Mailing List Archive

Basic questions - numbered
I know you guys are all very busy so I have numbered my questions. It's
too much to expect one person to answer all these questions and
comments. These questions are not directly related but they are my
difficulties and comments associated with Xen and the LIVE CD. With SuSE
planning to include Xen in a forthcoming distribution some of my
difficulties and comments may be worth considering or even altering and
included in a Wikki or FAQ section.

1. Beta2 Live CD-ROM -- boot issues

The greatest thing since sliced bread.

After over a month I got access to a high speed Internet connection and
downloaded LIVE CD Beta2.
It appears to work so my problem is local to my LAN machines. Perhaps
this problem should not be sent to the list. I do so only because of
the possibility that the LIVE CD (Beta2) won't boot.and there might be
a boot directive that will handle this problem. The problem is probably
caused by a combination of CD drive and Bios and I don't expect any help
with a hardware/firmware problem -- perhaps there is something I am
overlooking.

I tried it (the CD) on 3 relatively modern machines running XP and XP
Pro although that does not matter as it's a Live CD. The Live CD ran
okay as far as I used it (both in text and windows/graphic mode -- nice
job). Then I tried it on two Linux machines (again that does not matter
as it's a Live CD). But the Live CD would not boot. Ouch!!! After
several days of testing I discovered (even though the bios was properly
set so the CD should boot) that even an OS independent Smart Boot could
not force the CD to boot nor even detect it. Even a SuSE CD self booting
installation would not boot.

However, using a SuSE installation floppy disk (self booting) I could
launch a CD-ROM installation and the DVD and CD drives were operational.
Logs on the existing Linux installation show they were correctly
detected. In other words, this proved that the issue was confined to
"booting" associated with CD and DVD drives and the drives work okay
otherwise. On my machine these drives (DVD and CD) are controlled from
a Promise ultra 66 IDE extension card giving me a total of 8 drive slots
on the Linux machine. Although they (DVD and CD drives) are connected on
IDE slots as sr0 and sr1 (on the installed SuSE 6.4 system), they are
classified as scsi and given channel, id, and lun as well as drive
designations hdi and hdj. I think this is the reason why your Live CD
will not boot on my Linux system. It's the actual physical CD drive when
called by the Bios that is not detected or refuses to boot. This problem
is bound to reoccur many times once other's of my ilk and machinery get
involved with Xen and products like the Live CD. I'm only guessing, but
I think the only answer is a floppy image that will boot specifically
your Live CD (or behave in the manner of the SuSE installation floppy)
having previously been given some kind of drive information in order to
detect it or a modification to your Live CD boot process aimed at this
kind of problem.

I don't know how the Live CD will ever boot if the CD drive somehow
impairs that action -- if that is what is happening. If the SuSE
installation boot floppy can detect and use the CD but the CD drive
won't boot then we have to somehow chain boot from floppy to Live CD.
Since I have this identical problem on 2 Linux machines (one with the
Promise card and one without) I think perhaps the problem is in the boot
directive when the bios turns execution control over to the CD drive.
Both Linux machines seem to have the same Bios. If the Bios is the
problem and not the CD or DVD drive then the problem may not be fixable
and I will be unable to use either of these Linux machines. The XP
machines have a different Bios. All machines have different makes of CD.
Should I be replacing or relocating the CD drive??? Does it sound like a
Bios problem??? I don't expect help on a hardware/firmware problem but
can you give me any help (in the form of a boot directive) for this
problem??? I will provide any details you request.

2. Machine application processing architecture.

Let's say (on the Linux machine with 8 hard drive slots where 2 are for
the DVD and CD) I have taken one of these 6 hard drives, created a bunch
of logical partitions with the target usage being Xen usage. I am
assuming here that I can place (in each of these logical partitions) a
unique and unaltered Linux OS (like Mandrake, Gentoo, etc., as long as
they have a kernel versions of GNU/Linux 2.6.x ) which can then be
"virtualized" (as domains, 1, 2, 3, etc) by the Xen-based system using
the xm commands.

