Skip to content

Snapshots FAQ

Amir Goldstein edited this page Nov 14, 2018 · 4 revisions

Table of Contents

What is the development status of overlayfs snapshots?

Which local file systems can be used with overlayfs snapshots?

  • Any local file system that can be used as overlayfs upper layer:
    • Not network file systems
    • Not FUSE file systems
    • Not another overlay file system
  • The local file system must support NFS export operations.
  • Without reflink/clone support of local file system the performance overhead of overlayfs snapshots for opening a large file for write can be significant.

Which local file system is recommended for overlayfs snapshots?

  • The reflink support requirement makes XFS with reflink support the best candidate for local file system for overlayfs snapshot.
  • Btrfs could also be used as local file system for overlayfs snapshots, but it already has built in snapshots support.
  • Ext4 and XFS without reflink support can be used with overlayfs snapshots as long as long delays on opening large files for write can be tolerated.

Differences between overlayfs and Btrfs snapshots?

  • Overlayfs snapshots work on top of another local file system, btrfs included. Btrfs snapshots are built in.
  • Overlayfs snapshots are on directory sub-tree level. Btrfs snapshot are on [sub]volume level.
  • Overlayfs shares the inode cache of unmodified files among all snapshots. Btrfs uses a separate inode cache entry per snapshot.
  • Btrfs snapshots are writable and may be nested. Overlayfs snapshots are read-only.

Differences between overlayfs snapshots and LVM thin snapshots?

  • Overlayfs snapshots are on directory sub-tree level. LVM thin snapshot are on volume level.
  • Overlayfs snapshots use available local file system free space. LVM uses thinp pool blocks.
  • Overlayfs snapshots use local file system block allocators, which are aware of underlying disk geometry.
  • Overlayfs shares the inode cache of unmodified files among all snapshots. LVM uses a separate inode cache entry per snapshot.
  • LVM thin volumes mangle the underlying block layout, which may lead to fragmentation and performance degradation over time.
  • Overlayfs snapshot should have negligible performance overhead on file system operations.
  • LVM thin volumes incur copy on write cost per block write, because they do not work with native file system block size.

Can I take overlayfs snapshots for a test drive?

  • Yes. You can.
  • See Snapshots HOWTO.
  • More convenient installer for standalone module will be provided further down the road.

How can I contribute to overlayfs snapshots project?