With the new driver, we should be able to export more metrics, e.g. memory usage.
Check what numbers we can shake out of rocm-smi and include the useful ones in the Python script for export.
If there is a lot of additional metrics, consider refactoring the parsing/metric-assignment part of the code.
| Subject | Author | Repo | Branch | Lines +/- | |
|---|---|---|---|---|---|
| prometheus: Add more stats to AMD ROCm GPU exporter | Klausman | operations/puppet | production | +33 -1 |
https://docs.google.com/document/d/1OX7JsearaFNO5b8cT4ScLrUppdvFJmz8gWE1nG6_smc/edit# has an overview of parameters and their outputs. Will discuss with elukey which options might be useful to add.
Change 626386 had a related patch set uploaded (by Klausman; owner: Klausman):
[operations/puppet@production] prometheus: Add more stats to AMD ROCm GPU exporter
Change 626386 merged by Klausman:
[operations/puppet@production] prometheus: Add more stats to AMD ROCm GPU exporter