Configurability

[miconeilaws]

Hello team,

I'm really excited that NVIDIA is vending an agent with best practise and opinionated observability built in to expose the key GPU metrics and also health qualifications etc. Thats a really valuable tool and something I want to be able to leverage and not have to build myself.

However, I already have my existing observability stack and I want to be able to export the fleet intelligence telemetry there. This is one reason why I sent c1bab9c and thanks for your collaboration.

One challenge I still have is around configurability though, to set OTLP endpoints etc I need to use the fleetint CLI which is a bit cumbersome and not really scalable and natural for my IaC templates in ephemeral clusters that we spin up and down. I would ideally want to be able to configure the fleet intelligence agent at the code level; via env variables, a config file or any other mechanism.

Is this something on your roadmap? Would you be open to a contribution?

[miconeilaws]

@jingxiang-z @rsampaio

[jingxiang-z]

Hi @miconeilaws , the main purpose of this agent is to send the inventory, attestation evidence and telemetry to NVIDIA hosted Fleet Intelligence service. With that said, the CLI enroll command is to precheck the system, get the token from our backend and send the data to our backend. But this doesn't stop you from using it with your existing observability stack. Could you tell us more about how do you want to make it more configurable at the code level? Are you trying to configure the OTEL endpoint in a config file?