Re: fai 2.8.3 skipping packages?

2005-05-26 Thread Steffen Grunewald
On Wed, May 25, 2005 at 09:04:15PM +0200, Michael Tautschnig wrote:
> [...]
> > 
> > BTW:
> > I expect 2.8.4 to destroy my local setup again since the automated patching
> > of fai.conf etc would no longer work :-( and it's little fun to find
> > such changes the hard way.
> > 
> If you are willing to track all changes, you could ask Thomas to add your 
> email
> address to his CVS-commit list -- that might help.

Another idea: what about a fai-devel list? (not that I'm a developer
myself - I didn't even make it to the workshop at Geilenkirchen, but
I'd be willing to follow development a bit more closely, while still
too lazy to diff all the stuff myself)
Thomas?

> > Thanks for helping - if I had stayed with 2.6.x this would not have
> > happened (for the old error line would have shown up in error.log,
> > who reads the full log files?)
> Although 2.6.x would not have led to such (maybe?) strange behaviour, it would
> instead not have installed any packages, as the whole load of packages was 
> then
> installed at once and that command would have failed the same way.

Sure, but I would have found it at first glance.

> http://www.godslayer.org/pics/computers_46.php 

In the era of cloning (and man trying to to ${DEITY}'s job), this picture
merely shows Linux's flexibility :-)

Cheers,
 Steffen



Re: fai 2.8.3 skipping packages?

2005-05-25 Thread Michael Tautschnig
[...]
> 
> BTW:
> I expect 2.8.4 to destroy my local setup again since the automated patching
> of fai.conf etc would no longer work :-( and it's little fun to find
> such changes the hard way.
> 
If you are willing to track all changes, you could ask Thomas to add your email
address to his CVS-commit list -- that might help.

> -+-
> 
> Solution:
> I removed xv from the package list (the proper way would be to build it
> locally... until then remove it from the local repository), and everything
> looks nice now.
> 
> Thanks for helping - if I had stayed with 2.6.x this would not have
> happened (for the old error line would have shown up in error.log,
> who reads the full log files?)
> 

Although 2.6.x would not have led to such (maybe?) strange behaviour, it would
instead not have installed any packages, as the whole load of packages was then
installed at once and that command would have failed the same way.

As we are all trying to improve FAI, changes also imply incompatibilities,
unlike M$-systems:

http://www.godslayer.org/pics/computers_46.php 

:-)
Michael


Re: fai 2.8.3 skipping packages?

2005-05-25 Thread Holger Levsen
Hi,

On Wednesday 25 May 2005 17:37, Thomas Lange wrote:
> But with fai 2.8.4 beeing in  sdarge we will have a stable fai release for
> at least 3 years ;-) 

Please dont spread this, even if you might have been half joking. 
Self-fullfilling prophecy anyone ?

Lots of discussions, precautions and preparations have allready happenend, the 
current aim seems to be to release Debian/etch in 12-18 month, this means 
2006!


¡truly merry xmas!
Holger

P.S.: Re-read your mail, maybe you meant "fai 2.8.4 will still be available 
and solid when etch becomes stable, as then sarge will become oldstable, 
which will still be supported for a year or so" ? ;-)


pgpL6YF0io4oO.pgp
Description: PGP signature


Re: fai 2.8.3 skipping packages?

2005-05-25 Thread Thomas Lange
> On Wed, 25 May 2005 17:22:58 +0200, Steffen Grunewald <[EMAIL PROTECTED]> 
> said:


> BTW, it's very confusing that the fai/download/ area is kept so clean -
> it's almost impossible to find the previous version, let alone diffs.
> (There is one from 2.8 to 2.8.2...)
You can download old versions from 

http://snapshot.debian.net/
or from 
http://sourceforge.net/projects/fai

>> Now we often have the situation, when installing a
>> new FAI version, that it's very useful (or even needed?) to merge
>> changes in the FAI example configs into your own, customized FAI config

>> space in /usr/local/share/fai, which can, in case you really edit the
>> example config's scripts and things like that, be a nice little hell.
You can always do
 diff -r /usr/share/doc/fai/examples/simple /usr/local/share/fai


Installing a new FAI version does not overwrite your config space at
all. So you can use your old config space with newer fai version. This
does not always apply to file in /etc/fai but I try to keep changes in
those files to a minimum. But unfortunately we had to change things
there in the last few versions of fai. I'm sorry for that.


> - and it's quite a mess to patch the patch files with almost every
> new release (install package, overwrite files, read them, check with
> the patches, run patch --dry-run, patch the patch, dry-run again...)
Use dpkg-divert when overwriting a file in fai to keep you versions
when upgrading.

