[Gluster-devel] Exporting capabilities and type of volume

[M. Mohan Kumar]

Hello,

Block Device (BD) xlator v2 (http://review.gluster.com/#/c/4809/) is 
hybrid xlator that handles both posix and block bricks. BD xlator
leverages existing posix fops for most of the calls and it sits above
the posix xlator.

BD xlator overloads setxattr fop to create a Logical Volume (LV) and map
it to an existing posix file. Also BD xlator supports non-standard
features such as server offloaded file copy, snapshot, creating thin
provisioned LV instead of linear LV etc.

All these capabilities are listed as part of gluster volume info (and
xml) CLI output. But still clients that are not part of peer can't run
gluster volume info to get list of capabilities to start exploiting
them. Also consumers of libgfapi also need some kind of interface to get
these capabilities and volume type.

qemu-img command is used to create VM images in KVM
environment. qemu-img already supports gluster protocol to create a VM
image backed in a GlusterFS volume. GlusterFS block driver in QEMU needs
to know type of GlusterFS volume so that it can decide if it should send
additional glfs_fsetxattr command to create a LV and map it to the posix
file. Without user.glusterfs.bd xattr, LV will not be created and BD
xlator is not exploited.

There is a need for an interface providing type of Gluster volume and
capabilities of it. Client/libgfapi applications can use this interface
and start exploiting the features accordingly. IMHO getxattr FOP can be
used to inform type and capabilities of volume. ie when getxattr is
received for root gfid with xattr names "glusterfs.type" or
"glusterfs.caps" type or list of capabilities (in integer format) can be
returned.

Is there any other better approach to inform the type of volume and
capabilities it other than getxattr interface?

Note:
Wiki page for this feature
http://www.gluster.org/community/documentation/index.php/Features/exposing-volume-capabilities