Xen on CentOS 5 - Notes

From PDP/Grid Wiki
Jump to navigationJump to search

Docu on CentOS

Some links:

CentOS-5 Documentation including docs on Virtualization

check PCI-DMA: Out of SW-IOMMU space ERROR

CentOS 5

These are just notes taken during the installation, it is not meant as a howto or something like that

Hardware Dell 1950

Additional Stuff

mercurial is a distributed version control system. to use it on centos-5 X86_64 do:

compile libvirt

  • wget ftp://ftp.easynet.be/gnu/gettext/gettext-0.17.tar.gz; tar xvfz gettext-0.17.tar.gz; cd gettext-0.17; ./configure; make; make install
  • yum install gnutls gnutls-devel
  • getting latest libvirt see: libvirt cvs
  • cvs -d :pserver:anoncvs@libvirt.org:2401/data/cvs login; password anoncvs cvs -d :pserver:anoncvs@libvirt.org:2401/data/cvs co libvirt
  • cd to libvirt dir and apply patch patch -p1 < configure.patch
  • in libvirt dir do: ./autogen.sh; make; make install

virt-install needs:

export PKG_CONFIG_PATH=/usr/local/lib/pkgconfig

CentOS 5 X86_64, Minimal system

Additional Packages

Minimal System installed via kickstart, packages:

# Packages groups/list
%packages --resolvedeps --ignoremissing


Post Install, get the following packages

Some of these are not needed, has to be cleaned:

yum install \
gcc make zlib-devel zlib python-devel curses libncurses-devel ncurses-devel openssl openssl-devel \
`yum search xorg-x11|grep dev|grep x86_64|awk -F . '{print $1}'` \
bridge-utils tetex tetex-latex transfig libtool-ltdl dev86 glibc-devel

xen 3.1.2-rc from mercurial

For 64-bit Dom_0 and 32-bit Dom_U kombination:, the following is taken from: xensource install notes

Dell 2900 modules to include in mkinitrd

mkinitrd -v -f --with=sd_mod --with=scsi_mod --with=megaraid_mm --with=megaraid_sas --with=megaraid_mbox initrd-

dom0 64 bit and domU 32 bit from source code

Build and install xen:

make xen
make install-xen 

Build and install the dom0:

make linux-2.6-xen0-prep
cd build-linux-*-xen0-x86_64
make menuconfig
make install modules_install

And finally build and install 32 bit domU: 

make XEN_TARGET_ARCH=x86_32 linux-2.6-xenU-prep
cd build-linux-*-xen0-x86_32
make ARCH=i386 menuconfig
make ARCH=i386
make ARCH=i386 install modules_install

hg pull ...
cd xen-3.1
wget http://www.broadcom.com/docs/driver_download/NXII/linux-1.5.10c.zip
unzip, use broadcom sources for 
make world
make install
depmod 2.6.18-xen
mkinitrd -v -f --with=aacraid --with=sd_mod --with=scsi_mod  --with=megaraid_sas /boot/initrd-2.6.18-xen.img 2.6.18-xen

Note on: Network Problems Broadcom drivers have problems, network is OK when xend started without network bridges, see /etc/xen/xend-config.sxp

Starting bridge with: /etc/xen/scripts/network-bridge start results in a not working network configuration in dom0.

see network-bridge causes network to hang

swiotlb=128 has to be set in menu.list since we have 8GB RAM, without youll get error messages like :

Sep  7 13:14:22 test-11 kernel: PCI-DMA: Out of SW-IOMMU space for 65536 bytes at device 0000:02:01.0

see also: SW-IOMMU space

  • change/add /boot/grub/menu.list:
 title CentOS (xen-3.1.2-rc2 2.6.18 compiled)
       root (hd0,0)
       kernel /xen-3.1.gz
       module /vmlinuz-2.6-xen ro root=/dev/md1 swiotlb=128
       module /initrd-2.6.18-xen.img

Boot into your new system:

