NVIDIA / gpu-monitoring-tools

Tools for monitoring NVIDIA GPUs on Linux
Apache License 2.0
1.02k stars 301 forks source link

How do I mount a custom csv file with Kubernetes ? #169

Open dmrub opened 3 years ago

dmrub commented 3 years ago

I want to enable the DCGM_FI_DEV_GPU_UTIL metric and want to add a custom csv file to the Kubernetes ConfigMap and then mount it. There is a helm option extraVolumeMounts to mount and arguments options to specify alternate filenames for loading csv file. However, there is no option to define the ConfigMap volume or. Only the extraHostVolumes option for hostVolume is present.

shan100github commented 3 years ago

@dmrub temporarily I did build a dcgm-exporter image with DCGM_FI_DEV_GPU_UTIL enabled and its in docker hub. shan100docker/dcgm-exporter:2.1.8-2.4.0-rc.2-ubuntu18.04-v2

Feel free to use it if necessary. Also make sure to pass below as values.yaml & use helm install dcgm-exporter gpu-helm-charts/dcgm-exporter -n <NAMESPACE> -f values.yaml

image:
  repository: shan100docker/dcgm-exporter
  pullPolicy: IfNotPresent
  tag: 2.1.8-2.4.0-rc.2-ubuntu18.04-v2

arguments: ["-f", "/etc/dcgm-exporter/dcp-metrics-included.csv"]

This is just for temporary purpose only, please don't get offend Repo mods & contributors.