You're right that we've change some major things in /etc/fai files
several time in the last versions of fai. But with fai 2.8.4 beeing in
sdarge we will have a stable fai release for at least 3 years ;-)

-- 
regards Thomas


Re: fai 2.8.3 skipping packages?

2005-05-25 Thread Steffen Grunewald
On Wed, May 25, 2005 at 04:40:17PM +0200, Henning Sprang wrote:
> On Wed, 2005-05-25 at 15:20 +0200, Steffen Grunewald wrote:
> > answering to stupid me ...
> 
> Don't be so hard with yourself!

I am. I have to be. Made a lot of stupid mistakes all week. Coffee does
not help much. Need more sleep :-(

> > I suggest FAI to be distributed not only as complete packages but also as
> > a patch set ... would make it easier to find such small changes for
> > local personalization.
> 
> It would be a good idea in general, to better separate between things
> "included" in FAI not meant to be changed by users and things meant to
> be changed by users.

Thanks for seconding me :-)
BTW, it's very confusing that the fai/download/ area is kept so clean -
it's almost impossible to find the previous version, let alone diffs.
(There is one from 2.8 to 2.8.2...)

> Now we often have the situation, when installing a
> new FAI version, that it's very useful (or even needed?) to merge
> changes in the FAI example configs into your own, customized FAI config
> space in /usr/local/share/fai, which can, in case you really edit the
> example config's scripts and things like that, be a nice little hell.

That's what I'm going through. I have my local modifications to the
/etc/fai/* files, some patches to the disk partitioner (to keep the
contents of xfs filesystems if there are any while still allowing to
format a fresh replacement disk - the -f option to mkfs.xfs makes
things a bit hard for me) (for Alpha, I'm a bit out of sync now, but
still working on it - anyone else still interested in AlFAI?)
- and it's quite a mess to patch the patch files with almost every
new release (install package, overwrite files, read them, check with
the patches, run patch --dry-run, patch the patch, dry-run again...)

Of course (?) I could dpkg -x every new release into a tree, then run
a recursive diff - and still might miss some pre/post-install scripts
affecting my setup :-(


Hmmm. If I only could clone myself.

Cheers,
 Steffen

-- 
Steffen Grunewald * MPI fuer Gravitationsphysik (Albert-Einstein-Institut)
SciencePark Golm, Am Mühlenberg 1, D-14476 Potsdam * http://www.aei.mpg.de
* e-mail: steffen.grunewald(*)aei.mpg.de * +49-331-567-{fon:7233,fax:7298}
No Word/PPT mails - http://www.gnu.org/philosophy/no-word-attachments.html



Re: fai 2.8.3 skipping packages?

2005-05-25 Thread Henning Sprang
Hi, 

On Wed, 2005-05-25 at 15:20 +0200, Steffen Grunewald wrote:
> On Wed, May 25, 2005 at 11:49:20AM +0200, Michael Tautschnig wrote:
> 
> answering to stupid me ...

Don't be so hard with yourself!

> [...]
> 
> I checked the error.log - unfortunately the line "E: Sorry, broken packages"
> sometime in the past had been changed to "E: Broken packages", and my
> savelog.LAST (which had already been extended a lot) did not reflect this.
> I suggest FAI to be distributed not only as complete packages but also as
> a patch set ... would make it easier to find such small changes for
> local personalization.

It would be a good idea in general, to better separate between things
"included" in FAI not meant to be changed by users and things meant to
be changed by users. Now we often have the situation, when installing a
new FAI version, that it's very useful (or even needed?) to merge
changes in the FAI example configs into your own, customized FAI config
space in /usr/local/share/fai, which can, in case you really edit the
example config's scripts and things like that, be a nice little hell.


> 
> > ERROR: 25600 25600
> 
> Well, one would not expect a major problem to hide behind this line.
> Can someone please explain where this comes from, and what the numbers 
> mean? (I know that it's 256*100 or 0xa0**2 ... but what would 0x6400
> tell me? return code 100... what's that?)

I don't know about the numbers - but if any of the packages in one apt
call in software install has a failure, none of all the packages get
installed, but often an error is only displayed for the package with the
error - i had troubles to find that out, too once...

Henning




Re: fai 2.8.3 skipping packages?

2005-05-25 Thread Steffen Grunewald
On Wed, May 25, 2005 at 11:49:20AM +0200, Michael Tautschnig wrote:

answering to stupid me ...

> Are there any lines starting with "WARNING:" ? Were all the other packages
> listed in the line containing tivsm properly installed?

None of them - of the 99 only 58 were dropped, but the remaining ones already
had been installed by taskinst. If the numbers had been identical, I would
not have written the report...

