Skip to content

Compute lifecycle

Understand compute initialization, running state, suspension and sizing.

Updated View as Markdown

A compute runs PostgreSQL for a branch. Projects, branches and computes are separate resources; compute operations do not change project ownership.

States and actions

State or action Meaning and next step
creating Initialization is in progress. Wait and refresh; do not treat it as ready or as the start of billing.
running Initialization has completed according to the API. Verify a connection to the intended database before using it from an application.
Suspend Stops the compute; applications may lose query access. Read the impact before confirming.
Resume Starts a suspended compute. Wait for completion before connecting.
Restart May interrupt current sessions. Verify the connection afterwards.
Delete compute Deletes the compute runtime and metadata. It is not a substitute for deleting a project or creating a backup.

Change compute size

Select the intended compute and a CU value allowed by the plan. CU is a compute sizing unit, not a currency or CPU percentage. The MCP source accepts 0.25–8 CU; the plan, API and available resources can impose further limits.

Verify autoscaling, pooler and timeout behavior against your deployment. A configuration field does not establish complete runtime behavior. See limitations.

A create request returns 202 or 409

  • 202: accepted, with initialization still in progress. Read the compute state instead of immediately creating a replacement.
  • 409: a conflict, such as a branch that already has a compute or is being deleted. Check existing resources before retrying.

One compute per branch is a new source contract requiring a separate migration and rollout. These docs do not confirm that production has completed either.

Navigation

Type to search…

↑↓ navigate↵ open resultEsc close