error during online JFS installation

[RAJESH GANGADHARAN]

Hi
While trying to install Online JFS from swinstall GUI, I am getting some error message. Analysis phase completed during Installation phase;whilr executing swcleanup it is giving some error and coming out of installation. I found that a new kernel was build.

Following is the error message ( install_cleanup script was running with set-x option)

can any one suggest what could be the problem

LOGFILE
---------

* Beginning the Install Execution Phase.
* Filesets: 5
* Files: 45
* Kbytes: 576
* Installing bundle "B3929BA,r=B.11.00" .
NOTE: Saving the current system file at "/stand/system" to
"/stand/system.prev"
NOTE: The template file has been extracted from "/stand/vmunix"
It has been placed in "/stand/system" where it will be used
to build a new kernel.
* Installing fileset "AdvJournalFS.VXFS-ADV-KRN,r=B.11.00" (1 of
5).
NOTE: Building a new kernel based on template file "/stand/system"
WARNING: No such device, tunable, or cdio vxadv.
Ignoring the following line.
vxadv
Generating module: krm...
Compiling /stand/build/conf.c...
Loading the kernel...
Generating kernel symbol table...
* The current kernel function set directory at
"/stand/dlkm" has been moved to
"/stand/dlkm.vmunix.prev".
* The current kernel at "/stand/vmunix" has been moved
to "/stand/vmunix.prev".
* The newly built kernel has been moved to "/stand/vmunix".
* The newly built kernel function set directory has been moved
to "/stand/dlkm".
* Installing fileset "AdvJournalFS.AJFS-ENG-A-MAN,r=B.11.00" (2
of 5).
* Installing fileset "AdvJournalFS.AJFS-JPN-E-MAN,r=B.11.00" (3
of 5).
* Installing fileset "AdvJournalFS.AJFS-JPN-S-MAN,r=B.11.00" (4
of 5).
* Installing fileset "AdvJournalFS.VXFS-ADV-RUN,r=B.11.00" (5 of
5).
+ SUCCESS=0
+ FAILURE=1
+ WARNING=2
+ SHARED_ROOT=/export/shared_roots/
+ PRIVATE_ROOT=/export/private_roots/
+ print * Running install clean command /usr/lbin/sw/install_clean.
* Running install clean command /usr/lbin/sw/install_clean.
+ UTILS=/usr/lbin/sw/control_utils
+ [[ ! -f /usr/lbin/sw/control_utils ]]
+ . /usr/lbin/sw/control_utils
+ SUCCESS=0
+ FAILURE=1
+ ERROR=1
+ WARNING=2
+ EXCLUDE=3
+ GLOBAL_ERROR=11
+ PATH=/usr/lbin/sw/bin:/var/adm/sw/sbin:/sbin:/usr/bin:/usr/ccs/bin:/usr/sbin:/usr/lbin/sw
+ ROOT=/
+ SW_CTL_SCRIPT_NAME=install_clean
+ _pf=AdvJournalFS.AJFS-JPN-S-MAN
+ PRODUCT=AdvJournalFS
+ FILESET=AJFS-JPN-S-MAN
+ [ AJFS-JPN-S-MAN = AdvJournalFS.AJFS-JPN-S-MAN ]
+ RETURN_STRING=
+ RECURSE_LEVEL=0
+ [ -z /var/adm/sw ]
+ SW_TMPDIR=/var/adm/sw/tmp
+ SW_CU_DATESTAMP=200008
+ [[ / != / ]]
+ INSTALL_TLINK=/opt/upgrade/bin/tl_install
+ [[ -x /opt/upgrade/bin/tl_install ]]
+ /opt/upgrade/bin/tl_install
NOTE: tlinstall is searching filesystem - please be patient
ERROR: Can't read mount table
+ exit 1
ERROR: The command "/usr/lbin/sw/install_clean" failed. Check the
above output for details about the failure.
NOTE: The Install Phase has suspended. Check the above output for
reasons.


