Skip to content

Set limits per person#

On a shared pool of GPUs, one person can start enough environments and notebooks to hold every GPU. A workspace's limits per person stop that: how many environments a person may have running at once, and how many GPUs their environments and notebooks' runtimes may hold together.

Before you begin#

  • The admin role in the workspace to set them. Everyone in the workspace can read them; the console shows them in New environment.
  • For the API tabs, ASTRAEUS_TOKEN and API as in Create and change a runtime.

Set them#

  1. Open Hesperus → Environments and press Limits.
  2. Environments at once (Shells and IDEs running, per person) and GPUs at once (Across a person's environments and notebooks): whole numbers, 0 for no limit.
  3. Press Save. The list then says This workspace allows 2 environments running and 4 GPUs per person.

astra has no command for the limits. Use the console or the API.

$ curl -fsS -X PUT "$API/notebook-limits/default" -H "Authorization: Bearer $ASTRAEUS_TOKEN" \
    -H 'content-type: application/json' \
    -d '{"max_environments_per_person": 2, "max_gpus_per_person": 4}'
{"max_environments_per_person":2,"max_gpus_per_person":4,"updated_by":"user:7c1e2f4a-93b0-4d51-a6f2-0e8d3b9c1a55","updated_at":"2026-10-02T09:00:00Z"}
$ curl -fsS "$API/notebook-limits/default" -H "Authorization: Bearer $ASTRAEUS_TOKEN"

Field Default Description
max_environments_per_person 0 (no limit) Environments (shell and IDE) a person may have running at once. At most 1000.
max_gpus_per_person 0 (no limit) GPUs a person's runtimes — environments and notebooks' — may hold at once. At most 10000.

The name is always default. A non-admin's PUT is refused.

How they are applied#

  • When a runtime is made or started — an environment created or started, a notebook opened — against what its owner already has running (Pending, Starting or Ready). An admin who makes an environment for someone counts it against that person.
  • Over a limit is refused with 409 HESPERUS_LIMIT, naming what to stop:
    • this workspace allows 2 environments running at once per person, and its owner has 2 (dev, code): stop one first
    • this workspace allows 4 GPUs per person across their environments and notebooks, and its owner holds 3 (dev): this one asks for 2; stop one first or ask for fewer
  • Nothing running is stopped when you lower a limit; it applies to the next start.
  • It is a fair share, not a quota. Two starts at the same instant may both pass. The workspace's quota on the cluster, which the queue enforces for every run, is the hard ceiling (Quotas, pools and terms).