vmware on dl380 with sles9

[Daniele Bernazzi]

Hi, I am experiencing troubles with vmware gsx 3.1 on a server proliant dl380 g3. The host os is linux 2.6.5-7.147-smp (from suse sles9 patched); hyperthreading is enabled; the raid controller is smartarray 641, driver cpq_cciss-2.6.2-11. I did install vmware and did build a virtual machine windows 2003; windows installation start, but at the format time the system hang for about 15 mins, also host os is very slow. It shows high perc of waiting time. After this it goes on ok until reboot of the virtual machine. Looks like it has some difficults creating the paging file, in fact in the system events of windows I can see hard disk error creating the paging file. In the host log file I can see: SCSI host `cciss' has no error handling. Is anybody can give me some clue or some hint to troubleshoot the problem? Ora has someone already realized a working system with:
hp dl380 g3 with dual xeon 2.8G and smart array 641
host os: linux
vmware: gsx 3.0 or 3.1
guest os: windows 2000 or 2003
Thank you

[Kai-Uwe Schurig]

Hi,

first note that SLES 9 is not official supported as a host system for VMWARE GSX 3.1.
Anyway, i have sucessfully installed VMWARE GSX 3.1 Build 9089 on a HP DL 360 G3 with SLES 9 SP1 ;-)
The virtual machines (SLES 8, SLES 9, Windows Server 2003) are running without any trouble.
Here are some hints from my setup, see also

http://www.vmware.com/community/thread.jspa?threadID=9700&tstart=30

1. Build vmware modules for your kernel

- install latest kernel via YOU
- install kernel-source, gcc, make, binutils
- cd /usr/src/linux
- make cloneconfig
- make modules_prepare
- download latest Build of GSX
- rpm -ihv VMware-gsx-3.1.0-9089.i386.rpm
- vmware-config.pl
- None of the pre-built vmmon modules for VMware GSX Server is suitable for your running kernel.
Do you want this program to try to build the vmmon module for your system
(you need to have a C compiler installed on your system)? [yes]
- Using 2.6.x kernel build system.
...
...
- The module loads perfectly in the running kernel.
- configure networking an so on and finish your installation
- install VMware Management Interface (see 2. and 4. to avoid some trouble)
- Installing the VMware Virtual Machine Console (see 3.)

Note:

If you update your kernel later, you have to do this procedure again (install new kernel + kernel soure, make cloneconfig, make modules_prepare, run vmware-config.pl, build new modules, leave all other settings like neworking and so on).


2. Trouble with GSX Management server (Apache 1.3.28) and LDAP

The GSX Management server uses his own Apache Webserver.
If you use LDAP for user authentification, the GSX management server may produce a segmentation fault like this: /var/log/vmware-mui/error.log child pid 5572 exit signal Segmentation fault (11)
Workaround: disable LDAP in Yast and use local authentication (passwd)
Note tat this change may affect other services which are based on LDAP ;-)

3. Problem with xinetd, authentication and VMWare Remote Console

The default port for the remote console (902) may be already in use by another service
Solution: select a different port (e.g. 904) or comment out the settings for port 902 in /etc/services
Please note hat you have to select this port while connecting with the remote console ;-)
Also xinetd must be running on the server.

4. Problems with authentication (vmware-authd)

The default settings of vmware-authd did not work for mee, i changed the following settings in
/etc/pam.d/vmware-authd and /etc/vmware/pam.d/vmware-authd:
auth required /lib/security/pam_unix2.so
account required /lib/security/pam_unix2.so

Hope this helps,

Kai-Uwe

[Daniele Bernazzi]

Thank you Kay-uwe for your reply; unfortunately after your suggestions the problem is still on.
I did reinstall vmware with rpm (first time was installed from .tgz) and did rebuild as you wrote in steps 1 (except installation Virtual machine interface and virtual machine console, since i am using the console on my pc).
I had just discovered the problem will not arise if I use fat32 file system for the vm win 2003 server; I guess also for a linux guest os there will be no problem (may be I'll try). So, tomorrow i'll try to setup the server with fat32 disk and after to convert it to ntfs (I can bet at the convert time the problem will popup again).
In my opinion the problem is in the cciss driver, but I can't see an escape.
What kind of array controller you have in your dl360? You use cciss driver? Which level? As far as you know SLES8 is certified in my environment?
Thank you very much.
Daniele

[Kai-Uwe Schurig]

Hi,

i'm using the Smart Array 5i with the cciss version 2.6.4 which comes with kernel 2.6.5-7.145-smp. You use the cciss from the HP website, maybe a switch to the cciss from the kernel can help.

