Grafana is proud to be a Silver Sponsor of this year’s KubeCon India!

Running a homelab means running a diverse set of applications like NextCloud, readeck.org, usememos.com.

They are written in many languages and frameworks and in general lack a good way to monitor them. The old way to monitor them involved having a basic blackbox_exporter enabled synthetic monitoring. Even instrumented, each application has a different set of metrics, leading to complex alerting and dashboards.

Grafana Beyla, powered by eBPF and OpenTelemetry, combined with Prometheus promises to generate first class observability signals for services regardless of the language and frameworks used. We will walk through deploying Beyla for a homelab, and the challenges in having good observability even with Beyla.

We will walk through:

1. Deploying Beyla in your environment with Prometheus
2. Setting up dashboards for the data
3. Robust alerting over the signals, especially in low, sporadic traffic (homelab) scenarios.
4. The missing signals from eBPF and how you can fill those gaps

Modern AI workloads rely on large GPU fleets whose efficient utilisation is crucial due to high costs. However, gathering telemetry from these workloads to optimise performance is challenging because it requires manual instrumentation and adds performance overheads. Further, it does not produce telemetry in a standardised format for commonly used visualisation tools like Prometheus.

This talk explores the potential of leveraging eBPF to capture CUDA calls made to GPUs, including kernel launches and memory allocations. Data from these probes can be used to export Prometheus metrics, facilitating detailed analysis of kernel launch patterns and associated memory usage. This approach offers significant benefits as eBPF imposes minimal overhead and requires no intrusive instrumentation. Our implementation is also open-source and available on GitHub.