Auto-placement and the scheduler

When you spin up a VM with Host set to "Auto (best fit)," the scheduler picks the hypervisor for you. It runs on the server, so the choice is authoritative — the browser can never place a VM onto a host without room.

How it chooses:
1. It looks at every hypervisor host that is online and has enough free CPU and memory for the VM you asked for, after accounting for the VMs already running on it.
2. Among those, it picks the best fit by strategy:
• Spread (the default) — the host with the most free capacity wins, so load is balanced across your hosts.
• Binpack — the tightest host that still fits wins, so VMs are consolidated onto fewer hosts and big holes stay open for big VMs.

The preview:
With Auto selected, the spin-up form shows a live preview like "Scheduler → alpha · 14 vCPU free after load." Change the vCPUs or memory and the preview updates. If nothing can fit, it tells you before you try.

Why capacity is respected:
Each host advertises how many cores and how much memory it has. The scheduler subtracts the CPU and memory of the VMs already placed there, so it never promises capacity that isn't real. If every host is full, the VM isn't created and you get a clear "no capacity" message.