On 03/28/2013 02:22 PM, Robert Hajime Lanning wrote:
On 03/27/13 22:26, Justin Clift wrote:
Had a quick thought. Would a gluster volume command to
remove all of the xattr's from a brick be useful?
e.g. "sudo gluster volume clear /path/mybrick1"
It would remove all Gluster xattr's on the brick, letting it
be used in a new volume.
Pretty much exactly what we're giving people manual
directions for, with setfattr -x [xattrs here] /path/mybrick1.
Except people commonly have problems with the setfattr
approach, so we end up having to step them through it
manually on IRC until it works.
Worthwhile? (note, "clear" was a first idea for a
reasonable name, can be anything. :>)
I would expect this to be an option to the remove brick command.
You would never do this, without removing the brick from the volume.
If there is no volume, then I would expect it to be manual steps, as
it is today.
You might if you forgot to set the flag when you removed the brick from
the volume.
My preference would be a flag on volume create that allows the override
of the brick xattrs. GlusterFS takes the responsible approach to data
integrity and leaves everything on the brick when it's removed. I think
that's the correct move and should not be changed or optionally
overridden. There are occasionally reasons to use a previously populated
brick so allowing a way of forcing that on creation makes some sense.
There are a number of ways that using a previously used brick can also
fail, so unless someone knows why they're reusing a brick, it's better
for them not to and, instead, wipe that brick before use.
