Recover provisioning work

Find an unfinished provisioning job and retry failed guidance without replacing successful sets.

Save the provisioning job's id and engineId as soon as creation succeeds. Use GET /v2/provisioning/jobs/:id after a disconnected client or an uncertain wait. The job and per-kind states determine what needs recovery.

Find running work #

http
GET /v2/organizations/:orgId/provisioning/jobs?status=running
X-API-Key: YOUR_API_KEY

status=running is required. The response contains data, an array of visible running jobs, without cursor pagination. Add startedBy=me to restrict it to jobs started by the authenticated user. A service key has no user identity, so that filter returns an empty list for it.

If a new-engine create response was lost, also inspect GET /v2/organizations/:orgId/engines and the dashboard for the created engine. A job may already have completed and disappeared from the running list. Repeating the create request can create another engine.

Retry failed kinds #

http
POST /v2/provisioning/jobs/:id/retry
X-API-Key: YOUR_API_KEY

This request has no body. It returns a new job on the same engine, using the original sources. Only failed kinds that are still empty are selected; completed sets are retained. Save the returned job ID and follow it to completion.

NothingToProvision (409) means no failed kind remains eligible, for example because another action has since attached a set. ProvisioningJobRunning (409) identifies work already running on that engine. Read those resources before retrying.

If the original sources need changing, submit a new job on the existing engine with corrected sources and a deliberate Empty or Requested selection. The retry route reuses the original input.

Resolve limits and incomplete starts #

ProvisioningLimitReached (429) includes retryAt. Wait until that time before submitting again. Credit, plan, and access failures need the corresponding account or permission correction.

A failed workflow launch can leave a new empty engine behind even when the request returns a server error. Inspect the engine list before repeating new-engine creation. Reuse the existing engine when it is the one you intended to configure.

Once the job settles, review the generated guidance in Studio before translating with it.