I got my hands on a few DGX Spark units and have been experimenting with various use cases.
The major pain point throughout was being forced to continuously open the various CLI monitoring tools for hardware observability. I know other monitoring tools can be used but I wanted a simpler solution; once I got my clusters up and running correctly, I began working on a dashboard tool to make cluster monitoring easier.
I would like to share this tool with the community and I figured this is the best place to bring awareness.
Please see GitHub - chronosolidus/dgxsparkmonitor: Real-time cyberpunk-themed monitoring dashboard for NVIDIA DGX Spark GPU clusters via SSH · GitHub for more details.
I packaged it up in a docker image for simple deployment and I feel this will help many others from resorting to monitoring their spark clusters with legacy CLI interfaces or full monitoring solutions such as Prometheus.
Let me know what you think!