RE: [Sisuite-users] Grub issues with IBM HS20 Bladecenter

Mon, 09 Jan 2006 17:03:10 -0800

Bernard,

       Thanks for sending me that new version of systemconfigurator.  Unfortunately, it didn't seem to have any effect on the process.  I decided to go back to the begining and abandon my hacked kernel, initrd.img, and boel_binaries.  I'm now using the stock ones provided by Ole in his guide.  However, when I try to push the image to a new blade in the chassis it fails in the following manner:  The machine I'm pushing to receives its initial dhcp offer.  We were snooping the link the tftp works across and we saw the pxelinux.bin, initrd.img, and kernel come across.  However, once it starts trying to receive dhcp requests (after it shows messages for having loaded the kernel) it times out, giving a message about no dhcp offers.  The dhcp offer was sent by the server, we saw it in ethereal, but it never was acked by the client.

    At first we thought perhaps it was some issue with a module not being present, but, as I note below, someone here got an exact duplicate piece of hardware (except the BIOS version differing) working with the exact same software (kernel, initrd.img, etc.) from Ole's guide.  Our only thought is that the new version of BIOS is somehow setting something wonky with the ethernet after it has relinquished control of it after the initial successful DHCP offer (something with the ifup or down somehow?)  However, we can think of no way to confirm our hypothesis .  We've got some letters out to IBM to provide some information, but they've been of no help.  Can you think of something that the bios could set to make the ethernet on the client unresponsive to a dhcp offer or some module or piece of code we could check?

 Again, I appreciate any help or ideas that you or anyone else might have to offer to help reach a solution, or at least take a step d own the right road.

Steve


Side notes:
Another guy around here has gotten systemimager 3.5.2 to work with rhel4 following only the directions in Ole's guide.  His IBM Blades HS20s have BIOS version 1.03 and the blade chassis uses the broadcom trigons for ethernet.
Also, if recompiling the kernel is going to be necessary, I've compiled some kernels before, but how do you configure a kernel to be used for systemimager? It seems things need to be enabled differently (e.g. modules included)? Can you use a kernel you have already working for your rhel as a base somehow? And how do you know how exactly to build your initrd.img with mkinitrd, the man page wasn't exactly explicit.  Sorry about all the questions, but any help would be greatly, greatly appreciated.




Bernard Li <[EMAIL PROTECTED]> wrote:
Hi Steve:
 
Try this newer version of SystemConfigurator from OSCAR trunk and see if it works for you.  Supposedly it fixes some RHEL4 grub-related issues:
 

From: [EMAIL PROTECTED] on behalf of Kan Tok
Sent: Sat 07/01/2006 19:03
To: [email protected]
Subject: RE: [Sisuite-users] Grub issues with IBM HS20 Bladecenter

Bernard,

    Good evening.  I'm using the packages provided for in the guide, so it's the systemconfigurator-2.2.2-1.noarch.rpm package.  I had read of some interesting problems with grub when you have your machine raided, but I don't have a raid array se tup for these machines. .  Please let me know if there's any other relevant information I can provide. Thanks again.

Steve



How To being used:
http://www.opensubscriber.com/message/[email protected]/1483555.html

Bernard Li <[EMAIL PROTECTED]> wrote:
Hi Steve:
 
Which version of SystemConfigurator are you using?
 
Cheers,
 
Bernard

From: [EMAIL PROTECTED] on behalf of Kan Tok
Sent: Fri 06/01/2006 21:04
To: [email protected]
Subject: [Sisuite-users] Grub issues with IBM HS20 Bladecenter

Good day,

         We have s everal IBM Blade Chassis (with HS20s) that we had been successfuly administering with an older version of systemimager and rh9.  We were required recently to upgrade to rhel 4.0 and, as such, upgrade our version of systemimager to 3.5.2.  We followed the excellent guide and files provided by Mr. Nielsen.  With that guide, combined with a couple of hacks mentioned below, we go t the image to be grabbed from a golden client built by hand and then pushed to new blades successfully.  However, when we try to push the same image to a blade with the same exact hardware, but with a BIOS level of 1.06 (as opposed to a BIOS of 1.05 of the machines that are working), the machine hangs on its first reboot after image push.  As mentioned below, even on machines that we get to work, it seems there's some sort of problem with how the bootloader is being handled (not sure what finally arranges that in the process, systemconfigurator?)

  The symptom on the non-working machine is that it gets past loading the BIOS and displays only the word: GRUB   and is unresponsive then to keystrokes.  We've verified that the image is being pushed to the new blade with the BIOS 1.06 (we modified the .master script so that it wouldn't partition the drives and put a sleep command in there so we could c trl-c to get to the command prompt of ash and we see the directory structure loaded under /a/ as it seems like it should, including the /boot/grub files), but it seems like something is wrong with how grub is being loaded to the MBR or its not being loaded at all (as I don't see any messages saying it is going to stage 1.5 and no menu for 2 is show.)

  I'm afraid I'm new to these listings, but I have followed them for almost two years now and have been impressed with the knowledge of the participants.&n bsp; Any suggestions, comments or even anecdotes on reaching a solution to this problem would be greatly appreciated.  Thank you.

Steve



"What we had to hack to get it working on an IBM Blade":

Although the rpms and guide by Ole Holm Nielsen were good, there were several issues we had to work through to get systemimager working on the blade hardware.  Our final solution involved using our original initrd from the old blades we had working with the old systemimager, adding some files to the boel_binaries.tar.gz,  such as /etc/init.d/functions I think (sorry away from the chassis as I write this) so that systemimager would have what it would need to execute the .master scipt properly.  The image would push fine (with some messages about unresolved symbols) but then after it was done pushing and the reboot happened the machine would first go through the BIOS as normal but then instead of loading GRUB properly it would display as message about grub.  I forget the exact wording but it was a paragraph describing some basic grub shell functionality and then gave the grub shell prompt.

At this prompt, we then mounted the grub bootloader with
root (hd0,0)
setup (hd0)
reboot

in the prompt.  Unfortunately, as I mentioned above, this prompt never appears with the version 1.05 I BM Blade BIOS.  It just sits with GRUB on the screen, unresponsive.  We've been toying with including /sbin/grub (and other appropriate files) to try to load the bootloader into the MBR by hand after the image is pushed (by taking the reboot out of the .master script) but whenever we try and perform a root(hd0,0) it gives an error saying there's no such device (which I guess makes sense considering everything is mounted kinda wonky at that point.) 

BIOS issues:

We've done some research on IBM's support site, but they have a very t hin level of documentation and they seem to have stopped carrying the 1.06 BIOS lev el in their archives.  Upgrading to the newest BIOS 1.10  would be a serious difficulty for us given some issues with version control their trying to enact around here.

If you've read all this, thanks, I really appreciate the help.
Yahoo ! Photos
Got holiday prints? See all the ways to get quality prints in your hands ASAP.

Yahoo! Photos
Got holiday prints? See all the ways to get quality prints in your hands ASAP.

Yahoo! Photos
Got holiday prints? See all the ways to get quality prints in your hands ASAP.

Reply via email to