[root@kribbe ~]# xm dmesg

__  __            _____  _   ____              ____
\ \/ /___ _ __   |___ / / | |___ \    _ __ ___|___ \
 \  // _ \ '_ \    |_ \ | |   __) |__| '__/ __| __) |
 /  \  __/ | | |  ___) || |_ / __/|__| | | (__ / __/
/_/\_\___|_| |_| |____(_)_(_)_____|  |_|  \___|_____|
University of Cambridge Computer Laboratory
Xen version 3.1.2-rc2 (root@nikhef.nl) (gcc version 4.1.1 20070105 (Red Hat 4.1.1-52)) Tue Nov 13 15:02:14 CET 2007
Latest ChangeSet: unavailable
(XEN) HVM: VMX enabled
(XEN) VMX: MSR intercept bitmap enabled
(XEN) CPU0: Intel(R) Xeon(R) CPU            5150  @ 2.66GHz stepping 06


Due to performance reasons VMs should be installed in LVMs. For recovering a LVM in a Raid-1 see:Recovery of RAID and LVM2 Volumes

Installed Version:

Installed Packages
Name   : lvm2
Arch   : i386
Version: 2.02.16
Release: 3.el5
Size   : 5.5 M
Repo   : installed
Summary: Userland logical volume management tools

NOTE: for having the live-migration-posibility we have to look at Cluster-LVM or something like that.

To get started ....

Setting up LVM2 on kribbe

Using /dev/srv resp. /dev/md3 for LVM we have to:

  • comment out /dev/md3 in /etc/fstab
  • umount /dev/srv (only once)
  • Creating the physical volume and XenVG Volumegroup:
[root@kribbe ~]# pvcreate /dev/md3
 Physical volume "/dev/md3" successfully created
[root@kribbe ~]# vgcreate XenVG /dev/md3
 Volume group "XenVG" successfully created

Within this volume-group we can now manage the partitions for the virtual maschines using lvcreate, lvresize, .... commands

Example: lvm for a CentOS5 test VM

Root FS, only one partition

[root@kribbe tmp]# lvcreate -L10G -nXen_CentOS5_Test XenVG
 Logical volume "Xen_CentOS5_Test" created
root@kribbe tmp]# mkfs.ext3 /dev/XenVG/Xen_CentOS5_Test
mke2fs 1.39 (29-May-2006)
Filesystem label=
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
1310720 inodes, 2621440 blocks
131072 blocks (5.00%) reserved for the super user
First data block=0
Maximum filesystem blocks=2684354560
80 block groups
32768 blocks per group, 32768 fragments per group
16384 inodes per group
Superblock backups stored on blocks:
       32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632

Writing inode tables: done
Creating journal (32768 blocks): done
Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 28 mounts or
180 days, whichever comes first.  Use tune2fs -c or -i to override.


Note: swap in a lvm is not performant

[root@kribbe tmp]# lvcreate -L1G -nXen_CentOS5_Test_swap XenVG
 Logical volume "Xen_CentOS5_Test_swap" created
[root@kribbe tmp]# mkswap /dev/XenVG/Xen_CentOS5_Test_swap
Setting up swapspace version 1, size = 1073737 kB

Usage of LVM2 on Centos xvda disks

Taken from xen-users archive.

If you are using a lvm2 partition as a disk (xvda) in a dom_u and this lvm is itself partitioned you can't use just

mount -o loop <logical volume> <mountpoint>, instead you have to give the first sector of the partition you want to mount. This can be calculated by multiplying the start sector with its size


[root@hilde xen]# fdisk -l -u /dev/mapper/xenvg-hooiwagenc4532

Disk /dev/mapper/xenvg-hooiwagenc4532: 53.6 GB, 53687091200 bytes
255 heads, 63 sectors/track, 6527 cylinders, total 104857600 sectors
Units = sectors of 1 * 512 = 512 bytes

                           Device Boot      Start         End      Blocks   Id  System
/dev/mapper/xenvg-hooiwagenc4532p1   *          63      208844      104391   83  Linux
/dev/mapper/xenvg-hooiwagenc4532p2          208845     2249099     1020127+  82  Linux swap / Solaris
/dev/mapper/xenvg-hooiwagenc4532p3         2249100   104856254    51303577+  83  Linux

[root@hilde xen]# mount -o loop,offset=32256 /dev/mapper/xenvg-hooiwagenc4532 /mnt2