The reason I bring this up is that on this Linux machine I have SuSE 6.4
(kernel 2.2.17 -- not suitable for Xen) and I want to keep this
installation on the machine. The idea is to boot on the existing Linux
or boot on a Xen-based installation (perhaps several). I thought I
would try putting SuSE 9.1 (kernels 2.6.x) in the first of these logical
partitions on a dedicated hard drive. This would be the base OS to
become an Xen system. So, to get started, I formated (ext3) one of the 6
hard drives (that later I would partition into many logical ext3
partitions). But first I would simply try to install SuSE 9.1 on the
entire hard drive as a test of the installation process. Good thing I
tried this. To my surprise, the SuSE installation software package did
not like the idea of residing on the same machine as an existing SuSE
installation (it assumes one instance of the same SuSE installation --
not multiple instances) and all forced efforts on my part to do so
appeared to be leading to the destruction of the original installation
(even after going into the advanced manual steps of defining the
partition for the installation but the software wanted to treat it as a
change to a data partition). I want to use the 9.1 (or 9.2 or 9.3) SuSE
distribution to form my Domain 0 in the first of these partitions that I
want to create.

The Xen maual indicates we should simply install additional OS systems
from the disk(s) that make up the distribution. But if this problem
described above can happen -- as it did with me -- one cannot assume
that every effort to install another OS will be straight forward and it
might be advantageous to come up with a different manner of doing this.
SuSE support in any form (unless I buy the expensive Enterprise Server
support package) does not extend to this kind of installation situation.
Who needs problematic installations anyway. So I could do the
installations on another machine and "dd" that installation into the
target partition and with a way of booting up the Xen-based (SuSE).
But to create the Xen-based (SuSE) Dom 0 I first need to have the SuSE
installation bootable. I am assuming here that not only can I use these
logical partitions in this manner but I can move various Linux OS's into
the empty logical partitions in the same way (using the Linux "dd"
command). If there were hardware differences (hard drives, NIC, etc.) I
would have to change the appropriate config files. Although on the SuSE
list there was quite a discussion regarding how easy it was to move a
distribution from one machine to another or change the CPU board and
that sort of thing. Alternatively and probably desireable (this machine
has pluggable hard drives) and I could by moving hard drives around and
transfering the partition image by means of dd eliminate the
discrepencies introduced by different machines and avoid any
configuration discrepencies. This would necessitate a temporary hard
drive configured with the same partitions and plugged into the same hard
drive slot.

Are there important issues I have missed in considering the doing of
this???

If this seems feasible (moving a pre built Linux installation into a
logical partition for this purpose) could I have an example of a GRUB
configuration file as a kind of template for a dedicated multiple
logical partitions boot???

Is the "grub.conf" file associated with the LIVE CD a good example of
what I am looking for and how do I discover it's contents??? I cannot
easily check this out as the LIVE CD only runs on my XP production
machines (as mentioned above) and I cannot bring them down right now.

3. Live CD as hard drive installed example.

Can I do the following?

"dd" the Live CD into one of these partitions mentioned above and boot
to this "copy" on hard drive instead of CD???

If so does the partition need to be formatted (I assume the iso image
copied contains all the bits associated with both file system
construction as well as the installation existing within the file
system) so when booted should run identical to the LIVE CD without any
file system formatting (ext2, ext3, reiser, etc.)???

What would the GRUB boot file for this purpose look like -- could I
chainloader (hdx,y) + 1 to it -- or is there a better way???

If it works, I think the above "hard drive as CD" should behave as a CD
but much faster???

Is there a way of transforming the CD into an actual Xen-based hard
drive application? If it works, the above is simply an iso file
executed as would be a CD.

4. LIVE CD modifications:

The Live CD comes with its predefined IP addressing. I would like to
alter this so that the Dom0 and DomU virtualizations can access other
machines on my LAN. I have read both the development and user manuals
as well as list chatter and FAQ and there is a lot of material on this
subject but I'm still missing something.

Is it possible for someone to provide command lines relative to the
Live CD for the following steps specific to what I am trying to do as
the (well written) documentation is just a bit too vague for my level of
insight???

