Re: gvinum remains broken in 5.3-RELEASE?

On Mon, 8 Nov 2004 08:41:00 +0100 (CET), Lukas Ertl <le_at_FreeBSD.org>
wrote:

> On Mon, 8 Nov 2004, Andy Farkas wrote:
> 
> > gvinum found lots more old info than I initially saw.
> >
> > Apparently I now have 1 extra physical drive, 2 extra
> > volumes, and 8 extra subdisks. How do Iget rid of
> > these old objects?
> 
> gvinum rm -r <object>

Recently, when I tried to used "gvinum rm -r" to remove a plex belonging
to a mirrored volume, gvinum complained about the subdisk and wouldn't
remove the plex.  I was able to remove the plex successfully by doing a
"gvinum rm <subdisk>" followed by "gvinum rm <plex>".

Somewhat worryingly, after the "gvinum rm <subdisk>", its plex was still
marked "up" and not "down."  The system was in a quiescent state, so
nothing bad happened, but I couldn't help wondering what would have
occurred had the system been active and the mirrored volume been written
to (and hence writes attempted on the plex sans subdisk) between the two
commands.

BTW, I'd tried to "gvinum stop <plex>" the plex in question before
removing it (a la classic vinum), but discovered that despite the gvinum
internal help stating that functionality was supported, it isn't. 
Looking at gvinum.c confirmed this: gvinum's "stop" will only stop
gvinum itself (and attempt to unload the geom_vinum module), not
specific objects.

FWIW, this is on a very recent 6.0-CURRENT.

Cheers,

Paul.
-- 
e-mail: paul_at_gromit.dlib.vt.edu

"Without music to decorate it, time is just a bunch of boring production
 deadlines or dates by which bills must be paid."
        --- Frank Vincent Zappa