vfs_shadow_copy2 configuration

Hemanth Thummala hemanth.thummala at gmail.com
Thu Sep 12 14:41:43 CEST 2013


Hi Dan/Michael,

Does shadow_copy2 support snapshots in non-default format? We have
snapshots in .snapshot directory as hourly.0,hourly.1..
weekly.0,weekly.1...monthly.0.. yearly.0 etc.
I have tried the following configuration for our box.

        vfs objects = shadow_copy2
        shadow:sort = desc
        shadow:snapdirseverywhere = yes
        shadow:format = hourly.%H
        shadow:snapdir = .snapshot

With this we are able to get the list of labels in @GMT format with just
hours getting filled properly and rest with default values.
for example: @GMT-1900.00.00-09.00.00 which corresponds to my hourly.9
snapshot.

Where as later create checks getting failed with object not found as client
is unable to map this @GMT token to the actual hourly.X snapshot.

Is there any way we can work with these kind of snapshots? Or we should
have the snapshots always in time stamp based in the file system to make
previous versions to work?


Thanks,
Hemanth.


On Sun, Sep 8, 2013 at 9:31 PM, Dan Cohen1 <DANCOH at il.ibm.com> wrote:

> Hi Michael,
>
> I had another issue with shadow copy2: even when using relative paths
> restoring of files from the snapshots failed.
> Windows was giving me error messages like this one: "Could not find this
> item, This is no longer located in
> \\<machine_name>\<share_name>\<share_name>. Verify item's location and try
> again". Obviously the second <share_name> in the giving path is needless.
> I started to investigate it and kind of arrived to a dead end. So I
> decided to take your branch with your work on shadow copy2 vfs to see if
> it solves my issues and to give you some feedback as you asked me.
>
> The good news are that when eventually I've succeeded to apply your
> changes to vfs_shadow_copy2.c everything worked as expected.
> I applied the changes on top of branch based on samba 4.0.4. We use GPFS
> as the file system and I tried two configurations for shadow copy2:
> Configuration 1:
> shadow:snapdir = <absolute_path_to_snapshot_directory>
> shadow:basedir= <our base dir>
> shadow:snapdirseverywhere= no
>
> and configuration 2:
>
> shadow:snapdir = .snapshots
> shadow:snapdirseverywhere = yes
>
> both configurations worked perfectly, previous versions were presented by
> the windows tab and files/directories were restored correctly from the
> previous versions using the windows interface.
> So at least for the configurations I tested your work seems to solve the
> issues, including absolute path issue.
>
> Only problem was that I couldn't really fetch/clone your git repository to
> my host, so I had to manually merge the file vfs_shadow_copy2.c and fix
> some mismatches with my branch (using copy_smb_filename instead of
> cp_smb_filename).
> When I'm trying to clone/fetch your branch I'm getting the following error
> message:
> -----------------------------------
> dancoh at host176:/local/dancoh$ git clone
> http://gitweb.samba.org/obnox/samba/samba-obnox.git
> Cloning into 'samba-obnox'...
> error: Unable to find bc07864b77e10f1f231f9177ba168c222d0c555d under
> http://gitweb.samba.org/obnox/samba/samba-obnox.git
> Cannot obtain needed object bc07864b77e10f1f231f9177ba168c222d0c555d
> while processing commit d65db48db51006a766a8f58a53e7c1022e6c630c.
> error: Fetch failed.
> -------------------------------------
>
> I tried several times, from several machines,  with no success. Cloning
> the samba.git repository is obviously working fine for me.
> Do you have any idea what could be the problem? I guess I'm missing
> something... Should I also clone some other repo?
>
> Thanks,
> Dan Cohen
> IBM - XIV, Israel
> NAS Development Team
>
>
>
>
> From:   Dan Cohen1/Israel/IBM at IBMIL
> To:     Michael Adam <obnox at samba.org>,
> Cc:     samba-technical at lists.samba.org
> Date:   22/08/2013 15:09
> Subject:        Re: vfs_shadow_copy2 configuration
> Sent by:        samba-technical-bounces at lists.samba.org
>
>
>
> Hi Michael,
>
> Thanks for you answer.
> When you have this polished version of the patches I'll be happy to test
> it and give you some feedback (if I'll have anything clever to say...).
> The raw patches as they are now are a bit too messy for me to look/merge
> into my branch.
>
> Thanks,
> Dan
>
>
>
>
> From:   Michael Adam <obnox at samba.org>
> To:     Dan Cohen1/Israel/IBM at IBMIL,
> Cc:     samba-technical at lists.samba.org
> Date:   22/08/2013 13:17
> Subject:        Re: vfs_shadow_copy2 configuration
>
>
>
> Hi,
>
> I have recently worked on getting the shadow_copy2
> module into shape again for absolute snapdir
> and a couple of other things that worked in 3.6.
>
> The raw result of my work is here:
>
>
> http://gitweb.samba.org/?p=obnox/samba/samba-obnox.git;a=shortlog;h=refs/heads/master-shadow_copy2
>
>
>
> This is on top of master, but should apply to 4.0.
>
> The patches are very raw and show the progress when
> I approached the solution. They also contain some
> cleanup of the code.
>
> I am working on a polished/squashed version of the patches
> for pushing upstream and will post this here very soon.
>
> Any testing and feed-back would be highly welcome!
>
> Cheers - Michael
>
>
> On 2013-08-22 at 12:29 +0300, Dan Cohen1 wrote:
> > Hi,
> >
> > I had some problems with vfs_shadow_copy2 where I couldn't see my
> > snapshots in the windows "Restore previous versions"  tab.
> > After some debugging I realized that the shadow:snapdir parameter does
> not
> > support absolute paths anymore.
> > I saw this post
> https://lists.samba.org/archive/samba/2013-May/173098.html
> > regarding this issue but this still raise the question: is this a bug or
>
>
> > is it an intended behavior?
> > I couldn't find any ticket regarding this issue in the bugzila and this
> is
> > not documented in the man page or in the code. Actually the comment in
> the
> > head of the vfs_shadow_copy2.c file still states that absolute paths are
>
>
> > supported with shadow:snapdir.
> > Also, in the same matter, after looking at the code it seems  that
> > shadow:basedir is not in use anymore. Again this is not documented. And
> > again my question: is this a bug or is in purpose? (or am I missing
> > something?)
> >
> > Thanks,
> > Dan Cohen
> > IBM - XIV, Israel
> > NAS Development Team
>
> [attachment "signature.asc" deleted by Dan Cohen1/Israel/IBM]
>


More information about the samba-technical mailing list