpu_template, which is the name of the pod template. The value is a dictionary that contains the Kubernetes pod specification. When a pool is configured to use this pod template, the Kubernetes pod specification will be merged with the base pod configuration.

A pool can apply this pod template by adding the name to the common_pod_template array, and a platform can apply this pod template by adding the name to the override_pod_template array.

To learn more about pod templates, see Pod Template Concepts.