Ok... I've got more information. It's possible that I haven't got the right information, but maybe this will help.

At the end of the imaging process (right before it says "It's finished! Reboot already!") I get these messages.

total size is 0  speedup is 0.00
$VAR1 = [];
No module 1 found for kernel 2.4.22-21.mdk
WARNING: RH style ramdisk generation failed. at
/usr/lib/systemconfig/Initrd.pm line 96
WARNING: RH style ramdisk generation failed. at
/usr/lib/systemconfig/Initrd.pm line 96

Using a Mandrake 9.2 boot disk and going into rescue mode on the cluster node, I can reproduce this error by trying

mkinitrd 2.4.22-21mdk

which gives me the same error:
No module 1 found for kernel 2.4.22-21.mdk

Trying this same command on the head node works fine, and generates the initrd file of appropriate size. Copying this (presumably correct, from the head node) initrd-2.4.22-21mdk.img file into the /boot directory on the node did not make the node boot up correctly. Is it correct that for a given kernel, the initrd.img file should always be the same size? What is "module 1"? Why does this work on the head node and not on the cluster node?

Also curious: an fdsk /dev/sda in rescue mode on the node showed that /dev/sda5 was type 83 (linux), not 82 (linux swap), despite the following lines in my /var/lib/systemimager/scripts/oscarimage.master file

### BEGIN generate new fstab file from autoinstallscript.conf ###
cat <<'EOF' > /a/etc/fstab
/dev/sda5  swap swap defaults 0 0
/dev/sda1  / ext3 defaults 1 2
/dev/sda7       /biobits        ext3    defaults        1       2
/dev/sda6       /usr    ext3    defaults        1       2
/dev/fd0        /mnt/floppy     auto    noauto,owner    0       0
none    /proc   proc    defaults        0       0
none    /dev/pts        devpts  defaults        0       0
nfs_oscar:/home /home   nfs     rw      0       0
EOF
### END generate new fstab file from autoinstallscript.conf ###

The /etc/fstab is being created properly, but the partition table was not correct. Changing this manually (setting /dev/sda5 to swap) and booting the machine did not work either (though this is not surprising on its own.)

The head node is using the enterprise kernel
[EMAIL PROTECTED] scripts]# uname -a
Linux elf 2.4.22-21mdkenterprise #1 SMP Fri Oct 24 19:48:37 MDT 2003 i686 unknown unknown GNU/Linux


while the OSCAR default image has tried to use just the 2.4.22-21mdk kernel (no enterprise). I doubt that this is the problem, but what do you guys think? I'm grasping at straws here.

Finally, an OSCAR type question: is it necessary to re-associate the machines with MACs and the like to switch images? I already have a group of nodes associated with MACs. All I want to do is blast out a different image to them (after creating that image with the image builder, say with a different kernel). What is the best way to change the image that you want to send out to the same set of nodes? And how will you know if it's used the different image that you want? The only place I've seen in the OSCAR wizard for setting which image file to use is on the Define Clients dialog.

Jason


On Dec 12, 2003, at 1:08 AM, Jeremy Enos wrote:


At 07:46 PM 12/10/2003, Jason Hlady wrote:
Right! So: let's say I can figure out what the error is (I'll see if I can
get all of the error message at once, and probably post it here.)


Let's say I can go about trying to build a different kernel with the
appropriate support for my SCSI devices. Where does OSCAR grab the kernel
that it installs on the nodes?

I can offer you something here. OSCAR uses the kernel found in the base rpm list by default. The kernel would be the one that's rpm installed by the distro in that case. The SIS boot kernel (the one that is used for network booting and installing the nodes) is different and separate though, and can be found in /tftpboot. If you use the kernel_picker package, you have the option of selecting a different/custom kernel that you point to to replace these kernels if necessary.


The default kernel within the image can be configured with a config file... e.g.: /var/lib/systemimager/images/oscarimage/etc/systemconfig/ systemconfig.conf

Jeremy

How should I go about modifying it? I'm at
home right now, so I don't have the chunk of the manual that deals with a
custom kernel. I can probably manage to build a kernel with the appropriate
drivers linked in, if I just know what kernel source I'm supposed to be
modifying!


