preventing deadlocks in snapshot directories - unexplained

Eric Anderson anderson at
Fri Jan 13 13:21:06 PST 2006

user wrote:
> I'm sorry - I misspoke in my original post, and I don't think I made clear
> what it was I was witnessing.
> What I am saying is, I have snapshots that exist _two_ directories deep
> off of the mount point, such as:
> /.snap/snap_directory1/snapshot
> /.snap/snap_directory2/snapshot
> and what I am seeing is, if I:
> cd /.snap    (note, no snapshots in here)
> rm -rf snap_directory1/   (note, deleting a dir containing snapshot)
> What happens is, other operations inside of /.snap lock up.  Which is
> contradictory to what I expected, and to what you have explained thus far
> in response to my original post (many thanks for those responses).  Both
> you and I seem to expect that operations would only be locked for the
> /.snap/snap_directory1/ directory.
> I am going to try to reproduce this ... 
> The "solution" I proposed was :
> /.snap/snapshot
> /.snap2/snapshot
> /.snap3/snapshot
> Which isn't much different, but it prevents the existence of a
> non-snapshot directory that could accidently be put into use, or be a pwd,
> or otherwise statted while a snapshot is deleted, etc.
> Any thoughts / comments ?

What happens if you do:

rm -f snap_directory1/snapshot
rmdir snap_directory1

Does the same thing happen?


Eric Anderson        Sr. Systems Administrator        Centaur Technology
Anything that works is better than anything that doesn't.

More information about the freebsd-fs mailing list