Capacity of each environment

Optimize > Cost and savings (Preview) lists the Fabric capacity each environment's workspace is on, and each environment card in Data mesh > Environments shows the same line. It answers one question before you run something in Fabric: is the capacity there and running?

Everything is read only, with your own sign-in. Nothing changes a capacity, and nothing asks you to sign in or grant access. Work on this PC never needs any of it.

What it shows#

Field	Where it comes from
Capacity name, SKU, region and state	Fabric's capacity list (`GET /capacities`), for capacities you can see
Capacity id, region and assignment progress	The workspace (`GET /workspaces/{id}`), when Fabric sends them
Read	When it was read. Readings are kept for an hour; older ones are marked Stale until you press Refresh

Only fields Fabric actually returned are shown. An SKU the app doesn't know has no CU size, rather than a guessed one.

Running is not the same as having room#

Running means Fabric reports the capacity as Active: work can start. It doesn't say how busy the capacity is; a running capacity can still slow down or delay jobs when it is overloaded. How busy a capacity is shows in Fabric's Capacity Metrics app to capacity admins. There is no public per-capacity utilization API for ordinary users, so the app shows utilization as unknown, never as 0.

Not running means the capacity is Inactive: paused, suspended or an expired trial. Fabric runs, schedules and new lakehouses fail until an admin resumes it. Runs on this PC still work.

When it can't be read#

Shown	Means	What to do
The capacity isn't visible to you	Fabric only lists capacities you administer or may assign workspaces to. The workspace still runs on it.	Ask your capacity admin for its status, or check the Capacity Metrics app.
Capacity access not granted	Your sign-in hasn't accepted `Capacity.Read.All`. The capacity id still shows.	Sign in again and accept it, or ask IT. See Permissions.
You can't open the workspace	Fabric refused the workspace (401 or 403).	Ask a workspace admin for access, or assign a workspace you can open.
The workspace doesn't exist any more	Fabric returned 404.	Assign another workspace in Data mesh > Environments.
The workspace isn't on a capacity	The workspace has no capacity.	Ask a workspace admin to assign one (Workspace settings > License info).
Signed out, or Couldn't reach Fabric	No sign-in, or no network.	The last good reading stays, with its age. Work on this PC carries on.

Each environment is read on its own, so one failure doesn't hide the others. When Fabric can't be reached, the app stops trying after the first environment instead of waiting on each.

For developers#

sidecar/fabric_desktop/capacity.py builds the observations; capacity_health(refresh=False) in estate_service.py is the engine call (read only, audited as a read). The cost report includes the same document as capacities, and the capacity share per environment uses it. Readings are cached in capacity_observations.json in the app's data folder. FabricError.status now carries the HTTP status so "no access" and "not found" can be told apart.