--a4/Kg7jFoN9FtpDd
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Mon, Nov 20, 2006 at 12:30:47PM +1100, Daniel Carosone wrote:
> After the reboot and automatic fsck -p, in the first instance, the
> cleaner wasn't making forward progress, and wasn't recovering the
> space.  I shut the domain down again and forced a full fsck.  This
> found the unlinked file and cleaned it up - but left damage of its
> own. A second run fixed that.
>=20
> After a reboot and mount, the fs is still 107% full.. but the cleaner
> seems to be slowly trickling the space back.

No, it was worse than that.

The cleaner made a tiny amount of progress, then apparently got stuck
cleaning the same segs over and over again, each time thinking it had
succeeded.

So I shut down again, and fsck -f found the same "properly orphaned
inodes" that had been removed previously back again.

putting a "reboot" between each fsck run didn't help - but putting an
"xm destroy" (or presumably, a "reboot -n") between each run did work.

So it seems that, even with / mounted read only, reboot is causing the
filesystem to be closed in a way that is restoring the problems fsck
fixed.

NetBSD nblfs 4.99.3 NetBSD 4.99.3 (XEN3_DOMU) #37: Tue Nov 14 20:51:10 EST =
2006

--
Dan.
--a4/Kg7jFoN9FtpDd
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (NetBSD)

iD8DBQFFYShhEAVxvV4N66cRAmviAJ9m0TdBOGOGTPoOQdGYdc8Ub0kx4gCfe4Dv
gKNJ+qVirgBnR+gR/c7P9Xo=
=ZRiA
-----END PGP SIGNATURE-----

--a4/Kg7jFoN9FtpDd--