quotas & budgets

four budget dimensions per account. check them with usage, fill what's left with max sizing.

the four dimensions

Dimension	How it counts
Machines	a count of your machines, whatever their state.
vCPU	a per-machine ceiling: no single machine gets more cores than this. cores are not summed across machines.
Memory	a summed budget: all your machines' memory together must fit.
Disk	a summed budget: all machine disks plus your shared drive must fit.

note the vCPU limit works differently. with an 8-vCPU ceiling you can run five 8-core machines, but no 9-core machine.

checking your usage

ssh

$ ssh dashboard@ironwire.sh usage
2/5 machines · 6 vCPU in use (max 8/machine) · 3.0/8.0 GiB · 14.0/25.0 GiB disk

left to right: machines used vs allowed, total vCPU in use with the per-machine ceiling in parentheses, memory used vs budget, disk used vs budget.

what happens at the limit

nothing is throttled or deleted. a create, resize, or resume that would exceed the budget is rejected with an error saying what didn't fit:

$ ssh dashboard@ironwire.sh create worker
quota exceeded: machines 5/5
$ ssh dashboard@ironwire.sh create big --mem=8G
quota exceeded: memory 5120/8192 MiB (need 5120 more)
$ ssh dashboard@ironwire.sh create wide --cpu=16
cores exceed the per-machine maximum: 8 max (requested 16)

(same errors on every surface: one-shot, the in-machine ironwire CLI, or the dashboard's create form.)

free room: delete a machine, shrink memory with resize (disk can't shrink), or pause a machine you don't need running.

sizing with `max`

--cpu, --mem, and --disk accept the literal max on create and resize:

ssh

$ ssh dashboard@ironwire.sh create big --cpu=max --mem=max

cli

$ ironwire create big --cpu=max --mem=max

for cpu, max means your per-machine ceiling. for memory and disk it means whatever remains of your budget at that moment, so a --mem=max machine takes all the memory your other machines haven't claimed.

paused machines

pausing hands the machine's vCPU and memory back to your budget; only running machines count there. its disk keeps counting, and slightly more than usual: the root disk plus a state snapshot roughly the size of its memory. it still occupies a machine slot.

resume re-checks the vCPU and memory quota. if the machine no longer fits (say another machine took the memory), the resume is refused and the machine stays paused:

$ ssh dashboard@ironwire.sh resume big
quota exceeded: memory 8192/8192 MiB (need 4096 more)

the shared drive

your shared drive's full capacity counts against the disk budget, attached or not. growing it is checked against the remaining disk budget; deleting it frees the capacity back.