First, I need to revise the IP addresses, routing and IP configuration
for some of the Dom0 while leaving the Xen-based internal network
(virtualizations) stuff as it is. I need to both change and add stuff
and am somewhat confused regarding the specifics. From the literature I
know I must provide a MAC bridge to join the internal LIVE CD internal
network to the LAN network. That's how I read it. But this is where I am
most confused because the Xen-based system has it's own bridge
(assumedly to virtualize network connectivity) and I would think that
each DomU or virtualization including Dom0 would also need a bridge to
talk to my LAN machines or I've got this wrong and the bridges just join
networks. (Sorry, but I am missing a bit of network insight here). It
would have been nice if the Live CD could be extended to refer to
another network (LAN even if not existing) using 192.168.x.x static
and/or dynamically assigned. All my LAN stuff is hand crafted STATIC IP
addressing and DHCP is not allowed under 192.168.1.100 on the
192.168.1.0 LAN network. My gateways are 192.168.1.70 and 192.168.1.6.
The LAN has both Windoozzzz and Linux (yea,yea,yea) machines.

Second, I need to define the gateways for Dom0 as well as the DomU
instances.
FOR EACH_DOM DO
confused# route add default gw 192.168.1.70 dev eth0
confused# route add default gw 192.168.1.6 dev eth0
I'm not sure where Debian has this in a configuration file -- maybe
/etc/route -- Debian is all new to me.
ENDDO

Third, I need to change the /etc/host and other network related files so
that these domains recognize the LAN machines. In this regard -- are
SAMBA servers and clients as well as NFS servers and clients present on
the Live CD Dom0 and DomU entities. If not, and assuming I can "plant"
the Live CD onto hard drive as covered above, can I extend these
entities with additional packages???

Can these kind of changes (adding packages) be done during
virtualization or would virtualization have to stop until the physical
update installation is complete? I think the latter but I'm asking. In
this respect is DOM 0 updated differently than a DOMU. Based on my
short experience with the LIVE CD Xen seems to be always running and the
DOM 0 Guest (Debian in the case of the LIVE CD) can be started and
stopped. I had previously thought that DOM 0 Guest might be an
exception -- always virtualized?

It took me a while to understand the following -- I hope it's right???
If it's wrong even in the slightest will someone please correct it.

From a high level perspective, "Virtualization by Xen" is the
controlled execution of kernels associated with physical installations
of i86 based OSs. But it is more. Xen is integrated with the modified
kernel of a Linux i86 based OS installation -- the base installation
(and therefore the physical resources hard drive, NIC, etc. of that base
installation) which combination is called DOM 0. Xen is responsible for
the sharing of these physical resources (associated with DOM 0) by means
of time sliced (relatively concurrent) execution control of other
installed Operating Systems called DOMU for user level (because they do
not have certain privileges and permissions and when needed Xen manages
that need). Xen provides interfaces and processes to accomplish this
controlled execution of DOMU types of domains or even other Xen-based
DOM 0 domains where a domain is a primary or logical partition having a
file sytem (formatted as ext2, ext3 or other) and a OS (distribution)
as physical content. There is for every virtualizeable domain
possibility a uniqe physical OS installation. Virtualization as defined
here makes it possible to efficiently run many Operating Systems
(concurrently or more correctly speaking -- in parallel) on the same
machine. When applications in a domain (eg, domain 3) execute in
conjunction with their associated kernel (kernel for domain 3) the Xen
package (interfaces domain 0 Guest hardware and modifed kernel) -- (with
domain 3 kernel) -- in order to control both the time duration of that
execution (of domain 3 kernel) as well as the access (shared) of
domain 0 Guest physical block device resources (indirectly by domain 3
kernel). Xen lives in memory and operates from memory as do the kernels
of an OS installation being virutalized. So called priviledged kernel
calls from DOMU are interfaced to DOM 0 and memory is used to contain
memory device blocks of various types and sizes to facilitate this
virtualization or controlled interfacing and associated executing
processes between the DOM 0 and DOMU domains. The exact implementation
details are in the code which is under constant change, debugging and
enhancement. For example, a debate is now underway on the list regarding
how best to design console or terminal usage associated with DOMU.

Thanks in advance for any help and comments -- Ted



_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
Re: Basic questions - numbered [ In reply to ]
Some snippets answered, hopefully others will fill in the rest.

> 2. Machine application processing architecture.

Once you have got a SuSE 9.x install onto the drive, there is a YaST tool for
"install into a directory" you can use to create other domain filesystems.

