From: Ola Lundqvist (opal_at_debian.org)
Date: Fri 13 Jun 2003 - 09:55:38 BST
On Thu, Jun 12, 2003 at 03:04:31PM +0200, Herbert Poetzl wrote:
> On Thu, Jun 12, 2003 at 02:03:40PM +0200, Ola Lundqvist wrote:
> > Hello
> >
> > On Thu, Jun 12, 2003 at 01:53:43PM +0200, Herbert Poetzl wrote:
> > > On Thu, Jun 12, 2003 at 08:03:58AM +0200, Ola Lundqvist wrote:
> > *SNIP*
> > > > rsync -a --delete \
> > > > --exclude=/proc --exclude=/dev/pts\
> > > > $VSERVERS_ROOT/$FROM/ $VSERVERS_ROOT/$TO
> > > > mkdir -p $VSERVERS_ROOT/$TO/proc
> > > > mkdir -p $VSERVERS_ROOT/$TO/dev/pts
> > >
> > > hmm, so I have to modify this script, if I do
> > > additional mounting like mount --bind /var/tmp /tmp
> >
> > That is very dangerout. This means that your vserver will be able to
> > remove the motherhost temp files. It also means that they will be cleaned
> > when you restart the vserver.
>
> first, this was meant from 'inside' the vserver, which
> would be neither dangerous nor very useful, but it
> could be in some script ...
Ahh then that is an other thing... :)
> > > or something similar in the vserver?
> >
> > No I do not think so. Well if you start to use some other mounts to things
> > that change in the same manner as /proc and /dev/pts, yes but not otherwise.
>
> it actually affects any mount, because if for example
> you decide to put an RPM archive on each vserver, and
> you could do this for example via --bind mounts of an
> actually ro mounted partition ... or several mounts
> of this partition, if you fear the --bind magic ;)
Is that the reason why you can not give it ro permission if you bind to
a rw partition?
> > The problem is not that they are mounted. It is that these mounted points
> > behave very different from all other filesystems. Chey are linked directly
> > into the kernel.
>
> --bind mounts are actually VFS magic, or better a simple
> way to modify the VFS namespace ... I know that ...
>
> > /proc is very special
> > /dev/pts is also a bit special.
> > /tmp,/var/tmp ... It should not need to be copied either. It is (should) be
> > cleaned on every boot anyway.
>
> I do not think that /vservers/<NAME>/var/tmp is ever cleaned
> at all *hint*
Hmm, shouldn't it be cleaned when the vserver starts? Well not /var/tmp
maybe... :)
Regards,
// Ola
>
> > > best,
> > > Herbert
> > >
> > > > In Debian I have patched vserver to allow other directories than /vservers, so
> > > > that is the reason for $VSERVERS_ROOT.
> > > >
> > > > Regards,
> > > >
> > > > // Ola
> > > >
> > > > --
> > > > --------------------- Ola Lundqvist ---------------------------
> > > > / opal_at_debian.org Annebergsslingan 37 \
> > > > | opal_at_lysator.liu.se 654 65 KARLSTAD |
> > > > | +46 (0)54-10 14 30 +46 (0)70-332 1551 |
> > > > | http://www.opal.dhs.org UIN/icq: 4912500 |
> > > > \ gpg/f.p.: 7090 A92B 18FE 7994 0C36 4FE4 18A1 B1CF 0FE5 3DD9 /
> > > > ---------------------------------------------------------------
> >
> > --
> > --------------------- Ola Lundqvist ---------------------------
> > / opal_at_debian.org Annebergsslingan 37 \
> > | opal_at_lysator.liu.se 654 65 KARLSTAD |
> > | +46 (0)54-10 14 30 +46 (0)70-332 1551 |
> > | http://www.opal.dhs.org UIN/icq: 4912500 |
> > \ gpg/f.p.: 7090 A92B 18FE 7994 0C36 4FE4 18A1 B1CF 0FE5 3DD9 /
> > ---------------------------------------------------------------
-- --------------------- Ola Lundqvist --------------------------- / opal_at_debian.org Annebergsslingan 37 \ | opal_at_lysator.liu.se 654 65 KARLSTAD | | +46 (0)54-10 14 30 +46 (0)70-332 1551 | | http://www.opal.dhs.org UIN/icq: 4912500 | \ gpg/f.p.: 7090 A92B 18FE 7994 0C36 4FE4 18A1 B1CF 0FE5 3DD9 / ---------------------------------------------------------------