* Aborting the Install Phase.
Thanks in adv
Rajesh

[Bruno Vidal · Accepted solution]

Hi,
There is an hight probability that you have something wrong in your /etc/fstab.
What you can do:
-rm /etc/mnttab; bdf (one or two times).
mnttab is only an image of wich filesystem are mounted. If it doesn't exist, the system rebuild it. So if you have any mismatch between this file and the kernel you can have this kind of error.
-you can also have something wrong in your /etc/fstab. So edit it, and verify that everything is fine, then do "mount -aQ".

If everything is fine after that, it should be okay.
At last, I've already seen this kind of error because "/" wasn't in the fstab file, so verify that it is really okay.

Cheers.

[Massimo Bianchi]

Hi,
looks like you hace problems with what is already installed.

Do you have PHSS_18543 installed? This disables the vxadv, but the re-install procedure is VERY complex and may lead to an instable system. So pay attention if you have to re-install it. It is damned known as the Line In The Sand patch...

Most probably your situation is this.

If you need help in the reinstall of this patch, i suggest you to do a serach in the forum or to open a new thread.

Please check and post the output of :

/usr/sbin/swlist -l fileset -a state AdvJournalFS.VXFS-ADV-KRN | grep
AdvJournalFS.VXFS-ADV-KRN

Here are some notes for reinstalling the patch,
but i lack of a precise procedure.

PROBLEM
Customer reinstalled [PHKL_18543/PACHRDME/English] on his 11.0 system to activate AdvJfs but
he selected the "Reinstall filesets even if same revision exists" and
"Reinstall files even if same one already there" options. Swinstall
errored with "unsatisfied symbols" during kernel regen and the system panicked
when booting off the new kernel.

Reinstalling [PHKL_18543/PACHRDME/English] in this manner causes a problem because [PHKL_18543/PACHRDME/English] is a
"line in the sand" patch and shares filesets with many superseding patches. If
[PHKL_18543/PACHRDME/English] is installed with the reinstall options it overwrites filesets from
later patches.

RESOLUTION

The resolution is to reinstall the patches whose filesets were overlayed. Make
sure the customer has a full backup before proceding and is prepared in the
case
that the OS needs to be reinstalled.


1. If the system won't boot off the new kernel interrupt the boot process and
boot from /stand/vmunix.prev. After the system is booted make sure you copy
vmunix.prev to vmunix.good so it won't get overlayed by the next kernel build.
If there is no space in /stand then overlay vmunix with vmunix.prev. At this
point vmunix.prev is the only bootable kernel!

2. Make a list of the patches have [PHKL_18543/PACHRDME/English] as a prerequisite by running the
following command:

swlist -l fileset -a prerequisite PHKL_\*|grep 18543

3. Cross the patches off the list from step 2 if they have been superseded.
Identify superseded patches with:

swlist -l patch \*\.*,c=patch | grep superseded

4. Identify the patches remaining on the list that are corrupt by running
swverify against the patches. The command is:

swverify Patch_Name

If the patch is ok swverify will finish with "* Verification succeeded."

If swverify has errors such as the following then the swagent.log should be
examined to determine if the patch is corrupt.

"* The analysis phase failed for "hostname:/".
* Verification had errors.

5. All the corrupt patches on the list must be reinstalled. Both reinstall
options will have to be marked this time. Deselect the "Autoselect
dependencies
when marking software" to prevent other patches from being installed.
Swinstall
should be able to build a good kernel and reboot
successfully.


HTH,
Massimo

[Massimo Bianchi]

hi,
here is a link for the 18543,
but i suggest you to read all the forum thread about this.

They can be VERY VERY helpfull.

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x54f6e822e739d711abdc0090277a778c,00.html


HTH,
Massimo

[Caesar_3]

Hello!

Clean your mount table or disable in swinstall
mount option.
Also could be that you don't have some patch.

Caesar