How to update SUSE 10 sp2 without corrupting files

[Tekygirl]

I have a SUSE 10 sp2 server that I am using for hosting VMware and virtual machines. I have installed the basic host machine with only file server enabled (no dns, dhcp, print services or anything else). This is the fifth time I have had to reinstall because after I applied all the updates the system downloaded, it would not boot with a cd/dvdrom mounted. I tried everything I knew hot to do, which I admit isn't much (edited the fstab, tried to mount through command line) but nothing brought it back. this all started because I had the entire machine loaded with VMware, two virtual machines, and all was well until I decided to update the OES2 sp1 virtual machines. All manor of issues cropped up like corn in the summer. So I reloaded after a week because I couldn't fix the problems and decided in my limited knowledge the updates probably caused the issues, and there was no undoing the updates.
I assume (dangerous as that may be) that I applied too many updates to my bare system, but I would like a more experienced opinion. And by experienced, I mean definitely more than my six months!
What bare minimum of patches should I apply for this bare bones type of install before I put VMware 2.1 on it.
All help and insight is appreciated by this distraught Kansas girl.

Thanks.

Faith

[HROAdmin26]

It is likely that nothing is being "corrupted." Something may be "changed" that is causing the system not to boot. We do not know your hardware and do not know what patches you installed.

You can:

A) Install SLES 10.2 and call it a day - run it as-is.
B) Take more notes about what patches are being installed and/or troubleshoot the resulting boot failure. *In general,* watch out for grub and kernel patches. Build a test system in VMWare and install/upgrade/troubleshoot.

Most of the time, Linux boot issues are *typically* easy to identify. Resolving may involve fixing grub or rebuilding the initrd. 98% of Linux patches do not affect the boot process.

[Tekygirl]

First, the server actually booted, only the cd/dvdrom did not mount. I installed all of the updates that downloaded with the ZMD because I assumed the system would only download the updates I actually needed with my setup. It is obviously not that intuitive. I will be very careful what I actually apply to my system.
I have a sata Raid 5 with 1.8 tb space, 8 gb ram, and quad processors. I am using EVMS as my manager. I have set it up according to novell docs and it was working great until I tried to update the os. BAD IDEA! I will stick with the basic system, I suppose. We are a small shop and don't have the money to create another server just for testing purposes. Since I have been a Netware user for over a decade and it worked beautifully, this experience is very frustrating. I will end up calling Novell for support over this and hopefully they can tell me what to update and what NOT to update to keep everything working after the update.
Thanks for your insight. I appreciate you trying to help.

Faith

[HROAdmin26]

First, the server actually booted, only the cd/dvdrom did not mount.

So the entire OS was reinstalled because the CD didn't "mount?"

cat /etc/SuSE-release 
SUSE Linux Enterprise Server 10 (x86_64)
VERSION = 10
PATCHLEVEL = 2

ls -alh /dev/ | grep -i cdrom
lrwxrwxrwx  1 root root         3 2008-10-14 06:46 cdrom -> hda

mkdir /media/cdrom

mount /dev/hda /media/cdrom/

I too was a CNA for many years, and altho it was solid once configured, I really don't miss it.

[Tekygirl]

This last time with no virtual anything on it only caused my cd/dvd mount issue that I could tell. The very first time issues occurred from updating the vm's oes2 and suse 10 sp2 files. It caused my cd/dvdrom not to mount but also my vm's would not start back up after the reboot. The update completely hosed them. My partition tables were even messed up, although I am not sure how updates can do such a thing. Now I am trying to test just the host system with updates before I put anything else on the server. I am extremely frustrated with suse 10 and suppose I just need to take a deep breath and not sweat the small stuff...I am in the process of installing VMWare again to move forward.
Thanks.

[HROAdmin26]

The very first time issues occurred from updating the vm's oes2 and suse 10 sp2 files.

