My preferred method is to replace the servers!

First read all of this  (never mind that x86 is in the URL, it does apply to 
Don't forget to check that all of your SW supports it.

While you can upgrade, new servers are less risky and allows you the 
opportunity to backout really easily if something goes wrong.  
We will do just a tiny bit of upgrading here since our applications tend to 
agree about the risk.
It also affords them the opportunity to easily move to new versions of the 
software they might have installed on the servers.
If you don't have things that freak out about a hostname change (like 
authentication products/certificates/etc) you can do the old switcheroo.  Put 
up the new server at another IP and temp hostname and then when all ready, 
shutdown the old server and change the new one's hostname and IP to the old 

Your Oracle people are probably used to replacement servers as well.   They 
will just unload and reload things.   

My checklist for upgrades includes
1. Backup entire server
2. Enlarge / file system and /usr
3. Shutdown security sw
4. Set root password to something you know
5. Edit /etc/zypp/zypp.conf and change solver.onlyRequires = true   (avoids 
pulling in "recommended" packages - many of which I did not want on our servers)
        and comment out         # multiversion = provides:multiversion(kernel)
6. Do a chkconfig -l and shutdown and turn off everything that is not SUSE
7. Save /etc/sysctl.conf settings you may have turned on
8. shutdown
9. Login to the virtual machine.  Be in CMS
10. punch files and start the installation - how to do this is documented in 
the installation guide
11.  Choose, Start Installation, Upgrade, configure your network, 
12.  At this point I choose VNC.  You may have another preference.
13.  Click what you need.  I always add the System z HW Crypto support
14.  Go get coffee
15. It will eventually reboot it self
16. When it comes back up, figure out if you have orphaned packages and remove 
- they will never be patched!
17. Remove other things that your installation might not want installed that 
came along for the ride
18. Join it to SMT or SUSE Manager or whatever you use and  zypper update it
19. Upgrade other things that need it - mine need security sw upgrades
20. Customize your stuff. Again, ymmv.  Here's my list

systemctl enable snmpd.service
systemctl enable ntpd.service
systemctl disable display-manager.service
systemctl set-default

grub2-mkconfig -o /boot/grub2/grub.cfg
dracut -f 
chkstat --system --set
mkdir /var/log/journal

21. Turn your services back on
22. Special considerations for GPFS if you have that (kernel cmdline and apply 
a fix pack to get its systemd stuff)
23.  Reboot again.  May sure everything starts OK.

So 2-3 hours maybe per server.  You'd be looking at a couple of weeks, less 
maybe with practice.
If you have a good deploy process, you can probably build 88 in a day or two.  
And then get the app people and dba's to do their stuff.

-----Original Message-----
From: Linux on 390 Port [mailto:LINUX-390@VM.MARIST.EDU] On Behalf Of Victor 
Sent: Thursday, February 8, 2018 12:56 PM
Subject: [LINUX-390] Upgrade from SLES 11 to SLES 12

We have 88 guests with SLES 11SP4 on z/VM 6.4, the majority with oracle 11 and 
12. The cookbook only show how to made a new installation. We want to know is 
there a way to upgrade those servers to SLES12. If true:

  1.  How can we achieved it. DVD, NFS, FTP.
  2.  How many servers can upgrade at the same time.
  3.  This upgrade can be done without graphical interface.
  4.  This could be online or offline.


Victor Echavarry

System Programmer

Operating Systems

For LINUX-390 subscribe / signoff / archive access instructions, send email to with the message: INFO LINUX-390 or visit
For more information on Linux on System z, visit

For LINUX-390 subscribe / signoff / archive access instructions,
send email to with the message: INFO LINUX-390 or visit
For more information on Linux on System z, visit

Reply via email to