> As this is a "local" package, you won't find a "Get:" line for it, but you
> should find an "Unpacking ..." line. Still, it is possible that a bunch of
> packages has not been installed, because one of them was missing from the
> archive. Then you would find lines like 

It was there but had broken depends :-(

> E: Broken packages

I checked the error.log - unfortunately the line "E: Sorry, broken packages"
sometime in the past had been changed to "E: Broken packages", and my
savelog.LAST (which had already been extended a lot) did not reflect this.
I suggest FAI to be distributed not only as complete packages but also as
a patch set ... would make it easier to find such small changes for
local personalization.

> ERROR: 25600 25600

Well, one would not expect a major problem to hide behind this line.
Can someone please explain where this comes from, and what the numbers 
mean? (I know that it's 256*100 or 0xa0**2 ... but what would 0x6400
tell me? return code 100... what's that?)

-+-

BTW:
I expect 2.8.4 to destroy my local setup again since the automated patching
of fai.conf etc would no longer work :-( and it's little fun to find
such changes the hard way.

-+-

Solution:
I removed xv from the package list (the proper way would be to build it
locally... until then remove it from the local repository), and everything
looks nice now.

Thanks for helping - if I had stayed with 2.6.x this would not have
happened (for the old error line would have shown up in error.log,
who reads the full log files?)


Back to real work now,
 Steffen



Re: fai 2.8.3 skipping packages?

2005-05-25 Thread Steffen Grunewald
On Wed, May 25, 2005 at 12:03:33PM -0400, Eugen Paiuc wrote:
> Hi Steffen,
> 
> Puting the directory
> 
> /fai/files/packages 
> 
> in a local access http/ftp would/may resolve all.

No. Since I was able to install packages from there in other stages,
this is unlikely to be the proper solution.

Cheers,
 Steffen

-- 
Steffen Grunewald * MPI fuer Gravitationsphysik (Albert-Einstein-Institut)
SciencePark Golm, Am Mühlenberg 1, D-14476 Potsdam * http://www.aei.mpg.de
* e-mail: steffen.grunewald(*)aei.mpg.de * +49-331-567-{fon:7233,fax:7298}
No Word/PPT mails - http://www.gnu.org/philosophy/no-word-attachments.html



Re: fai 2.8.3 skipping packages?

2005-05-25 Thread Eugen Paiuc

Hi Steffen,

Puting the directory

/fai/files/packages 


in a local access http/ftp would/may resolve all.

(with the adding   of coresponding changies in
/etc/apt/sources.list and /etc/fai/sources.list,
and re_run apt-get update ).

This is my idea...


regards,
   Eugen

>Steffen Grunewald wrote:


Hi,

before filing a bug against fai (2.8.3) I'd like to ask here whether I'm
making stupid mistakes again...

...
Cheers,
Steffen

 





Re: fai 2.8.3 skipping packages?

2005-05-25 Thread Michael Tautschnig
> Hi,
> 
> before filing a bug against fai (2.8.3) I'd like to ask here whether I'm
> making stupid mistakes again...
> 
> (Before I start: this did work before, with 2.6.x.)
> 
> I'm trying to install a lot of packages (this is a server that gets 
> backed up by Tivoli and has some 3ware controller in it, so among the
> packages to install there are tivsm-api and tivsm-ba, and 3dm2 - which
> are located in files/packages and properly indexed into Packages.gz).
>
[...]

Are there any lines starting with "WARNING:" ? Were all the other packages
listed in the line containing tivsm properly installed?

As this is a "local" package, you won't find a "Get:" line for it, but you
should find an "Unpacking ..." line. Still, it is possible that a bunch of
packages has not been installed, because one of them was missing from the
archive. Then you would find lines like 

E: Broken packages
ERROR: 25600 25600

Please check those for such lines.

Thanks,
Michael


fai 2.8.3 skipping packages?

2005-05-25 Thread Steffen Grunewald
Hi,

before filing a bug against fai (2.8.3) I'd like to ask here whether I'm
making stupid mistakes again...

(Before I start: this did work before, with 2.6.x.)

I'm trying to install a lot of packages (this is a server that gets 
backed up by Tivoli and has some 3ware controller in it, so among the
packages to install there are tivsm-api and tivsm-ba, and 3dm2 - which
are located in files/packages and properly indexed into Packages.gz).

