offline upgrade to OES SP2 fails

[Frank Langner]

I want to upgrade a server from OES 2 SP0 to SP2 using the offline
upgrade method.

I've tested it with a server (VMWare session) without NSS volumes, that
worked fine.

As soon as the server (VMWare session set back to the beginning) get's a
NSS volume (on a separate disk, therefor *not* the ugly case with EVMS
on the boot disk!), the YAST upgrade does not find the boot partition
anymore. The partition list is just empty (in opposite to the first
test, when it listed the boot partition only). I can check the "show all
partitions" checkbox and after that I get a list, where some of the
partitions are listed twice: my original partitions "/dev/sda1" and
"/dev/sda3" (which is /boot and /), but also a "/dev/evms/sda1" and
"/dev/evms/sda3". These evms-partitions do not exist when the server is
booted regulary and therefor seem to be a result of a faulty partition
detection of the installation program. Also shown is a partition
"/dev/sdb1" for the NSS volume (shouldn't this be detected as an evms
managed partition!?).

Documentation says to select the partition with "evms" in its name, but
this does not work at all (and makes no sense to me, because the sys
disk partitions are not managed by evms).

Selecting the /dev/sda3 partition instead does allow the install to
proceed (but fails to mount /dev/sdb1, but I ignored that, because this
nss partition is not necessary for the upgrade).

But after the first reboot (during install), the system does not start
anymore. It does not find the / partition (I guess) and looking at
/etc/fstab makes clear why: all /dev/sdbax partitions have been changed
to /dev/sdbx!

After changing this back to /dev/sdax and reboot, the install continues,
but later on fails to upgrade eDir. During bootup I can see several
error messages from ipfilters etc., maybe the install messed up the
network card configuration also.

After all; a whole mess and no working server. Therefor: how to upgrade
a server with nss volumes (again: no evms on sys disk, just lvm)?

Thanks in advance!
Frank

[Frank Langner]

> Sounds like you have the EVMS partition on the same disk. Is this the
> case? If so, then it probably has a path of /dev/evms/sda1 or something
> like that. Is that the point where you are failing? If both are true
> then modify the path to remove evms from the path. It it is known issue
> and is documented. If not then sorry for the misunderstanding.

No, EVMS is not on the system disk. At least not as far as I had any
influence on it. I've installed the system by using the system disk only
and after the OS was installed completely, I have added the second disk
and used it to create a NSS volume. Therefor, if there is no automatic
change in the background, the system disk should be managed by LVM only.

This is the current fstab:

/dev/sda3 / ext3 acl,user_xattr 1 1
/dev/sda1 /boot ext3 acl,user_xattr 1 2
/dev/sda2 swap swap defaults 0 0
proc /proc proc defaults 0 0
sysfs /sys sysfs noauto 0 0
debugfs /sys/kernel/debug debugfs noauto 0 0
usbfs /proc/bus/usb usbfs noauto 0 0
devpts /dev/pts devpts mode=0620,gid=5 0 0
/dev/fd0 /media/floppy auto noauto,user,sync 0 0

I have a system with EVMS on the system disk (has an NSS volume on the
system disk), there fstab looks like:

/dev/evms/lvm/system/sys_lx / ext3 acl,user_xattr 1 1
/dev/evms/sda1 /boot ext3 acl,user_xattr 1 2
/dev/evms/lvm/system/swap swap swap pri=42 0 0

Therefor in my opinion the system I try to upgrade does *not* have EVMS
managed system partitions. Am I correct?

In addition: grub.lst shows "root=/dev/sda3" for the root volume,
whereas the other system (the one with EVMS on the same disk) shows
"root=/dev/evms/lvm/system/sys_lx" instead. Again a proof for me that
the system disk of the first system is not managed by evms.

But what is confusing me: when I take a look into evmsgui, all volumes
(including the sda1 to sda3) are displayed as /dev/evms/-volumes (Plugin
is DosSegMgr)! Does that mean that EVMS did indeed take control of these
volumes?? Why!? Or is it just a wrong display of evmsgui?

Thanks in advance,
Frank

[Frank Langner]

> You don't have it on the same disk if you didn't specifically put it on
> the same disk. And the mapping that you are seeing is normal. That is
> how NSS displays the mounts is through /dev/evms/.... Don't confuse this
> with the actual mount point under /media/nss or whatever you specified.
> This is just how NSS lists the devices.

Ok. Vexing, but ok.

But any idea how I can perform the update install without messing up the
system? The install program (booted from SLES10SP3-CD) can not mount the
'evms-version' of the root volume, whereas selecting the non-evms-
version creates a system that is not ok anymore (as described in my
first posting).

Frank

[Frank Langner]

Anyone having an idea regading this problem? Anybody having upgraded a
server with NSS volumes from OES 2 SP0 to SP2 ?

Thanks in advance,
Frank

[Frank Langner]

> I will attempt the upgrade on a system with an NSS volume that is NOT
> using EVMS and let you know how that turns out too. I do know that I've
> installed SLES 10 SP3 / OES2 SP2 directly from the media (my good media
> ;-) on a new server, and it worked fine.
>
> I am glad there is someone else in this boat!

Glad to hear that I am not the only one having this issue. :-) I already
raised a SR on this, but so far no useful response from Novell at all.
:-(

I keep you updated.

Frank

[DrumDude]

Yup I am toast 70% through the install. I now have a split kernel. I am
just wiping it out and starting from scratch and reinstall an eDirectory
server on OES after a crash scenario.
Dumb EVMS !
I still have one more server to upgrade in this scenario so I still need
to figger out what the EVMS issue really is.
Cheers

[Frank Langner]

In article <sQr4n.4786$q93...@kovat.provo.novell.com>, DrumDude@no-
mx.forums.novell.com says...

> Yup I am toast 70% through the install. I now have a split kernel. I am
> just wiping it out and starting from scratch and reinstall an eDirectory
> server on OES after a crash scenario.
> Dumb EVMS !
> I still have one more server to upgrade in this scenario so I still need
> to figger out what the EVMS issue really is.
> Cheers

If possible, please raise an SR on this issue. Maybe it helps when they
see that there are multiple persons having this issue. But so far,
assistance from "world class support" is very very poor.

Did you install all available online updates for your current platform
before you tried to perform the offline upgrade to SP2? This is a test
support wants me to do, but I can not because my test server does not
allow me to register my new activation codes (the next very annoying
problem).

[Frank Langner]

> Same problem. One identical server went via move_to_oes_sp1 and then to
> OES 2 SP2a SLES 10 SP3. This did not have the problem and is now OK.

Based on the SR, Novell will publish a fix for the problem within the
next few days. Will be a Driver Update Disk that has to be used in
conjunction with the original SLES10SP3 boot media. There's a bug in
hwinfo causing the issue.

[Frank Langner]

> Could you PM me your SR# so I can ask Novell for more details on this.

I would, but it seems I'm to dump to translate your address into a
proper email address... :-)

[Simon Flood]

Sorry I was thinking you were a HTTP user of forums.novell.com - I'll
send you an email.
--
Simon
Novell Knowledge Partner (NKP)

------------------------------------------------------------------------
Do you work with Novell technologies at a university, college or school?
If so, your campus could benefit from joining the Novell Technology
Transfer Partners (TTP) group. See www.novell.com/ttp for more details.
------------------------------------------------------------------------