Monitoring Amazon RES virtual desktop nodes

[americanjohnny1988]

Hi, we need to monitor Amazon RES VDI usage metrics in terms of CPU, RAM etc. for rightsizing purposes. It's useful to understand if the selected instance type (e.g. a g5.48xlarge) is not overprovisioned.

[nicolaven]

Thanks @americanjohnny1988 — accepted, and we're working on it.

RES desktops will be monitored on the same stack you already run on a ParallelCluster/PCS head node: they're discovered via the res:EnvironmentName tag and get dedicated RES Node List and RES Node Details dashboards covering CPU, RAM, and GPU utilization — so an overprovisioned instance type (e.g. an idle-GPU g5.48xlarge) is easy to spot for rightsizing. It's fully opt-in and won't affect existing deployments.

Will update this issue when it ships in a release.

[nicolaven]

Shipped in v2.10 🎉

RES desktops are now monitored on the same stack you run on your ParallelCluster/PCS head node:

1. Re-run the installer on the head node with RES_ENVIRONMENT_NAME=<your-res-env-name> set, then docker restart prometheus.
2. Add post-install.sh to your RES project launch script so each desktop runs the exporters.

You'll get RES Node List and RES Node Details dashboards (CPU, RAM, GPU, instance type, owner/project) — the GPU% column makes idle-GPU rightsizing candidates easy to spot. Full setup steps are in the README.

The core data path is validated, but the GPU% column and the installer on your specific RES-ready AMI are best confirmed in your environment — let us know how the rollout goes and if the Owner column needs a tag-key tweak for your RES version.