[Gluster-users] Parallel cp?

[Brian Candler]

On Mon, Feb 06, 2012 at 04:39:15PM +0530, Anand Avati wrote:
> > Is there something comparable a client can do on the mountpoint to get a
> > list of the component bricks in a volume?
> 
> You can do the getfattr on a directory (say the mount point directory)
> and get the list of all bricks.

Thanks, although the values it gives don't map to the brick paths on the
files themselves:

root at storage1:~# getfattr -n trusted.glusterfs.pathinfo /mnt/test-volumegetfattr: Removing leading '/' from absolute path names
# file: mnt/test-volume
trusted.glusterfs.pathinfo="(test-volume-dht-layout (test-volume-client-2 0 357913940) (test-volume-client-3 357913941 715827881) (test-volume-client-4 715827882 1073741822) (test-volume-client-5 1073741823 1431655763) (test-volume-client-6 1431655764 1789569704) (test-volume-client-7 1789569705 2147483645) (test-volume-client-8 2147483646 2505397586) (test-volume-client-9 2505397587 2863311527) (test-volume-client-0 2863311528 3221225468) (test-volume-client-1 3221225469 3579139409) (test-volume-client-10 3579139410 3937053350) (test-volume-client-11 3937053351 4294967295))"

root at storage1:~# getfattr -n trusted.glusterfs.pathinfo /mnt/test-volume/testfile
getfattr: Removing leading '/' from absolute path names
# file: mnt/test-volume/testfile
trusted.glusterfs.pathinfo="(<DISTRIBUTE:test-volume-dht> <POSIX:storage1:/data/sdm/test/testfile>)"

So I can tell from the mountpoint that there are 12 bricks, which is
somewhat useful, but not what the bricks are called.  This means it is going
to be difficult to decide whether

<POSIX:storage1:/data/foo/bar>

and

<POSIX:storage1:/data/baz/qux>

are on the same brick (storage1:/data) or on two different bricks
(storage1:/data/foo and storage1:/data/baz)

Regards,

Brian.