Using CERMM
Proxmox and Hyper-V
Put the agent on the host, and its page gets a Virtual machines section: every guest, its state and resources, and the buttons to start, stop and snapshot.
Proxmox VE
On a Proxmox VE node the agent (as root) asks pvesh: the cluster's resources every report, its status every 5 minutes. No API token, and it works behind NAT. It reads each guest's config under /etc/pve/nodes/<node>/ too.
Each node shows only its own guests (qm and pct act locally), so put the agent on every node.
What you see
- VMs and LXC containers: ID, state, uptime, CPU of its cores, memory, disk, snapshots, tags, HA and locks (Backing up…). One set to start on boot but stopped is red.
- Templates, apart.
- Storage pools with meters, and the cluster and its nodes.
What you can do
Click a guest: Start, Resume, Shut down (180 seconds), Reboot, Stop, and Snapshots: take one (with memory for a running VM), roll back (restarting it when the snapshot has no memory), delete. These are qm and pct commands sent as signed jobs through the Run drawer.
Hyper-V
Not tried on a real Hyper-V host yet. It's built and its parsing is tested against sample output, but nobody has run it against a real host. Tell us how it goes.
On a Windows host running Hyper-V (vmms.exe), the agent runs the Hyper-V PowerShell module in the background, at most once a minute, and again right after a job.
- In Hyper-V's own words: VMs by name, Saved, CPU as a share of the host, and Checkpoints (Standard or Production, by the VM's setting).
- Start, Resume, Shut down, Reboot, Save, Turn off; take, Apply and delete checkpoints. Commands are PowerShell addressed by the VM's ID.
Guests that have an agent
A guest is matched to an enrolled device by its UUID (Proxmox) or a network card's MAC address. That device's nameplate then says Host: pve-01, VM 101.
Not yet
- Alerts for guests (one that should be running, isn't).
- Backups per guest: see Backups.
- Hyper-V clusters and storage.