Checkpoints

A checkpoint is a named pointer to a specific version of an instance's disk manifest. Think of it like a git tag: cheap to create, holds the state in place, and lets you roll back to it later.

Checkpoints are distinct from snapshots — a checkpoint captures the disk manifest version only, while a snapshot captures full VM memory + disk for fast resume. Both live under an instance.

Endpoints

• GET /v1/instances/:id/checkpoints — list checkpoints on an instance
• POST /v1/instances/:id/checkpoints — create a checkpoint at the current state
• GET /v1/instances/:instance_id/checkpoints/:checkpoint_id — read one
• DELETE /v1/instances/:instance_id/checkpoints/:checkpoint_id — delete

Rolling back

Roll an instance back to a prior checkpoint via POST /v1/instances/:id/rollback. The instance's current_manifest_version is swapped atomically; running VMs are restarted on the older manifest.

Related

• Instances
• Checkpoints concept — how manifest versioning works