pk3 -> pk5 update, lsm problem help!

[Michael Schulte zur Sur]

Hi all,

I have updated tru64 5.1b to pk5 and after that the machine fails to boot. I says something about error in disk group configuration copies, no valid disk.. transactions are disabled.
Trying lsmbstartup at single user mode resulted in ipc error could not contact blabla.

Please help,

Michael

[Michael Schulte zur Sur]

One thing more.
I did the update in multiuser mode and the ipc error came because I tried a volrecover.
It says during lsmbstartup something about disk read error on dsk0h.
This I can't believe, because the system is mirrore by a kzpac and no disks are failed.

Michael

[Venkatesh BL]

Was dsk0h part of rootdg?. Do you have the root/usr/var under LSM?

What were you using the volrecover on?

[Michael Schulte zur Sur]

Hi BL,

thanks for posting so quickly.
Given my dislike for lsm I hope I did not do that. How do I find out if dsk0h is lsm managed?
I used volrecover without any parameters.
I am at home and have no access to the machine, so it has to go through the operator.

Thanks for trying,

Michael

[Venkatesh BL]

If dsk0 is accessible, then 'fstype' of dsk0h (from disklabel output) should tell you if LSM is using it.

[Michael Schulte zur Sur]

BL,

yes, it is managed by lsm.

Michael

[Venkatesh BL]

Can you provide the o/p of following:

1) # volprint
2) # voldctl mode
3) # voldctl list
4) # ps -A |grep vold

[Michael Schulte zur Sur]

BL,

volprint:
no diskgroups loaded
voldctl mode:
mode disabled
voldctl list:
volbootfile
version 3/1
seq# 0.3
hostid 15891:17:34:18
entries:
disk dsk6 type=sliced

Thanks,

Michael

[Michael Schulte zur Sur]

BL,

I have discovered that we have a software contract and I opened a case with HP.
After some trials the engineer gave up!
We are now waiting for someone to come to our place.

Michael

[Venkatesh BL]

hmm...actually, if your root/usr/var doesn't have any dependency on LSM, you should be able to atleast boot the system properly (may be by disabling the lsmstartup script).

[Michael Schulte zur Sur]

BL,

thanks for your time. We had a technician at our workplace. Actually I am not to sure what all was done but I will have to request a statement because the patch was the cause for the disaster. We managed to boot to single user mode by replacing the LSM label on usr, disabled LSM in sysconfigtab. We could mount usr, restored /, /var and /usr but still had some problems with restoring the LSM label. After booting from CD the technician was able to restore the label and fix the problem.

Michael

[Venkatesh BL]

Glad to hear that the problem is finally resolved!

[Johan Brusche]

Glad it's working again.
Some observations ....
Via voldctl list, /etc/vol/volboot content tells us dsk6h was used to create rootdg.

Do you remember if dsfmgr was used to move
or interchange dsk6 with dsk0 ? Is dsk6 still there does it have also a partiton-h
with size 4096 ?
Any reference to LSM in cnx-partitions-h ?
cfr clu_bdmgr -d --> /dev/disk/dskn|simp::

What was the uptime, and was this rolling patch-install the first member reboot since then ? Maybe the problem was already there,
just waiting for a reboot to manifest itself.

Disklabels on disks containing root partitions are already locked at boot, you can change in-memory copies, but not on-disk.

__ Johan.

[Michael Schulte zur Sur]

Johan,

we did not use dsfmgr and this is not a cluster. It is a standalone machine. The simp partition is on dsk0. I have now heard by rumor that this is not supported. Can you confirm this? If so, why is there no check on this?

thanks,

Michael

[Venkatesh BL]

Which is not supported?. Can you elaborate?

[Michael Schulte zur Sur]

BL,

my colleague was told by HP, that patching a stand alone system while in multiuser mode is not supported.

Michael

[Johan Brusche]

That HP person made a quite restrictive interpretation of the "Warning" notice, that you will find also in your /var/adm/patch/log/session.log

To me it means: multi-user ie. "init state 3" is supported, with restrictions...

----quote----

*** Installation Warning ***

You have chosen to install patches onto this system while it is running in
Multi-User mode. Some patches may directly affect core operating system
operations. To ensure the proper operation of all applications, it is
strongly suggested that you install these patches while the system is in
Single-User mode. If this cannot be done, install these patches when the
system is as lightly loaded as possible (i.e. not running production
environments, no users logged on, etc.).

Do you wish to continue? (y/n) [y]: y

----end_quote----

The critical things are p.ex. the substitution of shared libraries. You can understand that applications like Oracle in production mode will shoke while in-use shared libraries are disappearing.

___ Johan.

[Michael Schulte zur Sur]

Johan,

that I would understand. But I stopped all applications before and there were no errors in the log.

Thanks,

Michael

[Venkatesh BL]

Michael,

This is strange to me. I am not sure why he said that its not supported. If that's the case, they could have removed that option completely (from patch installation).

[Michael Schulte zur Sur]

BL,

that would have been a wise course of action. Check if machine is a member of a cluster. If not, remove multiuser option.

greetings,

Michael

[Michael Schulte zur Sur]

Johan,

I have got the official report on the case stating that HP requires single user mode when patching non cluster systems. This seems somewhat kind of a desperate try to explain something strange. Since there were no error messages whatsoever I have problems believing that there were files that could not be patched because they were open.

Please comment.

Michael

[Johan Brusche]

I still maintain my statement of support.
Below an extract of paragraph 1.8.1 in the dupatch "Installation Guide"

--- quote ---

1.8.1 When Single-User Mode Is Recommended

Although you can install patches in multiuser mode,
we recommend that you bring down your system to
single-user mode when installing patches that affect
the operation of the Tru64 UNIX operating system or
the product you are patching. If your system must
remain in multiuser mode, apply the patches when the
system is as lightly loaded as possible.

There are no restrictions on performing patch
selection and preinstallation checking in multiuser mode.
Patch removals can only be done in single-user mode.

--- unquote ---

I know you need this, because you cannot remotely connect to the console. Even for those situations, dupatch offers a menu-selection, that reboots the system with networking enabled. But if I remember well, you only have an X25 line to the site, it all depends if the destination site's X25-gateway allows you to telnet/rlogin towards the Tru64 to be patched.

__ Johan.