← Features

VM pools

A pool is an amount of vCPU and memory reserved for you in one region. The VMs you place in a pool share that capacity. One VM can use all of it, or many VMs can split it.

A pool is billed for as long as it exists, empty or full. Adding VMs doesn't change the cost; resizing does. Pools reserve CPU and memory, not disk.

Pools and plans

  • Personal has one pool: your account's capacity, from 2 to 16 vCPU, shared by all your VMs. You can't create more pools.
  • Work can have as many pools as it needs, billed by the hour.
  • Enterprise pools have custom capacity, dedicated hardware, and contract pricing.

See Personal, Work, and Enterprise for plan prices, and exe.dev/pricing for the full comparison.

Kinds of pools

Shared team pool. A team admin creates it with pool new. Any member of the team can create VMs in it.

Member pool. When default placement is member-pool, each member gets a pool of their own the first time they run new without --pool. Everyone on the team sees it in pool list, but only that member creates VMs in it.

Standalone VM. A standalone VM (sandbox) is a pool of one, sized to the VM and billed by the hour. See Standalone VMs.

Capacity, disk, and bandwidth

Pool capacity comes in units of 2 vCPU / 4 GB of memory. On Work, a pool can be any even number of vCPU from 4 to 512, with 2 GB of memory per vCPU.

A Work pool holds up to 25 VMs per vCPU, up to 1,000 VMs. A 4 vCPU pool holds 100 VMs, and a 32 vCPU pool holds 800. There is no team-wide VM limit.

Work's $150/month minimum spend covers 4 vCPU / 8 GB of capacity, counted across all of the team's pools and standalone VMs together. Capacity beyond that is metered at $0.105 per 2 vCPU per hour. For example, a team with one 8 vCPU pool pays the minimum spend plus $0.21 per hour for the 4 vCPU above the included capacity. A pool that is pending approval is not billed.

On Work, each pool also carries included disk and outbound bandwidth, based on its configured capacity:

  • Disk: 100 GB per 2 vCPU, averaged over the billing period.
  • Bandwidth: 50 GB per vCPU per billing period, prorated by how long the capacity existed.

The allowance belongs to the pool, including the 4 included vCPU. It covers that pool's VMs and doesn't move between pools. On Personal it belongs to your account.

Additional disk is charged per GiB-month. Additional bandwidth is metered and priced at the rate on exe.dev/pricing. See Usage for how both are measured.

Regions

A pool lives in one region, and its VMs run there. Pick any region pool new lists with --region=<code>, including one outside your account's region. Pools do not move between regions.

Sizes and approval

Pool sizes up to 32 vCPU take effect immediately. Larger sizes, up to 512 vCPU, need approval from exe.dev support:

  • pool new above 32 vCPU creates a pending pool that accepts no VMs until support activates it.
  • pool resize that grows an active pool above 32 vCPU records a request. The pool keeps its current size until support approves it.
  • Shrinking a pool takes effect immediately.

Default placement

You don't have to name a pool when you create a VM. If you leave out --pool, the VM goes where the team's default placement says. On Work and Enterprise, a team admin sets it, and switching to Work sets it to your first pool.

team settings vm-placement pool build             # one shared pool
team settings vm-placement member-pool --cpus=4   # a pool per member
team settings vm-placement poolless               # outside any pool (legacy)
team settings vm-placement default                # clear the setting
  • pool: every member's new VMs go to that pool.
  • member-pool: each member gets a pool of that size the first time they run new. --max-vms sets a lower VM limit.
  • poolless: new VMs stay outside pools, so you can migrate on your own schedule instead of having every new VM land in a pool right away. Available only while the team still has VMs outside pools.
  • default: new without --pool fails until an admin sets placement again, unless the team still has VMs outside pools.

team settings vm-placement shows the current setting. Admins can also change it at exe.dev/team. new --pool=<name> always overrides it.

Using pools

Only team admins can create, resize, adopt into, or delete pools.

Create a pool

pool new build --cpus=8 --region=lax

--max-vms sets a VM limit below the pool's size-based maximum.

Put VMs in a pool

new --pool=build
cp my-vm my-vm-copy --pool=build

A VM in a pool can be sized up to the pool's capacity. Without --pool, new follows the team's default placement.

Resize a pool

pool resize build --cpus=16
pool resize build --max-vms=200

Resizing a pool does not resize its VMs. To give a VM more of the new capacity, resize the VM with resize <vm>.

A pool cannot shrink below the size of its largest VM. Resize that VM first.

Move VMs in and out

pool adopt --vm=<vm> --pool=<pool> moves a team VM that is outside any pool into a pool. This can take several minutes.

pool detach --vm=<vm> turns a VM in a pool into a standalone VM with its own capacity, billed by the hour. This cannot be undone. The pool keeps its capacity and billing. Detaching follows the team's standalone setting, which is off until an admin sets it (see Standalone VMs).

VMs outside a pool

Team VMs from before pools stay outside any pool until an admin moves them in with pool adopt. See Migrating to pools.

Check usage

  • pool list shows each pool's region, size, VM count, and recent CPU use.
  • pool list build --usage --range=7d shows CPU history for one pool.
  • team usage shows pool capacity, included capacity, disk, and bandwidth for the billing period. It is limited to admins and billing owners.
  • exe.dev/team/pools and exe.dev/team/usage show the same in the browser.

Delete a pool

pool delete build refuses while the pool has VMs. pool delete build --force deletes it anyway, and its VMs keep running outside any pool. Billing for the pool stops when it is deleted.

A member pool cannot be force-deleted. Delete its VMs first, and switch placement away from member-pool before deleting it.

Next steps