[root@hilde xen]# ls /mnt2
bin   dev  home    lib         media  opt   root  selinux  sys  usr
boot  etc  initrd  lost+found  mnt    proc  sbin  srv      tmp  var

Virtualize Physical Nodes

  1. On DOM_0 Create a lvm for the DOM_U
  2. Mount this lvm on DOM_0
  3. On Physical Node change runlevel to s
  4. setup a exclude-list file, for details see eisxen
  5. cd /;tar csp --exclude-from=/rootfs/exclude.lst . |ssh <DOM_0> "cd /<mounted LVM>; tar xsp"
  6. change the fstab of the dom_u
Example exclude.lst, node specific, dont just copy, see also: eisxen
This list must not have an empty line at the end.
less /rootfs/exclude.lst

DOM_U is a centos-3.x

In order to run this as a virtual maschine, the dom_u might need a new version of e2fschk and module-init-tools. Therfore, on DOM_0 mount the DOM_U fs created in an earlier step, get e2fsprogs-1.39.tar.gz module-init-tools-3.2.tar.gz (these versions worked for us) to this directory. The following commands worked for me.

  • chroot to dom_u mountpoint
  • extract, compile, install e2fschk and module-init-tools
  • cp -r DOM_0:/lib/modules/<running version> DOM_U:/lib/modules
  • for e2fs do: tar xvfz e2fsprogs-1.39.tar.gz ;cd e2fsprogs-1.39; ./configure; make; make install
  • for modutils: tar xvfz module-init-tools-3.2.tar.gz; cd module-init-tools-3.2; ./configure --prefix=/ ;make moveold; make; make install

CentOS 5 with Virtualization package

  • Installed Version:
2.6.18-8.el5xen #1 SMP Thu Mar 15 19:56:43 EDT 2007 x86_64 x86_64 x86_64 GNU/Linux
  • Run yum update, Version:
2.6.18-8.1.15.el5xen #1 SMP Mon Oct 22 09:01:12 EDT 2007 x86_64 x86_64 x86_64 GNU/Linux
  • xen:
[root@kribbe ~]# xm dmesg
__  __            _____  ___   _____             ____       _ ____
\ \/ /___ _ __   |___ / / _ \ |___ /    _ __ ___| ___|  ___| | ___|
 \  // _ \ '_ \    |_ \| | | |  |_ \ __| '__/ __|___ \ / _ \ |___ \
 /  \  __/ | | |  ___) | |_| | ___) |__| | | (__ ___) |  __/ |___) |
/_/\_\___|_| |_| |____(_)___(_)____/   |_|  \___|____(_)___|_|____/

University of Cambridge Computer Laboratory

Xen version 3.0.3-rc5-8.1.15.el5 (mockbuild@) (gcc version 4.1.1 20070105 (Red Hat 4.1.1-52)) Mon Oct 22 08:25:42 EDT 2007
Latest ChangeSet: unavailable
(XEN) VMXON is done
(XEN) Dom0 has maximum 4 VCPUs
  • lvm2, bridge-utils
Installed Packages:
Name   : lvm2
Arch   : x86_64
Version: 2.02.16

Name   : bridge-utils
Arch   : x86_64
Version: 1.1

  • yum install vnc
  • use partition /dev/md3 mounted on /project for lvm
comment out ..... /project
umount /project
pvcreate /dev/md3
vgcreate XenVG /dev/md3
lvcreate -L10G -nXen_CentOS5_Test XenVG
mkfs.ext3 /dev/XenVG/Xen_CentOS5_Test
lvcreate -L1G -nXen_CentOS5_Test_swap XenVG
mkswap /dev/XenVG/Xen_CentOS5_Test_swap

Running various centOS Versions as VMs on kribbe

Virt-manager for installation

virt-install can be used for VM installation. No pypxeboot and patched udhcp needed.