In the software.log I can find they are requested for install:

[EMAIL PROTECTED] grep tivsm software.log
install_packages: executing chroot /tmp/target apt-get -y -o 
Dpkg::Options::="--force-confdef" -o Dpkg::Options::="--force-confold" 
--force-yes --fix-missing install iptables logcheck tivsm-api
tivsm-ba sysstat ethereal kernel-package xemacs21 nedit gnuplot-x11 gmt 
perlmagick xv xv-doc gs-gpl enscript cvsutils cvsreport galeon mozilla-firefox 
adobereader-enu xbase-clients xterm xlockmore-gl
xlibs-dev xutils xfs lesstif-dev xserver-xfree86 xfonts-base xfonts-75dpi 
xfonts-100dpi xfonts-scalable t1-xfree86-nonfree ttf-freefont type1inst 
xscreensaver gnome-applets gnome2-user-guide gdm
gnome-utils gnome-control-center sawfish-gnome menu gnome-session 
gnome-terminal gnumeric gnome-gv gnome-media gnome-apt gnome-games eject 
libstdc++2.10-glibc2.2 symlinks dlocate aptitude apt-utils
apt apt-howto-en grep-dctrl alien dpkg-ftp debootstrap debfoster deborphan 
dhelp gs gsfonts-x11 a2ps gv psutils poster pstotext tree mtools dosfstools 
dump w3m bc dc bsdmainutils cpio ed jove file
gawk libgmp3 libnet-perl libmail-sendmail-perl libpam0g patch perl-tk perl-suid 
perl-doc rcs sharutils tcllib time ash
[EMAIL PROTECTED] grep 3dm software.log
install_packages: executing chroot /tmp/target apt-get -y -o 
Dpkg::Options::="--force-confdef" -o Dpkg::Options::="--force-confold" 
--force-yes --fix-missing install minicom slang1-dev svgalibg1 git
ibritish iamerican ispell jargon lha bzip2 unarj unzip zoo zip unrar gimp groff 
man-db manpages manpages-dev info psgml debiandoc-sgml debiandoc-sgml-doc 
doc-linux-text doc-base doc-debian
doc-linux-de glibc-doc mgetty-docs diff-doc libapt-pkg-doc libpam-doc lilo-doc 
grub-doc lprng-doc sharutils-doc make-doc cfengine-doc texinfo emacs21 
emacs-lisp-intro emacs-goodies-el bbdb vm mp
metamail mime-support fvwm imagemagick gnuplot ncftp netselect ntp-doc finger 
whois mozilla mozilla-chatzilla mozilla-psm dialog discover textutils recode 
screen antiword tetex-extra transfig
cweb-latex auctex foiltex prosper latex-beamer electric-fence bin86 m4 byacc 
cvs gcc gcc-3.4 gcc-3.3 g++ g++-3.4 g++-3.3 g77 g77-3.4 g77-3.3 ddd indent 
autoconf automake1.8 automake1.9 3dm2 tw-cli
tw-doc apache libapache-request-perl fai fai-kernels mknbi
  3dm2 antiword apache apache-common apache2-utils arj auctex autoconf
Selecting previously deselected package 3dm2.
Unpacking 3dm2 (from .../3dm2_2.03.00.025_i386.deb) ...
Setting up 3dm2 (2.03.00.025) ...

but looking at the system itself, no tivsm-* is there.
After removing the hash in front of the /fai line in /etc/fstab, and mounting
the FS, I can do

deimos:~# apt-get install tivsm-api tivsm-ba
Reading Package Lists... Done
Building Dependency Tree... Done
The following NEW packages will be installed:
  tivsm-api tivsm-ba
0 upgraded, 2 newly installed, 0 to remove and 0 not upgraded.
Need to get 0B/17.2MB of archives.
After unpacking 43.7MB of additional disk space will be used.
Selecting previously deselected package tivsm-api.
(Reading database ... 107901 files and directories currently installed.)
Unpacking tivsm-api (from .../tivsm-api_5.2.3-0_i386.deb) ...
Selecting previously deselected package tivsm-ba.
Unpacking tivsm-ba (from .../tivsm-ba_5.2.3-0_i386.deb) ...
Setting up tivsm-api (5.2.3-0) ...
Setting up tivsm-ba (5.2.3-0) ...

so the packages are there, and can be installed.

Note that the packages have been created with alien - but they are not
the only ones, others (like condor) are already there.


Looking at the "apt-get" lines in software.log more closely, I find that
(obviously after taskinst has been run)
- clean
- install ... condor
- clean
- install tivsm-api tivsm-ba ... xv xv-doc ...
- clean
- install ... 3dm2 tw-cli ... fai ...
- clean
- install ...
- clean
- clean
Only the /fai/files/packages entries in the second install round
are missing (xv-doc too, which is a proper .deb).

I parsed *every* package listed in a apt-get install statement with
dpkg -s, and some more showed up with "unknown ok not-installed", 
totalling 58 pkgs, and including important stuff as logcheck, sysstat,
symlinks, gawk, time, ash ... which can, of course, be installed by 
hand afterwards...
If I do the parsing for each individual apt-get install line, I get the
counts: 0 58 0 0
so everything focuses to the 2nd install which consists of 99 packages.


I'm very confused.

Any ideas?

Should I copy all this stuff into a bug report?

Cheers,
 Steffen

-- 
Steffen Grunewald * MPI fuer Gravitationsphysik (Albert-Ei