> The Xen maual indicates we should simply install additional OS systems
> from the disk(s) that make up the distribution.

Take a look at the appendix on using debbootstrap for an easy way to install
Debian.

> short experience with the LIVE CD Xen seems to be always running and the
> DOM 0 Guest (Debian in the case of the LIVE CD) can be started and
> stopped. I had previously thought that DOM 0 Guest might be an
> exception -- always virtualized?

Dom0 also runs on top of Xen, as a guest (albeit with extra privileges to
access hardware).

> It took me a while to understand the following -- I hope it's right???
> If it's wrong even in the slightest will someone please correct it.
>
> From a high level perspective, "Virtualization by Xen" is the
> controlled execution of kernels associated with physical installations
> of i86 based OSs. But it is more. Xen is integrated with the modified
> kernel of a Linux i86 based OS installation -- the base installation
> (and therefore the physical resources hard drive, NIC, etc. of that base
> installation) which combination is called DOM 0.

That's about right. Xen controls access to CPU, memory, IO ports and
interrupt lines. That's it, it doesn't know or care about devices.

By talking to Xen's interface, privileged domains can talk to the actual
hardware. The first domain to start on the machine (dom0) is allowed to
access all the hardware, using its own device drivers.

> Xen is responsible for
> the sharing of these physical resources (associated with DOM 0) by means
> of time sliced (relatively concurrent) execution control of other
> installed Operating Systems

Xen shares out the resources listed above. However, since dom0 controls the
actual devices, it is responsible for sharing out access to block, network,
etc. to other domains.

> called DOMU for user level (because they do
> not have certain privileges and permissions and when needed Xen manages
> that need).

Strictly I think the U is for "unprivileged" - domain kernels don't
technically run in "user mode". It doesn't really matter, the terminology
evolves over time ;-)

> Xen provides interfaces and processes to accomplish this
> controlled execution of DOMU types of domains or even other Xen-based
> DOM 0 domains where a domain is a primary or logical partition having a
> file sytem (formatted as ext2, ext3 or other) and a OS (distribution)
> as physical content.

Domains can also run out of files in dom0's filesystem, LVM volumes in dom0,
network block devices in dom0...

Anything dom0 can see as a block device can be used as a domU's root
filesystem.

HTH,
Mark

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
Re: Basic questions - numbered [ In reply to ]
Hi Ted,

I have a few brief answers to some of your points below, to follow up
from Mark:

> Then I tried it on two Linux machines (again that does not matter
> as it's a Live CD). But the Live CD would not boot. Ouch!!! After
> several days of testing I discovered (even though the bios was properly
> set so the CD should boot) that even an OS independent Smart Boot could
> not force the CD to boot nor even detect it. Even a SuSE CD self booting
> installation would not boot.

Yep, sounds like your machines don't boot from CD-ROM. Things that are
worth trying in this situation:

- SBM (Smart Boot Manager) floppy: http://btmgr.sourceforge.net/download.html
- GrUB floppy: http://www.gnu.org/software/grub/grub-legacy-faq.en.html#q4

A floppy boot image for the demo CD is unlikely to happen because the
initrd the CD uses is significantly larger than will fit on a floppy. :)

> If this seems feasible (moving a pre built Linux installation into a
> logical partition for this purpose) could I have an example of a GRUB
> configuration file as a kind of template for a dedicated multiple
> logical partitions boot???

I'm not sure what your question is, but the GrUB runes for booting Xen
are discussed in the Xen user's manual:

http://www.cl.cam.ac.uk/Research/SRG/netos/xen/readmes/user/user.html#SECTION02240000000000000000

> Is the "grub.conf" file associated with the LIVE CD a good example of
> what I am looking for and how do I discover it's contents???

The grub configuration on the CD is in /boot/grub/menu.lst on the CD; you
should be able to read that file from a running system.

> Can I do the following?
>
> "dd" the Live CD into one of these partitions mentioned above and boot
> to this "copy" on hard drive instead of CD???

In theory, yes, but why would you? The filesystem on the demo CD has a
bunch of tweaks to make it run from CD (i.e. without writeable disks,
and running several OSes from the same image). Starting from that would
give you something that's not quite a proper Debian install, and
something that's not quite the proper Xen tools on top of it, which
would just cause problems later on.

