(bugfix) Bound snapshot storage metadata scans #100

Closed
opened 2026-06-08 23:16:14 +02:00 by parkel · 0 comments
Owner

Snapshot storage metadata is currently collected by walking the snapshot data tree after backup completion. On very large backup targets this can be slow and memory-heavy because the scan tracks allocated inodes for hardlink-aware accounting. This can make worker finalization fail or OOM after rsync has already completed.

Acceptance criteria:

  • Snapshot storage metadata collection has a bounded scan mode used by backup workers.
  • Partial measurements clearly mark when they were limited/truncated.
  • Existing exact behavior remains available for small/test/direct usage where appropriate.
  • Backup completion should not OOM solely because a snapshot has too many files.
  • Tests cover scan limits and metadata flags.

Related feature/design follow-up: #98.

Snapshot storage metadata is currently collected by walking the snapshot data tree after backup completion. On very large backup targets this can be slow and memory-heavy because the scan tracks allocated inodes for hardlink-aware accounting. This can make worker finalization fail or OOM after rsync has already completed. Acceptance criteria: - Snapshot storage metadata collection has a bounded scan mode used by backup workers. - Partial measurements clearly mark when they were limited/truncated. - Existing exact behavior remains available for small/test/direct usage where appropriate. - Backup completion should not OOM solely because a snapshot has too many files. - Tests cover scan limits and metadata flags. Related feature/design follow-up: #98.
parkel added the area: observabilitybugfix labels 2026-06-08 23:16:14 +02:00
Sign in to join this conversation.