For suported host systems with VMWARE GSX 3.1 look at:

http://www.vmware.com/products/server/gsx_specs.html#HostOS_linux

SLES 8 is supported, i have not tried it.

How have you configured your virtual hard disk? I use IDE 0:0 with a size of 10 GB (split into 2 GB files and disk space not preallocated), IDE 1:0 is my virtual cd-rom.
The virtual filesystem is NTFS, i had no trouble so far. Any hints in the vmware.log?

Regards,

Kai-Uwe

[Kai-Uwe Schurig]

One additional hint:

Make sure that you use the latest System ROM (version 4.09 P29-09/15/2004) and the latest firmware for the smart array 641 (version 2.34 A), they are available at:

http://h18004.www1.hp.com/support/files/server/us/locate/1110_4706.html

Kai-Uwe

[Daniele Bernazzi]

As supposed, after fat32 to ntfs conversion the problem reappear. On sles9 logs I can't see any more advise; while on win2003 log event I have this warning: "an error was detected on device \Device\Harddisk0 during a paging operation". I also have the error "device \Device\Scsi\symmpi1 did not respond within the timeout period", but this error arised also with fat32 fs.
The bios levels were already updated as you suggested; about the cciss driver I am not sure about the level: the version I reported is from rpm query, but I did install SP1 whith .147 kernel and I guess it did replace the driver because trying to erase the rpm, gave me the message: "The running kernel (2.6.5-7.147-smp) is not supported
by this binary RPM."
So I am very tired about this buggy SLES9 and if I don't know what to do: may be migrate to RH AS4 could be a good chance?
Thank you very much for your suggestions.
Daniele

[Kai-Uwe Schurig]

Hi,

what gives "modinfo cciss" ?
The message from the PSP setup indicates that the kernel you are using is not supported by the cciss from HP. But no problem, i also use the cciss 2.6.4 from the kernel and had no problems so far.
Any strange entries in the vmware.log ?
Enough space left on the hard disk?
Any other disk intensive tasks are running?
What about the virtual disk settings i asked about in my last response?
Have you tried the latest update kernel from Suse (kernel-smp-2.6.5-7.151), don't forget to rebuild the vmware modules after ther kernel update.


Kai-Uwe

[Daniele Bernazzi]

# modinfo cciss
license: GPL
description: Driver for HP Controller SA5xxx SA6xxx version 2.6.4
author: Hewlett-Packard Company
alias: pci:v0000103Cd00003220sv0000103Csd00003225bc*sc*i*
alias: pci:v0000103Cd00003210sv0000103Csd00003211bc*sc*i*
alias: pci:v00000E11d00000046sv00000E11sd0000409Ebc*sc*i*
alias: pci:v00000E11d00000046sv00000E11sd00004091bc*sc*i*
alias: pci:v00000E11d00000046sv00000E11sd0000409Dbc*sc*i*
alias: pci:v00000E11d00000046sv00000E11sd0000409Cbc*sc*i*
alias: pci:v00000E11d00000046sv00000E11sd0000409Bbc*sc*i*
alias: pci:v00000E11d00000046sv00000E11sd0000409Abc*sc*i*
alias: pci:v00000E11d0000B178sv00000E11sd00004083bc*sc*i*
alias: pci:v00000E11d0000B178sv00000E11sd00004082bc*sc*i*
alias: pci:v00000E11d0000B178sv00000E11sd00004080bc*sc*i*
alias: pci:v00000E11d0000B060sv00000E11sd00004070bc*sc*i*
depends: scsi_mod
supported: yes
vermagic: 2.6.5-7.147-smp SMP 586 REGPARM gcc-3.3

no strange entries in vmware.log
On the hard disk I have 2G used on 50G available; the server is just running an apache2 service with 2-3 connection/day, it's doing nothing. vmware is the first application I am trying to install and there are no process heavy working on disks:
# vmstat
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa
0 0 8 33164 52120 733600 0 0 4 5 2 4 1 0 98 1

my virtual disks are scsi lsi logic, 6G with space not pre-allocated; I also made a try with buslogic and ide with the same result. Also I installed win 2000 and I got the same behaviour.
Now I am getting last updates from suse.
Thank you, your hints are very appreciated.
Daniele

[Kai-Uwe Schurig]

Just for information:

I have upgraded my server with the latest patches from Suse (kernel 2.6.5-7.151-smp), rebuilded the vmware modules and had no problems with my Windows Server 2003 VM so far. I also installed Windows Server SP1 in my VM, again without trouble.

Regards,

Kai-Uwe

[Daniele Bernazzi]

Just overcomed! The problem was in the LVM layer. putting the guest machine files on a plain ext3 fs made it works smootly.
Thank you