On Mon, Feb 14, 2005 at 09:14:43AM -0800, Ben Hartshorne wrote:
> Hey folks,
> 
> So I guess I sorta solved my issue of yesterday...  Well, at least I've
> gotten further.
> 
> Here's my situation.
> 
> I used the mkbootpackage to make myself a directory (ben5) in
> /usr/share/systemimager/i386/boot/ that contains kernel, initrd.img, and
> boel-binaries.tar.gz.  I compiled my own kernel because the one that
> came with the SI suite just didn't do it for me at all.  
> 
> On a prosppective client, I install fedora and then lilo, and get it so
> it can boot.   I use the same kernel I've got on my golden server on
> this client to verify that the kernel is compiled well enough to boot
> the machine.
> 
> I run 'updateclient --autoinstall --server myserver' and it goes and
> gets all the relevant files.  I muck with lilo.conf a bit (b/c the one
> as generated doesn't work right - another story) and reboot. 
> 
> My boot gets a ways along, but when it gets to an rsync line, it
> segfaults.  The return code of rsync's segfault is 139.  When I poked
> around a bit, I found that ping also segfaults, same exit code.  
> 
> Because of these segfaults, certain sections of the boot script don't
> finish, and I'm left with a system that fauls its install.
> 
> Any thoughts?


Hey all,

Just an update on what I've tried, in case anybody's interested...

I found a friend who successfully uses SI to manage his cluster.  I
copied the initrd.img from him over to my machine, and the same symptom
was evident - ping and rsync both segfault.  

I copied my initrd.img over to him, and he could execute the binaries
without trouble.  

So, it seems theres' something about my setup (hardware?  kernel?
libraries?) that is having a hissy fit with the SI stuff.

I'm running these programs on a Dell SC1425 1U server with a Xeon chip.
I am using a nearly-vanilla install of Fedora Core 3.  (A few
applications have been added, but not much.)

My golden server is running the stock Fedora core 3 kernel.  

My test client is downloading, for this install process, a handbuilt
2.6.10 kernel based on the "config" file that comes with the SI install
(/usr/share/systemimager/boot/i386/standard/config).  I did a 'make
oldconfig' then added support for the devices I needed (SATA, Intel
GigE).  

After booting my test client sereval times and getting this segfault
each time, I started taking a bit of a shortcut to test - I unpack the
initrd.img onto my golden server, chroot into it, and run the binaries
from there.  The segfault is the same minus the 30 minute
boot/reboot/recover cycle.  

The procedure I have to run this test:

>> [EMAIL PROTECTED]:/tmp$ cp /usr/share/systemimager/boot/i386/ben5/initrd.img
>> .
>> [EMAIL PROTECTED]:/tmp$ cat initrd.img | gunzip > initrd
>> [EMAIL PROTECTED]:/tmp$ sudo mkdir /mnt/initrd
>> [EMAIL PROTECTED]:/tmp$ sudo mount /tmp/initrd /mnt/initrd -o loop
>> [EMAIL PROTECTED]:/tmp$ sudo chroot /mnt/initrd ash
>> 
>> 
>> BusyBox v0.60.3 (2004.06.29-02:52+0000) Built-in shell (ash)
>> Enter 'help' for a list of built-in commands.
>> 
>> # ping 128.32.136.9
>> PING 128.32.136.9 (128.32.136.9): 56 data bytes
>> 64 bytes from 128.32.136.9: icmp_seq=0 ttl=46 time=33.0 ms
>> Segmentation fault
>> #

Comments?

-ben


-- 
Ben Hartshorne
email: [EMAIL PROTECTED]
http://ben.hartshorne.net

Attachment: pgpsajzAXZ2H6.pgp
Description: PGP signature

Reply via email to