Boot failure after patching 5.3
 
*
Welcome, Guest. Please login or register. January 10, 2009, 12:48:41 AM


Login with username, password and session length


Pages: [1]   Go Down
  Print  
Author Topic: Boot failure after patching 5.3  (Read 1744 times)
0 Members and 1 Guest are viewing this topic.
oldwolf
New Member
*
Posts: 4


« Reply #7 on: August 20, 2007, 11:29:52 PM »

Short update of my own.
First, thanks for the pointer on the boot CD, it worked like a charm.
Once I was able to get into the system, I pulled down 5300-06 and ran the update which patched and repaired the system.
As an aside, I made no direct filesystem size changes for the patches. I'll dig through the logs a bit later and try and post any interesting findings from the initial failed patch.
Logged
Michael
Administrator
Hero Member
*****
Posts: 539


« Reply #6 on: August 20, 2007, 04:43:30 PM »

ok, a bit of an update here.

My nim server is at:
Quote
root@nim[ / ]:oslevel -s
5300-06-01-0000

/tmp is still 'small' only 48 MByte, and hd5 is 2x 16MB, or 32MByte

bosboot is 'reasonably' small....
Quote
root@nim[ / ]:bosboot -a

bosboot: Boot image is 32951 512 byte blocks.

Doing two client installs: one with all devices, one with all devices set to No.

All devices: number of filesets to install is: 843
Min. devices: number of filesets to install is: 287 - install time approx 14 minutes, followed by an error.
Quote
0301-152 bosboot: not enough file space to create:
              bootimage
              /tmp has 32366 free KB.
             bootimage needs 32530 KB.
BOS Install: Could not create boot image.
    ID#         OPTION
      1          Continue
      2          Perform System Maintenance and Then Continue
   Enter ID number:

I enter #2, enlarge /tmp, and run bosboot -ad hdisk0
Quote
# chfs -a size=+1 /tmp
Filesystem size changed to 98304
# bosboot -ad hdisk0

bosboot: Boot image is 20306 512 byte blocks

hd5 is one PP of 16 Mbyte, 20306 512 byte blocks is roughly 10M byte, it fits, so I really enter exit.
NOTE: through experience I have learned, the BOS Installer does not repeat the last failed statement (bosboot in htis case). You must bring the installation to the point it needs to be before using exit to complete the install. The text says: "creating the boot image" but that was only the message to display if the previous command was running successfully.

The ALL devices system has the same error message after 25 minutes...
The first bosboot attempt fails (/tmp is enlarged to same size above)....
Quote
# bos -ad hdisk0
dd: write error: No such device or address
32944+1 records in
514+1 records out

0301-161 bosboot: dd failed to copy bootimage

0301-165 bosboot: WARNING! bosboot failed - do not attempt to boot device.
I'll increase hd5 by 1 PP to 32M byte. And the command succeds....
Quote
# bosboot -ad hdisk0

bosboot: Boot image is 32945 512 byte blocks

Just a bit too large for a 16M byte boot logical volume.

Ok. The point of all this. When updating to a new kernel - verify that your new boot logical volume has been successfully created. Technology Levels, by default, install all devices.
Logged
Michael
Administrator
Hero Member
*****
Posts: 539


« Reply #5 on: August 17, 2007, 04:45:37 PM »

Well, as I took the quote from an AIX 5.2 system (my 5.3 system was not available at the time) there is sure to be some differences.

Note: if you regualrly use the /etc/rootvg.net don't forget to remove the entry ^./ or nothing will get on your backup.

Except that the *.data files may not be correct (prompt=no for example), you could probably use the command immediately, without any fake mksysb file to start with.

Testing....
« Last Edit: August 17, 2007, 04:49:15 PM by Michael » Logged
oldwolf
New Member
*
Posts: 4


« Reply #4 on: August 17, 2007, 03:49:49 PM »

Trying it now. When you go to create your image according to the instructions, it will probably error out. According to http://www-1.ibm.com/support/docview.wss?uid=isg1IY90008 , don't create the mksysb ahead of time, and you'll need to create a /etc/exclude.rootvg

Quote
Problem summary

The mkcd readme does not accurately reflect the current
way to create a boot only CD.

Problem conclusion

Updated mkcd readme.

Temporary fix

Create file /etc/exclude.rootvg using some editor and
put the following line in it:
   ª./
Execute the following command:
   mkcd -e -d /dev/cdx # Where cdx is your burner
OR
   mkcd -e -S  # This creates an image that can be
transferred to some system for burning.

