gpu_operator extend_metrics

Enable time-sharing in the GPU Operator ClusterPolicy

Parameters

include_defaults

  • If True, include the default DCGM metrics in the custom config

  • default value: True

include_well_known

  • If True, include well-known interesting DCGM metrics in the custom config

namespace

  • Namespace in which the GPU Operator is deployed

  • default value: nvidia-gpu-operator

configmap_name

  • Name of the ConfigMap where the configuration will be stored

  • default value: metrics-config

extra_metrics

  • If not None, a [{name,type,description}*] list of dictionnaries with the extra metrics to include in the custom config

  • type: List

wait_refresh

  • If True, wait for the DCGM components to take into account the new configuration

  • default value: True