10:49 p.m.
People,
If the livecd script is interrupted or crashes it leaves the img file
system in /var/tmp and I can't unmount it except by rebooting the
computer which is a real pain of course - is there some trick to
umounting this?
Thanks,
Phil.
--
Philip Rhoades
GPO Box 3411
Sydney NSW 2001
Australia
E-mail: phil(a)pricom.com.au
8:25 a.m.
On Sun, Sep 26, 2010 at 13:49:37 +1000,
Philip Rhoades <phil(a)pricom.com.au> wrote:
People,
If the livecd script is interrupted or crashes it leaves the img file
system in /var/tmp and I can't unmount it except by rebooting the
computer which is a real pain of course - is there some trick to
umounting this?
You have to umount things in the right order.
Which version of livecd-tools are you using? Later versions have lazy umounts
that work around this. They don't help if you control c out when it's in
some parts of the code, but it shouldn't happen on normal compose failures.
8:48 a.m.
Bruno,
On 2010-09-26 23:25, Bruno Wolff III wrote:
On Sun, Sep 26, 2010 at 13:49:37 +1000,
Philip Rhoades<phil(a)pricom.com.au> wrote:
> People,
>
> If the livecd script is interrupted or crashes it leaves the img file
> system in /var/tmp and I can't unmount it except by rebooting the
> computer which is a real pain of course - is there some trick to
> umounting this?
You have to umount things in the right order.
There seems to be only one file system mounted?
Which version of livecd-tools are you using? Later versions have lazy
umounts
that work around this.
livecd-tools-031-1.fc12.1.x86_64 (it seems to be the latest).
They don't help if you control c out when it's in
some parts of the code, but it shouldn't happen on normal compose
failures.
I have Ctrl-C'd out of the script on occasions and of course this causes
the problems . .
Thanks,
Phil.
--
Philip Rhoades
GPO Box 3411
Sydney NSW 2001
Australia
E-mail: phil(a)pricom.com.au
12:01 p.m.
On Sun, Sep 26, 2010 at 23:48:51 +1000,
Philip Rhoades <phil(a)pricom.com.au> wrote:
There seems to be only one file system mounted?
Usually it's the bind mounts of /dev, /proc and /sys that don't get unmounted
first.
livecd-tools-031-1.fc12.1.x86_64 (it seems to be the latest).
Try livecd-tools-033-3.fc12 in testing. If it works for you please add
karma at:
https://admin.fedoraproject.org/updates/livecd-tools-033-3.fc12
I have Ctrl-C'd out of the script on occasions and of course this
causes
the problems . .
It shouldn't, but it's trickier to catch. A called function from another
package is trapping it and exiting rather than using a more normal error
return. I think there may be a way to fix this, but it needs some looking into.
That only affects killing it when it is doing some things, at other times
things should get cleaned up in 033.
12:12 p.m.
On Sun, 2010-09-26 at 18:01 +0100, Bruno Wolff III wrote:
On Sun, Sep 26, 2010 at 23:48:51 +1000,
Philip Rhoades <phil(a)pricom.com.au> wrote:
> There seems to be only one file system mounted?
Usually it's the bind mounts of /dev, /proc and /sys that don't get unmounted
first.
and /var/cache/yum.
And sometimes you have to delete the loop device manually:
losetup -d /dev/loop0
(or whatever loop device was used).
It's boring enough that I've often intended to write a quick script to
do it (but, of course, I've never got round to it).
James
12:17 p.m.
On Sun, Sep 26, 2010 at 18:12:20 +0100,
James Heather <j.heather(a)surrey.ac.uk> wrote:
And sometimes you have to delete the loop device manually:
losetup -d /dev/loop0
(or whatever loop device was used).
It's boring enough that I've often intended to write a quick script to
do it (but, of course, I've never got round to it).
That should eventually be able to go away. I am not sure when it changed,
but you don't need to use losetup with at least recent stuff. I haven't
looked into the details, but we should stop using it when it safe to
do so for stuff that we care about.
12:19 p.m.
On Sun, Sep 26, 2010 at 12:01:06 -0500,
Bruno Wolff III <bruno(a)wolff.to> wrote:
Try livecd-tools-033-3.fc12 in testing. If it works for you please add
karma at:
https://admin.fedoraproject.org/updates/livecd-tools-033-3.fc12
That was a bit of a mistatement. If you try the test version, please fill
out the karma whether it works or not. But if it breaks something that
works with 031, it's even more important to supply negative karma, than
supplying positive karma if it looks OK.
7:16 a.m.
James,
On 2010-09-27 03:12, James Heather wrote:
On Sun, 2010-09-26 at 18:01 +0100, Bruno Wolff III wrote:
> On Sun, Sep 26, 2010 at 23:48:51 +1000,
> Philip Rhoades<phil(a)pricom.com.au <mailto:phil@pricom.com.au>>
wrote:
> > There seems to be only one file system mounted?
>
> Usually it's the bind mounts of /dev, /proc and /sys that don't get
unmounted
> first.
and /var/cache/yum.
