2026.02.09

The meeting covered a presentation that provided a technical deep-dive into collecting per-job resource metrics (CPU, memory, IO, network) in multi-tenant environments, examining different data sources (cgroup v2, /proc/ files, and future eBPF). The community reached consensus that cgroup v2-based collection for CPU, memory, and IO has no deployment blockers, with job-level granularity sufficient for initial implementation and configurable cgroup directory patterns needed to support different job schedulers.

Here is the link to the presentation slides: https://github.com/ovis-hpc/ovis-wiki/blob/main/Per-Job%20Monitoring%20Data%20Collection%20with%20LDMS%20--%20published-1.pdf