Bug 549158

Created attachment 323697 [details]
screenshot

I cannot see a problem related to partitioning.

Created attachment 324663 [details]
yast installation logs

same problem with 11-rc2 (openSUSE-KDE4-LiveCD-Build0339-i686.iso).

x86_64 RC1 & RC2 LiveCD
I got this problem when trying to install the 11.2 RC1 with the LiveCD.
and I thought it was about the the disk problem, and I got the same error 
in the 11.2 RC2 too.
Copying the live image to hard disk failed.
I tested it on the VMware Workstation 7.0 (Trial Version) just now, the same 
error.
all the md5sum of the livecd image is correct.
should it be the problem of LiveCD's filesystem ?
I did not see the error in the reloaded livecd of 11.1 and 11.2 M2.
hope this can be solved before 11.2 releases.

According to the log, the underlying command fails, however, I don't see its error message.

if you can reproduce that bug, once it occurrs and the system is still mounted, can you run following command?

mkfifo /tmp/YaST2-03169-la8Vgm/system_clone_fifo_1 ;
         mkfifo /tmp/YaST2-03169-la8Vgm/system_clone_fifo_2 ;
         tar -C / --numeric-owner -cSf /tmp/YaST2-03169-la8Vgm/system_clone_fifo_1 --one-file-system . &
         dd bs=1048576 if=/tmp/YaST2-03169-la8Vgm/system_clone_fifo_1 of=/tmp/YaST2-03169-la8Vgm/system_clone_fifo_2 >&2 &
         jobs -l >&2;
         tar -C /mnt --numeric-owner -xSf /tmp/YaST2-03169-la8Vgm/system_clone_fifo_2

and find out which process (and why) fails?

hmhm.

parallel execution of:

linux:~ # tar -C / --numeric-owner -cSf /tmp/YaST2-03200-Ik5EuY/system_clone_fifo_1 --one-file-system .
[skip]
tar: /tmp/YaST2-03200-Ik5EuY/system_clone_fifo_1: Wrote only 4096 of 10240 bytes
tar: Error is not recoverable: exiting now

and:

linux:~ # df -h && date
Filesystem            Size  Used Avail Use% Mounted on
/dev/loop0            3.0G  2.8G  195M  94% /
udev                  504M  248K  504M   1% /dev
/dev/sda6             2.0G  206M  1.7G  11% /mnt
/dev/sda1              69M  7.2M   58M  12% /mnt/boot
/dev/sda10             14G  167M   13G   2% /mnt/home
/dev/sda9             4.0G   72M  3.7G   2% /mnt/opt
/dev/sda7             6.0G  2.2G  3.5G  39% /mnt/usr
/dev/sda8             2.0G   82M  1.8G   5% /mnt/var
Fri Nov  6 15:36:30 MSK 2009
linux:~ # df -h && date
Filesystem            Size  Used Avail Use% Mounted on
/dev/loop0            3.0G  2.9G  134M  96% /
udev                  504M  248K  504M   1% /dev
/dev/sda6             2.0G  206M  1.7G  11% /mnt
/dev/sda1              69M  7.2M   58M  12% /mnt/boot
/dev/sda10             14G  167M   13G   2% /mnt/home
/dev/sda9             4.0G   72M  3.7G   2% /mnt/opt
/dev/sda7             6.0G  2.2G  3.5G  39% /mnt/usr
/dev/sda8             2.0G   82M  1.8G   5% /mnt/var
Fri Nov  6 15:36:49 MSK 2009
linux:~ # df -h && date
Filesystem            Size  Used Avail Use% Mounted on
/dev/loop0            3.0G  3.0G   16K 100% /
udev                  504M  248K  504M   1% /dev
/dev/sda6             2.0G  206M  1.7G  11% /mnt
/dev/sda1              69M  7.2M   58M  12% /mnt/boot
/dev/sda10             14G  167M   13G   2% /mnt/home
/dev/sda9             4.0G   72M  3.7G   2% /mnt/opt
/dev/sda7             6.0G  2.2G  3.5G  39% /mnt/usr
/dev/sda8             2.0G   82M  1.8G   5% /mnt/var
Fri Nov  6 15:37:58 MSK 2009

in order that fix this bug, size of /dev/loop0 must be greather than 3 Gb, right ?

Did you run all the commands in parallel? That's strange, since fifo should be emptied countinuously by the other processes...

no :)

now, i run one more test:

linux:~ # mkfifo /tmp/YaST2-03136-f8bFS4/system_clone_fifo_1 ; mkfifo /tmp/YaST2-03136-f8bFS4/system_clone_fifo_2 ;

linux:~ # tar -C / --numeric-owner -cSf /tmp/YaST2-03136-f8bFS4/system_clone_fifo_1 --one-file-system . & dd bs=1048576 if=/tmp/YaST2-03136-f8bFS4/system_clone_fifo_1 of=/tmp/YaST2-03136-f8bFS4/system_clone_fifo_2 >&2 & jobs -l >&2; tar -C /mnt --numeric-owner -xSf /tmp/YaST2-03136-f8bFS4/system_clone_fifo_2                                                                       
[1] 8821
[2] 8822
[1]-  8821 Running                 tar -C / --numeric-owner -cSf /tmp/YaST2-03136-f8bFS4/system_clone_fifo_1 --one-file-system . &
[2]+  8822 Running                 dd bs=1048576 if=/tmp/YaST2-03136-f8bFS4/system_clone_fifo_1 of=/tmp/YaST2-03136-f8bFS4/system_clone_fifo_2 1>&2 &                                         