CentOS 5 64-Bit in VM

set up kickstart file for VM, than

virt-install -p --location=http://www.dutchgrid.nl/mirror/centos/5/os/x86_64 --noautoconsole \
--file=/dev/XenVG/Xen_CentOS5_Test --name=domU001 --ram=1024  --mac=00:16:3e:ab:61:0b\
-x "ks=http://stal.nikhef.nl/ks/xen-test-koebrug.ks ip= \
netmask= dns= gateway="

MAC-ADDRESS: can be calculated from last the 3 numbers of you guest IP by (ex:):

printf 00:16:3e:%02x:%02x:%02x"\n" 171 97 11


This has to be defined in the dhcp servers dhcp.conf.

Connect to the console of the starting vm:

  • xm list, lists the VMs:
[root@kribbe ~]# xm list
Name                                      ID Mem(MiB) VCPUs State   Time(s)
Domain-0                                   0     7025     4 r-----   1056.7
domU001                                    8     1023     1 -b----    490.8
  • xm console <ID> or <Name> of the VM (s.a.) creates a console of the VM


Old Note: no vnc option in the virt-install command,

  • --vnc option removed, does not boot.
  • during boot of the VM answer: q: graphics support: no

But this one worked fine with a VNC console (both graphical and text based):

virt-install  --force -r 512 -m 00:16:3e:34:4b:6e -n mestkar -f /dev/vg0/mestkar.sys -l http://spiegel.nikhef.nl/mirror/centos/5/os/x86_64 -x "ks=http://www.nikhef.nl/grid/ndpf/files/local/install/mestkar.nikhef.nl.ks ksdevice=eth0" --vnc --vcpus=1 -b xenbr0

On VM:

  • koebrug.nikhef.nl
CentOS release 5 (Final)
[root@koebrug ~]# uname -a
Linux koebrug.nikhef.nl 2.6.18-8.el5xen #1 SMP Thu Mar 15 19:56:43 EDT 2007 x86_64 x86_64 x86_64 GNU/Linux

CentOS 4.5 64-Bit in VM

  • silo.nikhef.nl

[root@silo ~]# uname -a Linux silo.nikhef.nl 2.6.9-55.ELxenU #1 SMP Wed May 2 14:29:26 EDT 2007 x86_64 x86_64 x86_64 GNU/Linux [root@silo ~]# cat /etc/redhat-release CentOS release 4.5 (Final)

64 Bit DOM_U in 64 Bit DOM_0 seems to be no problem. 32Bit installations can not be done in this way, apropriate kernel needed Looking at virsh


Information on virsh can be found at: centos-wiki or Red Hat Virtualization

virsh is another virtualization management tool (built around the libvirt management API) which is now capable to deal with xen, future versions might also be able to manage kvm or qemu.

Since we plan to use xen we should stick with xm directly.

Virtual Maschine management with xm

based on: xen-user-docu

CentOS 5 i386-System -- Quattor generated


  • xenconsole: Could not read tty from store: No such file or directory
root@hilde xen]# xm create koebrug -c
Using config file "./koebrug".
Started domain koebrug
xenconsole: Could not read tty from store: No such file or directory

Try: /etc/init.d/xend restart

  • Post-Quattor steps to prevent quattor from revoking installation/configuration operations.
on kribbe:
chkconfig cdp-listend off
chkconfig ncm-cdispd off

on stal:
[root@stal ~]# aii-shellfe --boot kribbe.nikhef.nl

to give back the control to quattor (fresh install next boot, will also switch on cdp-listen and ncm-cdispd)
[root@stal ~]#  aii-shellfe --install kribbe.nikhef.nl
  • Mount external storage for files that should survive a reboot/installation
mount -tnfs tbn15.nikhef.nl:/export/array2/local/tbadmin /tmpmnt
  • Hostname: kribbe.nikhef.nl
  • Kernel: 2.6.18-8.1.8.el5xen
  • HD-Layout:
[root@kribbe ~]# sfdisk -l /dev/sdb

Disk /dev/sdb: 60801 cylinders, 255 heads, 63 sectors/track
Units = cylinders of 8225280 bytes, blocks of 1024 bytes, counting from 0

