compute

POST /v1/compute/servers/{serverId}/resize

Move a server onto another machine size.

All compute endpoints

All developer docs →

Authentication

Send an API key as a bearer token. The key must carry the sites.view permission; a key without it is refused with 403, not 404.

This endpoint takes no organisation id. Your key already identifies the organisation it belongs to, and the response is scoped to it.

Try it

Replace anything in angle brackets with your own values, and the key placeholder with a key from your dashboard.

curl -X POST https://api.zinndigital.com/v1/compute/servers/{serverId}/resize \
  -H "Authorization: Bearer zdk_live_…" \
  -H "Content-Type: application/json" \
  -d '{  }'

Signed in? The API console in your dashboard fills in your real organisation id and your own key, and runs the request against the live API so you can see the actual response. Open this endpoint in the API console

Details

Performs the upgrade and returns the server's new state. ⛔ This charges immediately at every provider we support, so it is only ever reached from a quote the customer has read (quoteComputeServerResize). The engine attaches an idempotency key derived from the server and the target size, so a retry cannot buy the resize twice. ⛔⛔ A size whose disk is smaller than what the machine is already using is refused with a 422, and that is a data-loss gate rather than a tidiness check. The provider's endpoint is an upgrade endpoint and usually refuses a shrink, but "usually" is not a promise to make about a customer's files. Before a shrink the engine reads the machine's live disk usage and requires the target to be strictly larger. It fails closed: if the provider does not answer, or answers -1 ("did not say"), the resize is refused rather than attempted — -1 < anything would otherwise read as "plenty of room". Growing is never gated, never pays for the extra read, and never fails because usage statistics were unavailable. RAM is not gated: running with less memory loses nothing on disk and is the customer's call. 404 for a server that is not the caller's; 422 for a size the provider does not offer, a machine that is not active — an order still running would have two writers, and a failed or released one has nothing to write to — or a shrink below the disk in use. Requires sites.view and billing.payment.manage — ⛔ not sites.create, which is what this was until #1682.

Parameters

NameTypeRequiredWhat it is
serverId (path)UuidYesThe server's id, as listComputeServers reports it. Ours (UUIDv7), minted when the order row was written — never the provider's own identifier for the machine.

Request body

NameTypeRequiredWhat it is
specstringNoThe target size, as getComputeCatalogue reported it.
plan_codestringNoThe plan to move to, as getComputeCatalogue's tiers[].plan_code.

Response

NameTypeRequiredWhat it is
idUuidYesOurs, and also the provisioning workflow's id and the machine's name at the provider — which is what makes the whole purchase path idempotent.
namestringYesWhat the customer called it. ⛔ At the provider a machine's name is its row id; confusing the two is a cross-tenant defect, so this is the customer's string and nothing else.
deploy_targetstringYesWhich compute range the machine belongs to.
plan_codestringYesThe tier it was bought on, or "" when it was ordered by size alone.
providerstringYesOur provider id.
specstringYesThe machine size it currently runs.
zonestringYesThe data centre it runs in.
statusstring<pending, provisioning, active, failed, suspended, terminated, resizing, terminating>YesHow far the order got — ours, and a different fact from power_state. pending is a row with nothing bought and nothing charged; failed is the state whose message is the…
messagestringYesWhy the order failed, in a sentence the customer reads. Blank unless failed.
vendor_statusstringYesThe provider's own word, verbatim and unrounded. Staff-facing detail; a customer client renders status and never this.
power_statestring<unknown, on, off, starting, stopping, rebooting>YesWhat the machine is doing right now. ⛔ Six values, not two, and the four extra ones are the point: a machine mid-transition is not "on". unknown is the honest answer when…
addressesstring[]YesPublic addresses if the provider states them. Empty means not stated, never "none" — a machine rendered as having no address reads as broken.
optimisationstringYesThe stack optimisation profile it was built with, or "" for none.
term_monthsintegerYesThe term it was bought on, or null when the provider states none.
renews_atstringYes⛔ Always null today, and honestly so: the provider publishes no renewal date, and one computed from the order plus the term would be our arithmetic rendered as the provider's…
created_atstringYesWhen the row was created, or "" when not stated.
terminated_atobjectNoWhen the machine was terminated; null while it exists. Owner decision D11 (2026-10-06): a terminated server leaves the servers list and the last five from the past 7 days appear…
operablebooleanYesWhether the customer may act on this machine — power, resize, renew, rebuild. ⛔ Read rather than re-derived from status: the engine refuses every one of those on anything but…
imagestringYesThe provider image the machine was built from. Blank on the resold range, which installs one managed image and offers no choice.
suspendedbooleanYes⛔⛔ Whether WE stopped it, which is not the same fact as power_state. A stopped machine is the customer's own choice and they may start it again; a suspended one is ours and…
suspension_reasonstringYesWhy, verbatim, so the customer reads the same sentence support wrote. Blank when the machine is in good standing.
capabilitiesstring[]Yes⭐⭐ What this machine's provider can actually do — capability tokens such as compute:snapshots, compute:console, compute:firewall, compute:backups_manage and…
billingComputeServerBillingYesHow a machine is paid for (W43-47, docs/687). term machines are charged ahead by a per-machine subscription. metered machines are charged per started hour from the account…
login_userstringYesThe account to sign in as (W43-49, docs/689). admin — key-only, passwordless sudo, root login refused — on a Zinn® Cloud machine built under the access model; root on one…
blueprintstring<, docker, k3s, k3s_argocd>YesThe software stack installed at first boot. Blank for none.
cluster_idstringYesThe cluster this machine is a node of, or blank.
cluster_rolestring<, server, agent>Yes—

Errors this endpoint can return

401 · 403 · 404 · 422 · 429 · 503