I can restore a snapshot, but snapshotting brings my DIR backend Containers down, copying does not, so I copy to another container as a nightly backup.
Oh, I think I see why snapshots are slower on dir than copies, that’s a bug, we should make copies be slow too
Looking into fixing that now.
So there is no way (besides Criu which does not work) to backup a container and it still be live while being copied/snapshotted?
Even with CRIU we would freeze it if your backend is dir or ceph.
There is no way to atomically copy an entire filesystem tree on those backends, so the only way to prevent anything from changing mid-snapshot (or mid-copy) is to entirely freeze all processes until the snapshot/copy is over.
All other storage backends (zfs, btrfs, lvm, cephfs) do support atomic snapshotting and so do not need such a temporary freeze.