PDA

View Full Version : IBM p615 SLES 10 boot with unused disks



trifox
09-Jan-2014, 22:39
IBM p615 7029-6e3
Five drives each with an OS: AIX 5.1, AIX 5.3, AIX 6.1, SLES 10, SLES 11. No problem booting any of these except SLES 10. If any drives in addition to the SLES 10 are powered up, the SLES 10 boot fails attempting to access the unconfigured drives. For example, SLES 11 boots and Yast shows the boot device /dev/sdb plus the other four devices as /dev/sda, sdc, sdd, and sde. Somehow SLES 10 is trying to do something with these other drives and fails. There must be some setting to at least background whatever it is doing but I cannot figure out what it is. Does this make sense and does anyone have some idea what I can set? I don't want to pull the other drives whenever I need to run SLES 10 (some customers need the older OS for our products) because my DASD backplanes are getting touchy.

trifox
10-Jan-2014, 01:35
Looking closer, it's not that it's trying to do something with the other drives but rather that it cannot load its own drive. The root device is /dev/sda3 which comes up fine w/o other drives powered up but otherwise cannot be accessed. Clearly IPL is able to access /dev/sda1 because the Penguin shows up, udevd starts, etc... yaboot.conf has the correct values but *somehow* having the other drives powered up conflicts with boot being able get to the root /dev/sda3 partition.

KBOYLE
10-Jan-2014, 05:51
trifox wrote:

> The root device is
> /dev/sda3 which comes up fine w/o other drives powered up but
> otherwise cannot be accessed

The first place I would start is the knowledgebase, if you haven't
looked already.
https://www.suse.com/support/kb/

I would also try to modify /etc/fstab and mount devices by device ID.
These names are persistent whereas device names (sda, sdb, ...) can
change leading to issues similar to what you describe.

--
Kevin Boyle - Knowledge Partner
If you find this post helpful and are logged into the web interface,
show your appreciation and click on the star below...

trifox
10-Jan-2014, 18:22
Hi Kevin,
I looked through the kb but did not get any hits. /etc/fstab does not come into play yet because the root fs has not been loaded, which is the problem. I think there is something about "hardcoded" device names though. IIUC the boot process works like this:

1. p615 firmware (SMS) ipls the disk chosen by the user
2. This loads the initial SLES boot code
3. The initial code creates the RAMDISK fs
4. "Somehow" the boot code gets the long boot disk identifier from the firmware and writes it into yaboot.conf
5. Also somehow the root device name gets written into yaboot.conf and I think this is the problem.

IIRC when I looked in yaboot.conf after the boot failure, it had /dev/sda3 in it which may be the problem. I think that it should have been /dev/sde3. I have five disks arrayed as

AIX 6.1
SLES 11
AIX 5.3
AIX 5.1
SLES 10

and when I boot SLES 11, it uses /dev/sdb3 as the root so I think the SLES 11 boot code assigned /dev/sda,b,c,d,e to the disks in their slot order. I entered various parameters on the SLES 10 boot: line without success. If I could get the correct entry for the boot line, I am fine with doing a manual process. I will experiment some more.

KBOYLE
10-Jan-2014, 20:10
trifox wrote:

> I entered various parameters on the SLES 10 boot: line without
> success. If I could get the correct entry for the boot line, I am fine
> with doing a manual process. I will experiment some more.

SLES10 is old... There have been many changes. It's difficult to
remember what has changed. Also, I have not worked with the IBM pSeries
and I expect there are differences in the way it boots when compared to
Intel platforms. If anyone else has any ideas, please feel free to jump
in.

Have you tried to compare /boot/grub/menu.lst to see what might be
different between SLES10 and SLES11?

Do you have console access when the boot fails? does "dmesg" provide
any clues?

Can you boot from a SLES10 CD and then successfully boot from the disk
where SLES10 is installed?



--
Kevin Boyle - Knowledge Partner
If you find this post helpful and are logged into the web interface,
show your appreciation and click on the star below...

smflood
11-Jan-2014, 00:39
trifox Wrote in message:

> IBM p615 7029-6e3
> Five drives each with an OS: AIX 5.1, AIX 5.3, AIX 6.1, SLES 10, SLES
> 11. No problem booting any of these except SLES 10. If any drives in
> addition to the SLES 10 are powered up, the SLES 10 boot fails
> attempting to access the unconfigured drives. For example, SLES 11 boots
> and Yast shows the boot device /dev/sdb plus the other four devices as
> /dev/sda, sdc, sdd, and sde. Somehow SLES 10 is trying to do something
> with these other drives and fails. There must be some setting to at
> least background whatever it is doing but I cannot figure out what it
> is. Does this make sense and does anyone have some idea what I can set?
> I don't want to pull the other drives whenever I need to run SLES 10
> (some customers need the older OS for our products) because my DASD
> backplanes are getting touchy.

Just to confirm I've understood the above correctly, you have five
separate drives each with a different OS installed and they all
boot with other drives present except for SLES10?

Which disk(s) is/are set up with Boot Loader? That's where'd I'd
start looking. If the SLES10 disk is bootable then I'd compare
the /boot/grub[2]/menu.lst from SLES10 and SLES11
disks.

HTH.
--
Simon
SUSE Knowledge Partner

----Android NewsGroup Reader----
http://www.piaohong.tk/newsgroup

KBOYLE
11-Jan-2014, 02:34
Simon Flood wrote:

> If the SLES10 disk is bootable then I'd compare
> the /boot/grub[2]/menu.lst from SLES10 and SLES11
> disks.

Isn't that what I said? :-)

Good suggestion!

--
Kevin Boyle - Knowledge Partner
If you find this post helpful and are logged into the web interface,
show your appreciation and click on the star below...

trifox
11-Jan-2014, 05:08
This is what I discovered.:
1. Boot SLES 11, look at bootloader configuration. The device dropdown menu (I'm writing this from memory) only shows one item, the very long ppc PCI+SCSI+partition identifier which I presume is somehow passed to the bootloader by the ppc firmware during IPL.
2. Boot SLES 10 after taking out other disks and the same dropdown shows only /dev/sda and /dev/sdb. When the boot fails, it drops me into a shell where I was able to mount /dev/sde3 but of course the boot will not continue. I'm wondering if I only had two drives running when I initially installed SLES 10.

I could not find any possibility to edit the dropdown menu to add /dev/sdc, etc... so I "solved" the problem by delicately (one of my DASD backplanes is very touchy) swapping the disks around so that the SLES 10 disk is the first one in order. Now all five boot correctly. Thank you both for your suggestions. I think they are correct and I probably could have figured how to get SLES 10's bootloader config to include the correct /dev/sde3. I know SLES 10 is ancient but then so am I and some of my customers still need executables linked with ancient GLIBC versions. Heck some of them still support customers on AIX 5.2 which is why I need to keep AIX 5.1 around! Thanks again.

KBOYLE
12-Jan-2014, 19:34
trifox wrote:

> Now all five boot correctly.

That's great!

Sometimes it's important to learn exactly why something isn't working
correctly. Often, a workaround is sufficient. Your solution is simple
and effective. I appreciate your taking the time to describe it.

Good luck.

--
Kevin Boyle - Knowledge Partner
If you find this post helpful and are logged into the web interface,
show your appreciation and click on the star below...