[Gluster-devel] Get GFID for a file through libgfapi
Niels de Vos
ndevos at redhat.com
Tue Jan 10 10:17:28 UTC 2017
On Tue, Jan 10, 2017 at 03:05:06PM +0530, Shyam wrote:
> On 01/10/2017 02:55 PM, Niels de Vos wrote:
> > On Mon, Jan 09, 2017 at 08:58:07PM +0000, Ankireddypalle Reddy wrote:
> > > Neils,
> > > Thanks for pointing to the sample code. The use case is to use GFID as a unique key for maintaining indexing information about a file when it is being backed up from GlusterFS storage. Extending GFAPI to extract GFID for a file would be great.
> > > As per the example If I need to find GFID for a path p1/p2/p3/p4 on a glusterfs volume then should I do a look up for every level?
> > > LOOKUP (/)->LOOKUP(p1)-> LOOKUP(p2)-> LOOKUP(p3)-> LOOKUP(p4)
> >
> > No, that is not required. You can use glfs_h_lookupat() with the full
> > path. Note that glfs_h_extract_handle() just does a memcpy() of the GFID
> > into the given (unsigned char*), the format is of 'uuid_t'.
>
> The handle is an abstraction and hence we do not identify it as a GFID
> per-se, isn't asking to assume the handle is the GFID breaking this
> abstraction?
>
> IOW, if we publish this, then the handle is always the GFID and vice-versa,
> are we fine with this?
I see this as a possible workaround until we offer a way to get the GFID
through real function calls. Doing it like this is rather hacky, the
same would be true for doing glfs_getxattr().
What the most suitable API to get a GFID is, I do not know yet. At the
moment I am pondering about a glfs_xstat()/glfs_fxstat() that can get
multiple flags for information to resolve. This would then fill an
'struct glfs_xstat'. Accessor methods for this struct can then return
the GFID or any other extensions (location/replication awareness?).
(idea based on https://lwn.net/Articles/394298/ )
Feedback most welcome :)
Thanks,
Niels
>
> >
> > Attached is the modified test that shows the UUID without the need for a
> > lookup of each component of the directory (a LOOKUP will be done by
> > gfapi if needed).
> >
> > $ make CFLAGS="-lgfapi -luuid" resolve
> > cc -lgfapi -luuid resolve.c -o resolve
> > $ ./resolve storage.example.com media resolve.log
> > Starting libgfapi_fini
> > glfs_set_volfile_server : returned 0
> > glfs_set_logging : returned 0
> > glfs_init : returned 0
> > glfs_set_volfile_server : returned 0
> > glfs_set_logging : returned 0
> > glfs_init : returned 0
> > glfs_h_extract_handle : returned 0
> > UUID of /installation/CentOS-7-x86_64-Everything-1503-01.iso: b1b20352-c71c-4579-b678-a7a38b0e9a84
> > glfs_fini : returned 0
> > End of libgfapi_fini
> >
> > $ getfattr -n glusterfs.gfid -ehex /lan/storage.example.com/media/installation/CentOS-7-x86_64-Everything-1503-01.iso
> > getfattr: Removing leading '/' from absolute path names
> > # file: lan/storage.example.com/media/installation/CentOS-7-x86_64-Everything-1503-01.iso
> > glusterfs.gfid=0xb1b20352c71c4579b678a7a38b0e9a84
> >
> > HTH,
> > Niels
> >
> >
> > >
> > > Thanks and Regards,
> > > Ram
> > >
> > >
> > > -----Original Message-----
> > > From: Niels de Vos [mailto:ndevos at redhat.com]
> > > Sent: Monday, January 09, 2017 3:39 PM
> > > To: Ankireddypalle Reddy
> > > Cc: Gluster Devel (gluster-devel at gluster.org); integration at gluster.org
> > > Subject: Re: [Gluster-devel] Get GFID for a file through libgfapi
> > >
> > > On Mon, Jan 09, 2017 at 05:53:03PM +0000, Ankireddypalle Reddy wrote:
> > > > Hi,
> > > > I am trying to extract the GFID for a file through libgfapi interface. When I try to extract the value of extended attribute glusterfs.gfid through libgfapi I get the errorno: 95. This works for FUSE though. Is there a way to extract the GFID for a file through libgfapi.
> > >
> > > It seems that this is a case where FUSE handles the xatts special. The glusterfs.gfid and glusterfs.gfid.string (VIRTUAL_GFID_XATTR_KEY and
> > > VIRTUAL_GFID_XATTR_KEY_STR) are specifically handled in xlators/mount/fuse/src/fuse-bridge.c.
> > >
> > > There is a way to get the GFID, but it probably is rather a cumbersome workaround for you. The handle-API is used heavily by NFS-Ganesha (because NFS uses filehandles more than filenames), and extracts the GFID from the 'struct glfs_object' with glfs_h_extract_handle(). A basic example of how to obtain and extract the handle is in https://github.com/gluster/glusterfs/blob/master/tests/basic/gfapi/bug1291259.c
> > >
> > > Could you explain the need for knowing the GFID in the application? We can extend gfapi with fetching the GFID if that would help you.
> > >
> > > Niels
> > >
> > >
> > > PS: we have a new integration at gluster.org where external projects can ask gfapi related questions. The gluster-devel list tends to be a little heavy on traffic for non-Gluster developers.
> > > ***************************Legal Disclaimer***************************
> > > "This communication may contain confidential and privileged material for the
> > > sole use of the intended recipient. Any unauthorized review, use or distribution
> > > by others is strictly prohibited. If you have received the message by mistake,
> > > please advise the sender by reply email and delete the message. Thank you."
> > > **********************************************************************
> > >
> > >
> > >
> > > _______________________________________________
> > > Gluster-devel mailing list
> > > Gluster-devel at gluster.org
> > > http://www.gluster.org/mailman/listinfo/gluster-devel
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL: <http://www.gluster.org/pipermail/gluster-devel/attachments/20170110/ce0d796e/attachment-0001.sig>
More information about the Gluster-devel
mailing list