Did those but still couldn't umount:
/var/tmp/imgcreate-2A4KXU/install_root
And sometimes you have to delete the loop device manually:
losetup -d /dev/loop0
I get:
loop: can't delete device /dev/loop0: Device or resource busy
(or whatever loop device was used).
It's boring enough that I've often intended to write a quick script to
do it (but, of course, I've never got round to it).
Yes, it is a bit annoying . .
Thanks,
Phil.
--
Philip Rhoades
GPO Box 3411
Sydney NSW 2001
Australia
E-mail: phil(a)pricom.com.au
7:26 a.m.
Bruno,
On 2010-09-27 03:19, Bruno Wolff III wrote:
On Sun, Sep 26, 2010 at 12:01:06 -0500,
Bruno Wolff III<bruno(a)wolff.to> wrote:
>
> Try livecd-tools-033-3.fc12 in testing. If it works for you please add
> karma at:
> https://admin.fedoraproject.org/updates/livecd-tools-033-3.fc12
That was a bit of a mistatement. If you try the test version, please fill
out the karma whether it works or not. But if it breaks something that
works with 031, it's even more important to supply negative karma, than
supplying positive karma if it looks OK.
Still builds an iso OK but breaking out of a build still causes the
umount problem - see other note . .
Regards,
Phil.
--
Philip Rhoades
GPO Box 3411
Sydney NSW 2001
Australia
E-mail: phil(a)pricom.com.au
7:43 a.m.
On Mon, 2010-09-27 at 13:16 +0100, Philip Rhoades wrote:
James,
On 2010-09-27 03:12, James Heather wrote:
> On Sun, 2010-09-26 at 18:01 +0100, Bruno Wolff III wrote:
>> On Sun, Sep 26, 2010 at 23:48:51 +1000,
>> Philip Rhoades<phil(a)pricom.com.au <mailto:phil@pricom.com.au>>
wrote:
>> > There seems to be only one file system mounted?
>>
>> Usually it's the bind mounts of /dev, /proc and /sys that don't get
unmounted
>> first.
>
> and /var/cache/yum.
Did those but still couldn't umount:
/var/tmp/imgcreate-2A4KXU/install_root
Then something's still running that's using that directory or file. Have
you made sure that the shell you're using to umount doesn't have that
directory as its working directory? (And if you used 'su' to get to
root, is the user shell you came from currently in that directory?)
If that doesn't help, you can run
lsof /var/tmp/imgcreate-2A4KXU/install_root
to find out what process has that directory or something beneath it
open.
> And sometimes you have to delete the loop device manually:
>
> losetup -d /dev/loop0
I get:
loop: can't delete device /dev/loop0: Device or resource busy
Yes, you won't be able to delete the loop device until after you've
successfully umounted, because the install root is mounted via the
loopback device. When you've managed to umount everything, the losetup
-d should succeed.
James
7:59 a.m.
> I get:
>
> loop: can't delete device /dev/loop0: Device or resource busy
Yes, you won't be able to delete the loop device until after you've
successfully umounted, because the install root is mounted via the
loopback device. When you've managed to umount everything, the losetup
-d should succeed.
I usually use
for x in $(cat /proc/mounts |grep ' /var/tmp/.'|cut -d' ' -f2|tac); do
umount $x; done
for l in $(losetup -a|cut -d: -f1); do losetup -d $l; done
It will fail for mounts and loppbacks still in use, but running it twice
usually resolves all the dependency chains I have.
(There was a similar issue also involving crypt/dm chains when shutting
down the live os. See https://bugzilla.redhat.com/show_bug.cgi?id=519155
- I'm not sure if that still is an issue.)
/Mads
8:07 a.m.
Mads,
On 2010-09-27 22:59, Mads Kiilerich wrote:
>> I get:
>>
>> loop: can't delete device /dev/loop0: Device or resource busy
>
> Yes, you won't be able to delete the loop device until after you've
> successfully umounted, because the install root is mounted via the
> loopback device. When you've managed to umount everything, the losetup
> -d should succeed.
I usually use
for x in $(cat /proc/mounts |grep ' /var/tmp/.'|cut -d' ' -f2|tac); do
umount $x; done
for l in $(losetup -a|cut -d: -f1); do losetup -d $l; done
It will fail for mounts and loppbacks still in use, but running it twice
usually resolves all the dependency chains I have.
Perfect! - many thanks!
(There was a similar issue also involving crypt/dm chains when
shutting
down the live os. See https://bugzilla.redhat.com/show_bug.cgi?id=519155
- I'm not sure if that still is an issue.)
Will keep it mind . . just in case . .
Regards,
Phil.
--
Philip Rhoades
GPO Box 3411
Sydney NSW 2001
Australia
E-mail: phil(a)pricom.com.au
4952
days inactive
4953
days old
livecd@lists.fedoraproject.org
11 comments
4 participants
participants (4)
-
Bruno Wolff III -
James Heather -
Mads Kiilerich -
Philip Rhoades