RHEL 3 to RHEL 4 upgrade, EMC SAN and LVM

[Leibniz]

Hi everybody:

I performed an upgrade on a server yesterday from RHEL 3 AS to RHEL 4 AS but had to back out due to SAN connection issues. I'm hoping someone can shed light on this problem.

Environment:
HP BLp class Blades
RHEL 3 AS 32bit / RHEL 4 AS 32bit
EMC Symmetrix

The specific problem after the upgrade to RHEL4 was that , when we plugged the SAN cables back in, the boot process stopped immediately after it remounted / to rw, as the LVM management process was working. The end situation was LVM not being able to find any VGs, including the local disk VG. From there it dropped into maintenance mode.

Shutting down, unplugging cables and booting without SAN, and the system booted up fine. Even local VG.

As you will know, the upgrade from RHEL3 to RHEL4 means moving from LVM1 to LVM2

So specifically this problem looks to be focused on the LVM2 reading the LVM1 information from the SAN devices.

As I previously (last week) upgraded a very similar server successfully, I obviously focused on what I did differently from one upgrade to the last one.

The only thing that I recall is the order in which I upgraded and plugged the SAN back in.

i.e. First upgrade (successful)
1. shutdown, unplug SAN (commenting out SAN devices in /etc/fstab of course)
2. kickstart to upgrade, perform upgrade, shutdown.
3. plug SAN back in, boot, reimport previously exported VGs, uncomment SAN devices in /etc/fstab.
4. register and then up2date
5. Done. Successful.

Failed attempt.
1. shutdown, unplug SAN (commenting out SAN devices in /etc/fstab of course)
2. kickstart to upgrade, perform upgrade, reboot.
3. register/patch/full up2date to RHEL 4.8
4. plug SAN back in, boot,
> Errors. Can't vind any VGs.
> back out of upgrade.

So, my specific question is this: would a system update of the LVM2 package (while up2date-ing) BEFORE the SAN was attached, cause this problem?

That is what I suspect, but I would like to get confirmation on that idea before I invest hours to re-plan and re-execute.

Thank you for your assistance.

[Rob Leadbeater]

Hi,

Firstly, Red Hat do not support or recommend upgrading between major versions of RHEL.

You might get things working, but you'll be on your own with regards official support...

That said, looking at the differences in your two attempts, did you export the volume groups on the failed one ?

Cheers,

Rob

[Leibniz]

Thanks for your reply.

Although we can not do the upgrade in all instances, (i.e. from RHEL 3 to RHEL 5 or from 32bit to 64bit - re-imaging from scratch is required) in this scenario we have done this hundreds of times successfully. We will have no trouble with support for the upgraded servers.

Yes, I deactivated and exported the VGs. In fact, I had to import them back when we backed out of the change in order to get back to square one. All went successfully.

The problem is specifically on the how the new LVM2 first sees the LVM1 VGs on the SAN, and it seems to be related to the state of LVM2 (i.e. first incarnation of that RPM as opposed to getting updated via up2date.)

[Rob Leadbeater]

Hi again,

If you "...will have no trouble with support for the upgraded servers." then why not log a call with Red Hat and ask them ?

This KB doc is quite clear though:

https://access.redhat.com/kb/docs/DOC-3246

"WARNING: Use of the "upgrade" option during the installation process can cause erratic system behavior in various cases and is unsupported by Red Hat. If you experience problems with an upgraded system and are seeking assistance from Red Hat Support, you will be instructed to perform a clean installation."

Cheers,

Rob

[Leibniz]

Thanks again.

Well that certainly is clear.

I will initiate a call with RedHat to see where that gets us.