Re: [Xen-users] Howto install domU CentOS 5 directly on partition

[Nico Kadel-Garcia <nkadel@xxxxxxxxx>]

Liam Kirsher wrote:
> Paul,
>
> I think someone else already furnished basically the same info, but here
> is more on what I did.  I just wanted identical CentOS5 virtual
> machines.  Below, I give them the name domN, where N is some integer.
>
> First, I created a template installation using virt-install with these
> variables
> virt-install --paravirt --nographics --name=domTemplate \
>         --location=nfs:192.168.1.102:/linux/CentOS-5.0-x86_64 \
>         --ram=512
>
> Others have pointed out that the virt-* programs need at least 500M of
> RAM to function, otherwise they hang.
> I didn't see how to use virt-manager to re-partition -- it kept giving
> me the LVM volumes, but maybe it's possible.
>   
You do it in your CentOS kickstart, or by hand-configuring disks in your 
CentOS installation as ou did.

> Anyway, during installation with virt-install I removed all the volumes
> and partitions and created a single new one that used the entire space
> (which consisted of a previously created logical volume of 2G, named
> domTemplate_data).
> Ignore the warning about not having swap. The template doesn't need
> swap, that gets created for each domU.
>   
Swap is not evil in such a simple setup. You can add it later by adding 
another DomU "disk" and partition.

> I unchecked all the packages to be installed, because I wanted a small,
> generic installation to start with.  It came out to be around 1.2G, if I
> recall. You could make it slimmer, but that was good enough for my purposes.
>   
Yeah, this is RedHat's fault. I've growsed at them about it. I'm sorry, 
but a "base" installation should not include bluetooth and ISDN widgets.
> After the domTemplate is created, you might want to check to make sure
> it works before going any farther:
> xm create -c domTemplate
>
> I borrowed from the following how-to:
> http://www.redhat.com/archives/fedora-xen/2006-November/msg00257.html
>
> - Create lvm partitions:
> lvcreate -L 1G -n domN_swap VolGroup01
> lvcreate -L 3G -n domN_data VolGroup01
>
> - Map the domTemplate_data partition:
> kpartx -av /dev/VolGroup01/domTemplate_data
>
> - Use dd to copy to the new partition:
> dd if=/dev/mapper/VolGroup01-domTemplate_data of=/dev/VolGroup01/domN_data
>
> - Create swap space:
> mkswap /dev/VolGroup01/domN_swap
>
> - Unmap the domTemplate_data partition:
> kpartx -dv /dev/VolGroup01/domTemplate_data
>
> - Mount the new data partition:
> mount /dev/VolGroup01/domN_data /mnt/domN
>
> - In new host, edit the following files and change 'template' to new name.
> - I had named the host "template" to remind me which one I was working with.
> vi /mnt/domN/etc/hosts
> vi /mnt/domN/etc/sysconfig/network
>
> - Also on new host, set the correct IP address and MAC address:
> vi /mnt/domN/etc/sysconfig/network-scripts/ifcfg-eth0
>
> - Unmount the new data partition:
> umount /mnt/domN
>
> - Copy the 'image' xen configuration file and edit it:
> cp /etc/xen/domTemplate /etc/xen/domN
> vi /etc/xen/domN
>         set MAC address to unique!
>         set memory
> (I forgot to set the MAC address once, and ended up with perplexing
> networking problems.  It should be unique, and should match the MAC
> value set in ifcfg-eth0.  I removed the unique id, I think that should
> be okay, not sure why I would need it.
> You may also want to put a link in the /etc/xen/auto directory, so the
> domain is created when the system boots up.)
>
> RESIZE FILE SYSTEM if domN logical volume is larger than the template! 
> Make sure it's not running, first.
> e2fsck -f /dev/VolGroup01/domN_data
> resize2fs  /dev/VolGroup01/domN_data
>   
No need for this. You can store a tarball of the file system, not the 
file system itself, which is inevitably more space efficient, and 
restore the tarball to a new filesystem in similar fashion. I've done 
this before, to about 13,000 servers in one month, so I'm sure the 
technique works well. It's also one heck of a lot faster than using 
Anaconda and most other OS installers.
Getting "grub" loaded for use by pygrub in my approach of setup is.... a 
bit of an art form to discuss another time.
Your write-up is overall quite good, and appropriate for a FAQ.

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users