Thanks for the help, and by the way, your earlier message was right on the
money--I hadn't copied in the boel_binaries.tar.gz file--I got lucky enough
to figure this out along the way.


Jason


----- Original Message ----- From: "Frank Crawford" <[EMAIL PROTECTED]> To: "Jason Hlady" <[EMAIL PROTECTED]> Cc: <[EMAIL PROTECTED]> Sent: Wednesday, December 10, 2003 7:22 PM Subject: Re: [Oscar-users] Imaging successful but image doesn't work


> Jason, Bernard,
> One other issue that may be coming in is that SIS builds it's own
> initrd.img file on the nodes and may not be including all the required
> SCSI modules during booting.
>
> The problem is almost certainly a kernel issue, and as such you will
> need to watch closely during the boot for errors, etc. It probably will
> give some message about failing to find or load some SCSI module, but
> with out seeing the kernel messages, it is pretty hard to diagnose.
>
> Frank
>
> On Thu, 2003-12-11 at 09:37, Bernard Li wrote:
> > Hey Jason:
> >
> > Jason Hlady wrote:
> >
> > > I haven't tried installing Man 9.2 on the cluster nodes, only on the
> > > head node (which is almost identical to the cluster nodes.) Still,
it's
> > > probably worth doing, and I'll try that first thing tomorrow morning.
> >
> > If they are the same hardware then that probably won't make much of a
> > difference.
> >
> > > My colleague indicates that he's seen the "unable to mount root fs,
> > > kernel panic" message before when trying to install Mandrake on IBM
> > > x330/335s, and that he worked around it by building a different
kernel.
> > > Perhaps it's a kernel problem and not a configuration one. Does
OSCAR
> > > simply use the kernel which the head node uses? If so, and if the
> > > kernel distributed with Man 9.2 works on both the head and the cluster
> > > nodes, then at least it should be clear that it's not a kernel
> > > problem... I think. :)
> >
> > I am not too sure about kernel stuff but I think SIS uses its own kernel
> > when building images (not related to the kernel on the headnode). You
> > might want to look into the OSCAR package kernel-switcher to see if that
> > gives you any insight. I would also recommend going through the
> > mailing-list archives - I have seen 'similar' questions asked in the
> > past, so you might actually be able to get something out of it.
> >
> > > I'll have to do a little more work on this. Any more suggestions from
> > > anyone else would be appreciated. I'm not completely against RH9.0,
but
> > > the fact that it's no longer supported by RedHat anymore, coupled with
> > > the fact that we're a Mandrake shop around here and have more
expertise
> > > in Mandrake, makes me want to try and get it working with Mandrake.
Of
> > > course, I could always try RH using one of the cluster nodes as a head
> > > node and go from there.
> >
> > I think it'd be cool if after this exercise you emerge and actually got
> > it working under Mandrake 9.2 and we can all learn from this (and
> > improve our current support).
> >
> > But if it is any incentive, we are also using the IBM x330's and we have
> > no issues with RH 9.0 :-)
> >
> > Good luck!
> >
> > Cheers,
> >
> > Bernard
> >
> >
> >
> > -------------------------------------------------------
> > This SF.net email is sponsored by: SF.net Giveback Program.
> > Does SourceForge.net help you be more productive? Does it
> > help you create better code? SHARE THE LOVE, and help us help
> > YOU! Click Here: http://sourceforge.net/donate/
> > _______________________________________________
> > Oscar-users mailing list
> > [EMAIL PROTECTED]
> > https://lists.sourceforge.net/lists/listinfo/oscar-users
> --
> ac3
> Suite G16, Bay 7, Locomotive Workshop Phone: 02 9209 4600
> Australian Technology Park Fax: 02 9209 4611
> Eveleigh NSW 1430
>




-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/
_______________________________________________
Oscar-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/oscar-users

--------------
Jason Hlady, B. Sc., M. Sc. (Chem), Adv. Cert. (Comp. Sci.)
Programmer/Analyst (Bioinformatics Specialist)
U of Saskatchewan, Bioinformatics Research Laboratory (BIRL)
[EMAIL PROTECTED] (306) 966-2075



-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/
_______________________________________________
Oscar-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/oscar-users

Reply via email to