Services
A stateful container, a workspace, and an explicit lifecycle.
Service lifecycle
| State | What it means |
|---|---|
queued | An operation has been accepted for processing. |
starting | Resources or the container are starting. |
running | The runtime reports the service running. Check your application's readiness separately. |
sleeping | Compute is stopped; provisioned storage can remain. |
stopping | Resource release is in progress. |
resizing | The service is moving to a different resource configuration. |
deleting | Service and service-owned storage cleanup is in progress. |
deleted | The service has been deleted. |
error | Inspect service error information and events before retrying. |
Actions
Start/wake requests an allocation. Stop/pause requests release while retaining storage. Restart replaces allocated compute with the same configuration and workspace. Resize changes resources and can restart processes. Delete removes the service and its service-owned workspace; independent storage survives.
Keep valuable files under /workspace and back them up. None of these actions promises process-memory checkpointing.
Revisions and retries
Configuration updates are revision-checked so a stale edit cannot silently overwrite a newer one. Read the current service before editing it. Lifecycle changes require an idempotency key for the same logical operation.
After a timeout, read the service and its events before retrying. Reuse the original key only for an identical retry. API request conventions.
Related guides
Configuration · Restart · Sleep and activity · Endpoints · Logs and commands