Basically, the demo CD is intended to be just that: a demo. It's *not*
meant as a production platform.

You would really be *much* better off installing a linux distro from
scratch and then installing Xen into it. Recent versions of Debian have
excellent support for installing in difficult cases, and it sounds like
your Suse install floppies work too.


> That's how I read it. But this is where I am
> most confused because the Xen-based system has it's own bridge
> (assumedly to virtualize network connectivity) and I would think that
> each DomU or virtualization including Dom0 would also need a bridge to
> talk to my LAN machines or I've got this wrong and the bridges just join
> networks. (Sorry, but I am missing a bit of network insight here).

The demo CD doesn't use the network card of the machine. Domain 0 talks
to all the other domains via the 'xen-br0' bridge, which has the IP
address 10.171.45.1. You should be able to bring up real LAN
interfaces separately with the usual linux tools. You can then setup
NAT between the inside (other domains via xen-br0) and the outside
(LAN). Read the man page for iptables for more details on how to do athat.

Tim.

--
Tim Deegan (My opinions, not the University's)
Systems Research Group
University of Cambridge Computer Laboratory

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
Re: Basic questions - numbered [ In reply to ]
Tim

Thankyou for your reply. Sorry that my question was confusing but
sometimes confused people ask confusing questions. I found this question
very difficult to word and would beg your indulgence if I am now too wordy.

To recap, you said:

I'm not sure what your question is, but the GrUB runes for booting Xen
are discussed in the Xen user's manual:

You referred me to:

title Xen 2.0 / XenLinux 2.6.9
kernel /boot/xen.gz dom0_mem=131072
module /boot/vmlinuz-2.6.9-xen0 root=/dev/sda4 ro console=tty0

Question part 1 -- clarification: I was actually looking for
comfirmation on using a group of logical partitions with each
containing an install of a Linux distribution such as SuSE, RH, DEB, etc.

Also, instead of partitioning one of the six hardrives as first
primary, second primary, third primary , and fourth primary partition
(with extended partition containing logical partions), I would only have
a first primary (extended partition) with umteen logical partions.
with each of these logical partitions containing an installed linux
distribution.

I realized that partioning this way is no problem but until recently I
had (mistakenly) always thought that linux had to be installed on a
primary partition on 0 (first) IDE 0 slot or the 1 (second) IDE 0 slot
or the typical traditional master locations. I did not realize that I
could create linux installations in logical partitions and refer and
boot them on that basis. I had always used logical partitions for
/boot, /usr, swap, and that sort of thing. In addition, I was under the
impression that an extended partition having logical partitions had to
be the last partition on a hard drive so if there were two primary
partitions the extended partition had to be last. So if there is only a
primary partition with an extended partition having many logical
partitions it is the last as well as the first.

Therefore, I had never attempted to create complete Linux installations
within a logical parition -- only parts of an installation (like /boot,
/usr, etc.) that collectively defined the installation.

So, theoretically speaking, if I were to use /dev/hde (the fith (5th)
drive -- in the sequence a, b, c, d, e, f, if there were no gaps) I
would have /hde1 and if on hde I allocated an extended partion with 64
logical partitions (numbered 1 to 64) I could install a bootable (using
Grub) linux distribution in each of these logical paritions.

BTW: I am trying to stay at a theoretical level but my question also
comes out of the situation with my current SuSE where I've actually got
on the 6.4 SuSE install -- b and d missing leaving
(hda(hda1,hda2,hda3), hdc(hdc1,hdc3,hdc4), hde, hdf, hdg, hdh, and a
tape driveSeagate STT8000A ATAPI tied to (fd0) floppy, hdi--DVD Samsung
ATAPI, and finally hdj--HP CD+r/w). This means I have a,c,e,f,g,h,i,j
accounting for all 8 IDE slots but apparently the letter designation, as
in hda, is not reserved but only follows the lettering sequence. So the
5th drive is desigated with "g". But I don't think this variability is
the case with Grub and the designations are reserved, as in (hd4, n)
which always refers to the 5th hard drive and a partition. The SuSE
Reference manual states that:
(hd0,0) is the first primary partition.
(hd0,1) is the second primary partition.
(hd0,2) is the third primary partition.
(hd0,3) is the fouth primary partition (usually an extended partition)
(hd0,4) is the first logical partition .
(hd0,5) is the second logical partition.

