Docker container for nvidia gpu monitoring with Telegraf
This Dockerfile
is the merge between the two used by nvidia and telegraf to craft their containers. Is nothing more than a PoC. Today it works, do not expect that tomorrow it will.
To build your local version
docker build . -t telegraf-nvidiasmi
Pull the upstream version on dockerhub
docker pull mconcas/telegraf-nvidia
Start Telegraf agent with the nvidia runtime, and monitor your local nvidia GPU
docker run -d [--net=possible-net] --hostname=$HOSTNAME --name=telegraf --runtime=nvidia -v $PWD/telegraf.conf:/etc/telegraf/telegraf.conf:ro mconcas/telegraf-nvidia-runtime telegraf