Changes inside a VM have *zilch* to do with the host machine/OS. If you don't like what a VM is doing or you think a VM is affecting the host machine, shut down VMWare (/etc/init.d/vmware stop) and I doubt you will see any difference. It's also a good practice to ensure floppy/cdrom's are started in a "disconnected" mode from the VM's. Only one VM or the host machine can access the cdrom at a time.

The update completely hosed them.

So in that instance, the *VM* was upated and would not boot. Again, I would point to grub/kernel updates *inside the VM.* Also remember that the hardware that a VM sees is *not* the physical HW on the host machine. Pretty much only the CPU model carries down inside the VM. The chipset, HDD controller, video, etc. are all a standard config presented by VMWare.

Also remember that when VMWare is installed, you run "vmware-config.pl" - this compiles/installs modules *tied* to your current kernel. If you then change the kernel, you will likely see these modules fail to load and VMWare will not work until you run vmware-config again.

[gogalthorp]

Also a good idea is to put the VM on a separate partition. It is easy to do when installing just add an addition partition and mount it. On my machine I mount a 30gig partition as /vm. Then when installing or configuring VMware I just tell it to put the VM files into the /vm directory which happens to be a seperate partition. Then if I need to reinstall I just tell the installer to NOT format the vm partition and mount it as /vm and also do the same with my home partition mounted as /home Then only the root partition gets reformatted and my VM's and home data are safe. So I can update the OS without touching /home or /vm

[Tekygirl]

I do exactly that for my vm. I put it on it's own partition, but I didn't realize I could just reload the boot partitions and keep my other partitions. Interesting. I will have to test it.

I have a couple of questions regarding setting up my host server. Should I use EVMS or LVM for partitioning? I plan on using nss volumes on my virtual machines but not on my host server. I use EVMS on the virtual partitioning, but is it necessary for the host server? And will EVMS still work on the virtual servers if I don't use if for the host server?

I appreciate the great feedback and insightful knowledge bank. Thanks. My frustration is great learning Linux but I am getting there.

[Tekygirl]

I have used the /vmware-config.pl. Thanks for all of your help.

[HROAdmin26]

The VM's and host are completely separate machines. The VM's only see a clean, unused disk when they first boot. The VM is running as a *process* under the host machine - but the VM *thinks* it is on its own hardware.

I have a number of volume managers in use - from LVM across the spectrum to VxVM.

In the case of LVM, I have tested and found a noticeable overhead to using it. Because of that, I would only recommend LVM when one of the following are *required*:

> Dynamic volume resizing - adding a new disk and increasing the volume size. (Filesystem must also support "growing.")
> Volume snapshots - although if the volume is hosting active VM's, LVM snaps will not guarantee a valid backup image.

So unless one of those is important, I skip LVM altogether and just create partitions > format > mount.

I have not done extensive testing on performance with EVMS. But I am sure there are some comparisons that Google can dig up.

For hosts running "non-critical" VM's, I tend to stripe the disks for max space and performance. For production usage, the disks are usually RAID10 (or RAID1.) Your RAID is provided at the hardware level (RAID5) and thus really isn't a concern at the OS level. If you need better performance, you may want to consider a RAID10 volume - lose some space, gain some performance.

Example numbers:

Software raid (md driver) - /dev/md1 is reiserfs:

hdparm -Tt /dev/md1

/dev/md1:
 Timing cached reads:   9000 MB in  2.00 seconds = 4502.81 MB/sec
 Timing buffered disk reads:  748 MB in  3.01 seconds = 248.87 MB/sec

Software raid (md driver) - /dev/md2 is xfs (VM volume):

hdparm -Tt /dev/md2

/dev/md2:
 Timing cached reads:   8448 MB in  2.00 seconds = 4227.36 MB/sec
 Timing buffered disk reads:  662 MB in  3.00 seconds = 220.38 MB/sec

If I make those /dev/mdX devices managed by LVM, the sequential reads will typically drop by about 50MB/sec.