Struct SnapshotCatalog 

pub struct SnapshotCatalog { /* private fields */ }

Manages the on-disk snapshot catalog for all VMs.

Layout:

{data_dir}/snapshots/{vm_id}/{snapshot_id}/
    vmstate
    mem          (full only)
    meta.json
{data_dir}/snapshots/{vm_id}/.{snapshot_id}.partial/   (being written)

Every catalogued directory carries a readable meta.json, because a snapshot is written into a .partial staging directory and renamed in as a whole (see PendingSnapshot). Readers therefore treat a missing or unparseable record as corruption rather than as a normal lifecycle state.

Implementations

impl SnapshotCatalog

pub fn new(data_dir: &str) -> SnapshotCatalog

Create a new catalog rooted at {data_dir}/snapshots.

pub fn begin(&self, vm_id: &str) -> Result<PendingSnapshot<'_>, SnapshotError>

Start a new snapshot for vm_id.

Creates the staging directory; the snapshot enters the catalog only on PendingSnapshot::commit.

pub fn list(&self, vm_id: &str) -> Result<Vec<SnapshotInfo>, SnapshotError>

List all snapshots for a VM, sorted by creation time (oldest first).

pub fn get( &self, vm_id: &str, snapshot_id: &str, ) -> Result<SnapshotMeta, SnapshotError>

Look up a single snapshot by ID.

pub fn delete( &self, vm_id: &str, snapshot_id: &str, ) -> Result<(), SnapshotError>

Delete a snapshot directory from disk.

pub fn sweep_incomplete(&self)

Remove staging directories left behind by a crash mid-checkpoint.

A committed snapshot is renamed into place atomically, so anything still staged when the process died is unfinished by definition — and can be holding a full memory dump.

pub fn find_by_id( &self, snapshot_id: &str, ) -> Result<SnapshotMeta, SnapshotError>

Find a snapshot by ID alone, searching across all owner directories.

pub fn list_all(&self) -> Result<Vec<SnapshotInfo>, SnapshotError>

List all snapshots across every owner directory, sorted by creation time.

pub fn referenced_rootfs_paths( &self, ) -> Result<BTreeSet<PathBuf>, SnapshotError>

Rootfs images that catalogued snapshots depend on.

Restore rebuilds the dm-snapshot origin from SnapshotMeta::rootfs_path, and the guest memory in the snapshot was captured against those exact bytes — re-converting the same image with a different vm-agent produces a different file and is not a substitute. Anything that garbage-collects rootfs images must treat these as pinned.

pub fn delete_by_id(&self, snapshot_id: &str) -> Result<(), SnapshotError>

Delete a snapshot knowing only its ID (searches across all owner directories).