Thus my concern!!! The numbering conventions seem to be static or fixed
which would make my GRUB sequence in the below line WRONG? Just
thought you would like to know the basis of the question.

Continuing theoretically: If that drive contained just logical
partitions I would have on hde (hde1, hde2, hde3...hden) where n=1 to 64
-- I believe 0 refers to the extended partition itself but I am not sure
as it is not used just assumed.
And if this were translated into GRUB terminology I would have ((hd4,0),
(hd4,1), (hd4,2)...(hdn-1,n-1) where n =1 to 64

The above are my assumptions and I think I am WRONG in the above GRUB
line???

Now the actual Question.

Question 2nd part: are these sequences correct for both Linux and Grub
or are the first 3 (Linux 1,2, 3, and Grub 0,1,2 ) "reserved numbers"
making my sequences incorrect for the 5th hard drive (having only an
extended partition with umteen logical partitions each containing an
installed OS).

Thanks -- Ted.


Tim Deegan wrote:

>Hi Ted,
>
>I have a few brief answers to some of your points below, to follow up
>from Mark:
>
>
>
>> Then I tried it on two Linux machines (again that does not matter
>>as it's a Live CD). But the Live CD would not boot. Ouch!!! After
>>several days of testing I discovered (even though the bios was properly
>>set so the CD should boot) that even an OS independent Smart Boot could
>>not force the CD to boot nor even detect it. Even a SuSE CD self booting
>>installation would not boot.
>>
>>
>
>Yep, sounds like your machines don't boot from CD-ROM. Things that are
>worth trying in this situation:
>
>- SBM (Smart Boot Manager) floppy: http://btmgr.sourceforge.net/download.html
>- GrUB floppy: http://www.gnu.org/software/grub/grub-legacy-faq.en.html#q4
>
>A floppy boot image for the demo CD is unlikely to happen because the
>initrd the CD uses is significantly larger than will fit on a floppy. :)
>
>
>
>>If this seems feasible (moving a pre built Linux installation into a
>>logical partition for this purpose) could I have an example of a GRUB
>>configuration file as a kind of template for a dedicated multiple
>>logical partitions boot???
>>
>>
>
>I'm not sure what your question is, but the GrUB runes for booting Xen
>are discussed in the Xen user's manual:
>
>http://www.cl.cam.ac.uk/Research/SRG/netos/xen/readmes/user/user.html#SECTION02240000000000000000
>
>
>
>>Is the "grub.conf" file associated with the LIVE CD a good example of
>>what I am looking for and how do I discover it's contents???
>>
>>
>
>The grub configuration on the CD is in /boot/grub/menu.lst on the CD; you
>should be able to read that file from a running system.
>
>
>
>>Can I do the following?
>>
>>"dd" the Live CD into one of these partitions mentioned above and boot
>>to this "copy" on hard drive instead of CD???
>>
>>
>
>In theory, yes, but why would you? The filesystem on the demo CD has a
>bunch of tweaks to make it run from CD (i.e. without writeable disks,
>and running several OSes from the same image). Starting from that would
>give you something that's not quite a proper Debian install, and
>something that's not quite the proper Xen tools on top of it, which
>would just cause problems later on.
>
>Basically, the demo CD is intended to be just that: a demo. It's *not*
>meant as a production platform.
>
>You would really be *much* better off installing a linux distro from
>scratch and then installing Xen into it. Recent versions of Debian have
>excellent support for installing in difficult cases, and it sounds like
>your Suse install floppies work too.
>
>
>
>
>>That's how I read it. But this is where I am
>>most confused because the Xen-based system has it's own bridge
>>(assumedly to virtualize network connectivity) and I would think that
>>each DomU or virtualization including Dom0 would also need a bridge to
>>talk to my LAN machines or I've got this wrong and the bridges just join
>>networks. (Sorry, but I am missing a bit of network insight here).
>>
>>
>
>The demo CD doesn't use the network card of the machine. Domain 0 talks
>to all the other domains via the 'xen-br0' bridge, which has the IP
>address 10.171.45.1. You should be able to bring up real LAN
>interfaces separately with the usual linux tools. You can then setup
>NAT between the inside (other domains via xen-br0) and the outside
>(LAN). Read the man page for iptables for more details on how to do athat.
>
>Tim.
>
>
>
Re: Basic questions - numbered [ In reply to ]
Hi,

I did not ready the entire email, just flash thru it and believe
your questions can be answered as below. If I am wrong, please
just delete this email.