  Device Boot Start     End   #cyls    #blocks   Id  System
/dev/sdb1   *      0+     32      33-    265041   fd  Linux raid autodetect
/dev/sdb2         33   51025   50993  409601272+  fd  Linux raid autodetect
/dev/sdb3      51026   58580    7555   60685537+  fd  Linux raid autodetect
/dev/sdb4      58581   60800    2220   17832150    5  Extended
/dev/sdb5      58581+  60147    1567-  12586896   fd  Linux raid autodetect
/dev/sdb6      60148+  60669     522-   4192933+  fd  Linux raid autodetect
/dev/sdb7      60670+  60800     131-   1052226   fd  Linux raid autodetect


[root@kribbe ~]# cat /proc/mdstat 
Personalities : [raid1] 
md0 : active raid1 sdb1[1] sda1[0]
     264960 blocks [2/2] [UU]
md3 : active raid1 sdb2[1] sda2[0]
     409601152 blocks [2/2] [UU]
md5 : active raid1 sdb3[1] sda3[0]
     60685440 blocks [2/2] [UU]
md2 : active raid1 sdb6[1] sda6[0]
     4192832 blocks [2/2] [UU]
md4 : active raid1 sdb7[1] sda7[0]
     1052160 blocks [2/2] [UU]
md1 : active raid1 sdb5[1] sda5[0]
     12586816 blocks [2/2] [UU]
unused devices: <none>

Example installation of a CentOS5 VM

Create a VM in lvm partition

  1. Setup lvm as stated above
  2. Mount VM root fs: mount /dev/XenVG/Xen_CentOS5_Test /mnt
  3. extract archive:
  4. [root@kribbe tmp]# bunzip2 centos.5-0.20070424.img.tar.bz2
  5. [root@kribbe tmp]# tar xvf centos.5-0.20070424.img.tar
  6. loopmount image
  7. [root@kribbe tmp]# mkdir /img
  8. [root@kribbe tmp]# mount -o loop centos.5-0.img /img
  9. cp image to logical volume
  10. [root@kribbe img]# cp -a * /mnt/
  11. Change Root PW in VM:
[root@kribbe mnt]# chroot /mnt
bash-3.1# pwconv
bash-3.1# passwd
Changing password for user root.
  1. Set Hostname of VM: /etc/sysconfig/network
  1. no swap in VM: /etc/fstab
#/dev/sda2               none                    swap    sw       0 0

setup of VM start files

SELinux, Postinstallation steps

  • Cleaning up daemons, pay attention to ntp, does the dom_U has the correct time?
  /sbin/service microcode_ctl stop
  /sbin/chkconfig --del microcode_ctl
  /sbin/service smartd stop
  /sbin/chkconfig --del smartd

If you are using SELinux, it is important to check that the image has the correct security context (xen_image_t), or access to the virtual disk will be denied to the domU system. You can check this with ls:

# ls -Z /srv/xen/mailserver.img 
-rw-r--r--  root root user_u:object_r:xen_image_t      /srv/xen/mailserver.img

If you are having trouble setting the right file context, please have a look at the Xen tips and tricks page. At any rate, turning off SELinux, as some howtos on this subject advise is a very poor workaround. Reading two manual pages (semanage(8) and restorecon(8)), for an extra layer of security is a good trade!

Memory and network problems

The CentOS 5 (RHEL5) release of Xen is not really stable; it changed it's behaviour over time from 'not working at all' in CentOS 5.0 and 5.1 to 'working' in '5.2', and with further updates it appears as though rather important changes find their way into the Xen part of this otherwise stable distro.

Recently, and this is 5.3 on April 22, 2009, I started to see spontaneous reboots of my Dom0. Not much in the syslog, but some suspicious entries about memory squeeze:

Apr 22 10:29:09 toom kernel: xen_net: Memory squeeze in netback driver.
Apr 22 10:29:14 toom kernel: printk: 8 messages suppressed.

This relates to a known bug; a suggested solution is to limit the Dom0 memory to prevent 'ballooning'. This is what I did:

in /etc/grub.conf:

kernel /xen.gz-2.6.18-128.1.6.el5 com2=9600,8n1 console=com2,vga dom0_mem=1024M

and in /etc/xen/xend-config.sxp:

(dom0-min-mem 1024)

Making 1024 (megabyte) both the minimum and maximum of the dom0 memory. This could affect performance, but I haven't seen the squeeze message since.