Re: Plans for 1.8+ (2.0?)

["Stilyan Tsenov" <naylits@xxxxxxxxx>]

2007/2/1, Thomas E. Spanjaard <tgen@netphreax.net>
:> A snapshot of a filesystem needs support from the filesystem to be
> snapshottable, e.g. temporary suspension of all activities on that
> filesystem while the snapshot-taking is in-progress.

Let suppose the new DragonFly filesystem will support filesystem
snapshots.I try to imagine how that will look to the user.The closest
technologies  we now have are union and null filesystems.The
snapshotting filesystem will not be a single hierarchical namespace
(directory tree),but more like with union(null) mounted parts of the
tree,that looks like it is a single tree.Timestamped layers of the
filesystem namespaces under the active namespace.
That was the story considering an alone DF system.When I add it to a
networked cluster ,with nodes which are synchronizing as much of the
data as they can,I conclude that the whole cluster is supposed to
operate as there is only one single filesystem namespace(the cluster
filesystem namespace),and snapshots of older filesystems are only used
for archiving and synchronization purposes.Am I right ?At least on the
conceptual level?
What about the data I don't want to share on that cluster
namespace?Should I be able to use UFS or every other filesystem
storing it ?Generally  are we to be locked in this single clusterable
filesystem solution ?