Comments
Logged
Michael
Administrator
Hero Member
*****
Posts: 539


« Reply #3 on: August 17, 2007, 08:52:55 AM »

After some time to think and reflect on my experiences....

I have been installing 5306 (to give it a name) from my nim server, and what I notice is that /tmp is too small for the boot image to be built when all devices are being installed. Also, hd5 might be too small - especially in partitions and PP sizes of rootvg smaller than 64M byte.

My guess is that either /tmp or hd5 was too small is what went wrong with the update originally.

What I am going to try doing, and I suggest to you now, is to make a bootable CD, but with no mksysb image on it, just the boot stuff. This should be enough to boot your nim server, and do regular maintenance.

Hints on how to this are in /usr/lpp/bos.sysmgt/{mkcd|mkdvd}.README.{txt|html}

The interesting part is:
Quote
Creating boot-only media

   There maybe circumstances where boot only media is needed for instance to
   perform maintenance or to install using non-bootable media.  This boot media
   needs to match the level of system that maintenance is being performed on or
   the non-bootable install media being used.  Therefore it is preferrable that
   the boot-only media have been created on the same system that maintenance is
   to be performed or that the non-bootable backup came from.  Additionally,
   installp failure messages may appear toward the end of the install because
   installp is attempting to find additional software to install, but no
   additional software is on the boot-only media.  These messages can be
   ignored.

   Execute the following commands to create boot-only media:

  # cd /var/adm/ras
  # ls ./bosinst.data ./image.data | backup -ivqf/tmp/fakemksysb
  # mkcd -m /tmp/fakemksysb -d /dev/cdx     # Where cdx is an attached cdwriter
    OR
  # mkcd -m /tmp/fakemksysb -S   # Will create a CD image which can be
                                 # transferred to a system with a cdwriter

   Note:  The final mkcd command in the previous example makes an image that
   could be transferred to another system (AIX or non-AIX) for burning.
   Note:  The bosinst.data file must be set for a prompted install (PROMPT = yes
).

Keep us informed!

p.s. I am going to make mine a CD rather than a DVD, to be more portable.
p.p.s. Everyone trying this: make sure you have installed ALL DEVICES beforehand or it may not boot that one system you want it to! And make sure /tmp is large to handle the very large bootimages.
Logged
oldwolf
New Member
*
Posts: 4


« Reply #2 on: August 16, 2007, 11:50:59 PM »

It seems unfortunately not. chroot doesn't want to let me get my temporary tree moved to the working root. When I try to run bosboot anyway, I get complaints about bootinfo being called with -v flag.  Huh Generally I'd just call a re-image from my NIM server, except this _is_ the NIM server.  Undecided Do you know of a way to force the CD to boot a 64 bit kernel for maintenance? It looks like it's starting a 32 bit kernel.
Logged
Michael
Administrator
Hero Member
*****
Posts: 539


« Reply #1 on: August 16, 2007, 10:59:20 PM »

Are you able to run bosboot while in maintaince mode - after mounting the filesystems by hand?

I am guessing that the problem is that the kernel on your CD is not the same as the kernel on disk, and that is giving you your problems.

A nim server with a spot at the same level (TL level) should do the trick, if the CD boot doesn't work.
Logged
oldwolf
New Member
*
Posts: 4


« on: August 16, 2007, 10:00:44 PM »

I installed some of ML 05 patches for AIX 5.3 on a 9110 yesterday. After reboot it only gets as far as the welcome to AIX and kernel identification message and then reboots. I can boot from CD, and as long as I call the shell before mounting the filesystem, I can hand mount the filesystems and they all appear to be intact. As soon as I leave the shell, and it attempts to mount the filesystems itself, or if I choose to let it mount them prior to bringing up the shell, it gets a scrolling list of getrootfs killed messages before being dropped into the kernel debugger. Since I can't seem to get into maintenance mode and try to back out the patches, beyond a re-install/restore, I'm out of ideas, any suggestions?
Logged
Pages: [1]   Go Up
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.2 | SMF © 2006-2007, Simple Machines LLC

Valid XHTML 1.0! Valid CSS! Dilber MC Theme by HarzeM
Page created in 1.167 seconds with 18 queries.




eXTReMe Tracker

Terms of Use and Privacy and Security Policies
Copyright 2001-2008 Michael Felt and ROOTVG.NET