Snaplane

Snaplane is a Kubernetes operator for block-volume backup and restore.

Overview

• BackupPolicy schedules and dispatches backups
• Backup represents one transfer execution
• VolumeSnapshot is the consistency point and queue carrier
• changed-block input is modeled through CBT providers
• writer sidecars store backup data on node-local storage under /var/backup
• restore is requested through PersistentVolumeClaim.dataSourceRef -> Backup
• restore integration depends on VolumePopulator

Components

• controller manager
  • runs the BackupPolicy and Backup controllers
• writer sidecar
  • runs on backup-target nodes
  • exposes mTLS gRPC write APIs
  • reports endpoint and capacity through Lease annotations
• restore populator
  • registers sourceKind=Backup
  • restores mock-image-v1 and cas-v1 backups

Current Status

Implemented:

• v1alpha1 BackupPolicy and Backup APIs
• strict serial dispatch with queue metadata on VolumeSnapshot
• destination node pinning on source PVCs
• writer-sidecar mTLS path
• retryable backup execution
• restore populator flow
• CAS repository publish and restore
• CAS compaction execution in writer-sidecar maintenance
• dependency-safe CAS manifest pruning
• SnapshotMetadata-backed range collection and live snapshot-reader data path

Open work:

• no repo-local TODOs are currently tracked; external environment work is captured by the e2e gate documentation

Track open work in TODO.md.

Quick References

• architecture baseline: ARCHITECTURE.md
• retained design docs: docs/design-docs/index.md
• open work: TODO.md
• e2e lanes: test/e2e/README.md
• release runbook: docs/runbooks/release.md