Find the VMs working hardest
Which VMs are working hardest?
Ask for the top CPU and memory users across the cluster, right now. Then read the busiest node yourself.
What you get: the few VMs that are under pressure, and the node they sit on, before a slowdown turns into an outage.
Ask Quox
Section titled “Ask Quox”Ask in plain words:
Ask the Proxmox agent HYPERION which VMs are using the most CPU and memory right now. Read only please.
▤ AOCL │ L1 identity complete · allowed · 0ms
▤ AOCL │ L3 policy gate complete · proceed · 11ms
◐ DELEGATE │ delegate_to_agent agent_id=atlas ✔ DELEGATE │ delegate_to_agent agent_id=atlas │ 1 tools used, 1 tool outcomes · 30.4s
⛓ AEE │ envelope opened · stream_success
I'll hand this to HYPERION for a read-only check. Read complete, nothing changed (source: proxmox_read). Across the cluster, 81 of 104 VMs are running. Here are the two rankings. Top 10 by CPU (CPU % is the VM's share of its own vCPUs, 0 to 100) # │ VMID │ Name │ Node │ CPU % │ Mem used │ Mem alloc ───┼──────┼───────────────┼────────┼───────┼──────────┼────────── 1 │ 451 │ ci-01 │ hyp-04 │ 100.3 │ 8.6 GB │ 32 GB 2 │ 450 │ app-01 │ hyp-04 │ 52.0 │ 187.5 GB │ 195.3 GB 3 │ 9250 │ hyp-pool-9250 │ hyp-05 │ 36.9 │ 7.4 GB │ 8 GB 4 │ 446 │ sec-11 │ hyp-04 │ 12.7 │ 4.5 GB │ 32 GB 5 │ 206 │ uptime-12 │ hyp-02 │ 11.5 │ 1.5 GB │ 10 GB 6 │ 452 │ eval-12 │ hyp-04 │ 10.4 │ 3.4 GB │ 16 GB 7 │ 305 │ vpn-db-11 │ hyp-03 │ 7.5 │ 3.4 GB │ 16 GB 8 │ 430 │ docs-01 │ hyp-04 │ 6.7 │ 1.7 GB │ 2 GB 9 │ 401 │ site-12 │ hyp-04 │ 6.6 │ 2.4 GB │ 8 GB 10 │ 445 │ eval-11 │ hyp-04 │ 6.4 │ 5.2 GB │ 128 GB Top 10 by memory used # │ VMID │ Name │ Node │ Mem used │ Mem alloc │ Mem % │ CPU % ───┼──────┼───────────────┼────────┼──────────┼───────────┼───────┼────── 1 │ 450 │ app-01 │ hyp-04 │ 187.5 GB │ 195.3 GB │ 96.0 │ 52.0 2 │ 502 │ web-01 │ hyp-05 │ 46.6 GB │ 125 GB │ 37.3 │ 4.4 3 │ 203 │ security-01 │ hyp-02 │ 38.8 GB │ 40 GB │ 96.9 │ 0.6 4 │ 150 │ ci-12 │ hyp-01 │ 25.8 GB │ 32 GB │ 80.7 │ 0.1 5 │ 101 │ bastion-01 │ hyp-01 │ 21.1 GB │ 64 GB │ 32.9 │ 0.3 6 │ 9248 │ hyp-w2-423dry │ hyp-03 │ 15.3 GB │ 16 GB │ 95.7 │ 0.8 7 │ 707 │ app-17 │ hyp-07 │ 12.5 GB │ 24 GB │ 52.1 │ 0.5 8 │ 9245 │ hyp-story-a │ hyp-03 │ 12.3 GB │ 16 GB │ 76.6 │ 3.9 9 │ 309 │ vpn-11 │ hyp-03 │ 11.5 GB │ 24 GB │ 47.7 │ 2.8 10 │ 9205 │ hyp-pool-9205 │ hyp-03 │ 11.2 GB │ 12 GB │ 93.3 │ 3.0 A few things worth flagging: • app-01 (450, hyp-04) tops both lists: 187.5 GB of 195.3 GB (96%) and sustained 52% across 24 vCPUs. It is the single largest consumer on the cluster and is close to its memory ceiling. • ci-01 (451, hyp-04) is pinned at 100.3% CPU. For a 4 vCPU sanity box that is a runaway or a saturated workload. Worth a look with the VM doctor if it is not expected. • security-01 (203), hyp-w2-423dry (9248), hyp-pool-9205 (9205) and app-01 are all above 93% memory, so they have little headroom left. • hyp-04 carries most of the CPU-heavy VMs (5 of the top 10), so that node is the hot spot. If you want, I can run the VM doctor on ci-01 or app-01 (read-only) to explain what is driving them. Nothing was changed here. Tools used: proxmox_read HYPERION's read is above: 81 of 104 VMs are running, and nothing was changed. The two to look at first are app-01 (450), at 96% memory and 52% CPU, and ci-01 (451), pinned at 100% CPU.
43.7s · 1 tool · 0 approvals
quox.✔ Machine verified against the CLI on 17 Oct 2026 · test report mvctv1
HYPERION, the Proxmox agent, read every running VM and changed nothing. It ranked the top five by CPU and by memory, then said which VM to look at first and which node carries the most load.
Do it yourself
Section titled “Do it yourself”quox vm node hyp-04{"success": true,"node": "hyp-04","cores": 64,"memory": { "used": 229442363392, "total": 270227722240, "usedPercent": "84.9"},"rootfs": { "used": 24237346816, "total": 100861726720, "usedPercent": "24.0"},memory.usedPercent is how much of the node’s memory its VMs are using. A node near 100% has no room left for a VM to grow.
When one VM looks wrong, ask for a diagnosis. It reads the VM’s recent history and gives a verdict with the evidence:
quox vm doctor <name-or-vmid>What just happened
Section titled “What just happened”HYPERION read the cluster’s live numbers. Reading changes nothing, so no approval was needed. Moving or resizing a VM is up to you. A stop or a reboot is a governed write that can wait for a person to approve it.