[skip]

tar: ./usr/share/zoneinfo/posix/W-SU: Cannot hard link to `./etc/localtime': Invalid cross-device link
tar: ./usr/share/zoneinfo/posix/Europe/Moscow: Cannot hard link to `./etc/localtime': Invalid cross-device link
tar: ./usr/share/zoneinfo/W-SU: Cannot hard link to `./etc/localtime': Invalid cross-device link
tar: ./usr/share/zoneinfo/Europe/Moscow: Cannot hard link to `./etc/localtime': Invalid cross-device link
2+71272 records in
2+71272 records out
2215915520 bytes (2.2 GB) copied, 442.381 s, 5.0 MB/s
tar: ./home/linux/.pulse: time stamp 2009-11-06 18:55:33 is 6472.239163428 s in the future
tar: ./home/linux/.kde4: time stamp 2009-11-06 18:55:28 is 6467.23890968 s in the future
tar: ./etc/sysconfig: time stamp 2009-11-06 18:54:20 is 6399.238626445 s in the future
tar: ./var/run: time stamp 2009-11-06 19:42:00 is 9259.238583706 s in the future
tar: .: time stamp 2009-11-06 18:53:57 is 6376.238490741 s in the future
tar: Exiting with failure status due to previous errors
[1]-  Done                    tar -C / --numeric-owner -cSf /tmp/YaST2-03136-f8bFS4/system_clone_fifo_1 --one-file-system .
[2]+  Done                    dd bs=1048576 if=/tmp/YaST2-03136-f8bFS4/system_clone_fifo_1 of=/tmp/YaST2-03136-f8bFS4/system_clone_fifo_2 1>&2


looks like a bug in timezone-2009m-2.2.i586 package:

linux@linux:~> rpm -q --scripts $(rpm -qf /etc/localtime)
postinstall scriptlet (using /bin/sh):
if [ -f /etc/sysconfig/clock ];
then
    . /etc/sysconfig/clock
    if [ -n "$TIMEZONE" -a -f /etc/localtime -a -f /usr/share/zoneinfo/$TIMEZONE ]; then
        new=$(mktemp /etc/localtime.XXXXXXXX) || exit 1
        cp -l /usr/share/zoneinfo/$TIMEZONE $new 2>/dev/null || cp -fp /usr/share/zoneinfo/$TIMEZONE $new
        mv -f $new /etc/localtime
    else
        [ ! -f /etc/localtime ] || echo "WARNING: Not updating /etc/localtime with new zone file" >&2
    fi
fi


"cp -l" make hardlinks instead of copies.

probably same bug in the "smolt" or "libzypp" pakage:

tar: ./etc/smolt/uuid: Cannot hard link to `./var/lib/zypp/AnonymousUniqueId'

hardlinks in live system is very, very bad idea. installation will always fails, because it is impossible to create cross-device hardlinks.

Ah, now I see the problem, many thanks for this analysis.

I don't know whether there is any option for tar to solve that when un-tarring (I at least didn't find any), what can be done is to use --hard-dereference option for the tar command creating the archive (according to the documentation). This should ensure that hardlinks will be handled properly.

In general, hard links in live system make sense - they save the disk space, which is in case of CD very limited, however, causes a problem here.

crutches for 11.2 release installation:

1. boot into live system (kde or gnome);
2. run terminal emulator;
3. type "su -";
4. cd /usr/share/YaST2/modules;
5. open ImageInstallation.ycp in editor (vim ImageInstallation.ycp);
6. go to line 910;
7. add "--hard-derefernce" into tar options;
8. ycpc -c ImageInstallation.ycp;
9. run installation;
10. ...
11. PROFIT

patch:

--- ImageInstallation.ycp.orig  2009-11-13 08:21:03.150679080 +0000
+++ ImageInstallation.ycp       2009-11-13 08:21:34.172020411 +0000
@@ -907,7 +907,7 @@ global boolean FileSystemCopy (string fr
     cmd = sformat (
        "mkfifo %3 ;
         mkfifo %4 ;
-        tar -C %1 --numeric-owner -cSf %3 --one-file-system . &
+        tar -C %1 --hard-dereference --numeric-owner -cSf %3 --one-file-system . &
         dd bs=1048576 if=%3 of=%4 >&2 &
         jobs -l >&2;
         tar -C %2 --numeric-owner -xSf %4",

Does that mean that with this change live installation works for you?

Yes

Thanks for confirmation.
Coolo, do you plan to release updated live media for 11.2?

not now, but e.g. the kde team plans to release a spin off with updates.

OK,  then I will fix it in SVN also in the 11.2 branch (apply patch from comment#12) and in case new release is planned, it can be grabbed.

Patch applied in SVN for 11.3, 11.2 and SDK11-SP1.

This one burnt me and I couldn't find anything preventing it from burning other users in the near future, so I added it to 

http://en.opensuse.org/Bugs:Most_Annoying_Bugs_11.2#Installation

and also added a link to that page from 

http://en.opensuse.org/OpenSUSE_11.2

I think we could be a bit better about communicating these kinds of issues.

(In reply to comment #16)
> not now, but e.g. the kde team plans to release a spin off with updates.

Does this mean the GNOME live image won't be fixed?  I saw the problem with both GNOME and KDE images.

i affected currently by this bug in my installation of OpenSuse 13.2 edulife version.