I am using Xen with SuSE 9.2 Professional. And installed it in
this way:

01) Install SuSE into the hostess machine, and, as I needed 9 VMs
and decided to go for LVM, I partitioned my SCSI 18GB disk into
3 partitions: /boot (128MB), / (2GB) and a LVM partition with
the rest.

02) The LVM partition were divided into a swap (384MB) for the
xen0 domain, plus 1 root and 1 swap partitions for each VM.
Each VM root partition is 800MB and each VM swap is 200MB.
Do not forget to format each VM root partition for reiserfs.

03) After finishing installing, I installed the compiler and
utilities, and all Xen dependencies into the hostess OS.
It is still SuSE 9.2 Pro default kernel.

04) Untar Untar/gunzipped xen-2.0.5 into /root/xen-2.0.5,
download the linux 2.6.10 kernel from kernel.org and copied
it to the same diretory.

05) make xen
make tools
make kernels
make install

06) Go to /boot/grub and changed menu.lst to use the xen kernel
to boot, the way you can see in the web page - just make
sure to use the correct command, see other entries in the
file menu.lst, especially the one for the default kernel.

07) Move /lib/tls to /lib/tls.disabled

08) Reboot, and choose Xen to boot.

09) After boot, create a directory /mnt/vmroot and mount the
VM root partition of VM 1 into it. As I named my LVM group
as v0, my device is /dev/v0/vm1root.

10) Now you can use YaST and choose "Install into directory".
It will prompt you for the directory, you write /mnt/vm1root,
then choose the "minimal installation" and "next". SuSE will
install itself into the VM1 root partition.

11) You change to the VM1 root partition, go to etc directory
and change everything you want :) Do not bother about GRUB,
just copy all modules for the xenU kernel to a modules
directory inside the VM1 root partition:

# (cd /lib/modules/; tar cf - 2.6.10-xenU)| \
(cd /mnt/vm1root/lib/modules; tar -xf -)

12) Unmount the /mnt/vm1root partition, create the /etc/xen/vm1
configuration file, like this:

kernel = "/boot/vmlinuz-2.6.10-xenU"
memory = 64
name = "fwInt"
nics=1
vif = [ 'mac=aa:00:00:00:00:11, bridge=xen-br0' ]
disk = [ 'phy:v0/vm1root,hda1,w', 'phy:v0/vm1swap,hda2,w' ]
root = "/dev/hda1 ro"
extra = "3"

13) Execute "xend start" in order to have the controller running,
then "xm create vm1" and you are in!

Hope this can help...

--
Bye,
Fernando Maciel Souto Maior
fernando@araujo.com.br
http://www.araujo.com.br
+55+31 3270-5886
LPIC/1 # 31908

AVISO-------------------------------------------------------------
Esta mensagem pode conter informacao confidencial ou privilegiada.
Se voce nao for o destinatario ou a pessoa autorizada a receber
esta mensagem, nao pode usar, copiar ou divulgar as informacoes
nela contidas ou tomar qualquer acao baseada nessas informacoes.
Se voce recebeu esta mensagem por engano, favor avisar o remetente
imediatamente, respondendo o e-mail e em seguida apagando-o.
Obrigado pela cooperacao.

DISCLAIMER--------------------------------------------------------
This message may contain confidential and/or privileged information.
If you are not the addressee or authorized to receive this for the
addressee, you must not use, copy, disclose or take any action based
on any information herein. If you have received this message in
error, please advise the sender immediately by replying to this
e-mail and delete this message. Thank you for your cooperation.

------------------------------------------------------------------
This email was sent using SquirrelMail - http://squirrelmail.org

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
Re: Basic questions - numbered [ In reply to ]
Hi Ted,

On Fri, Apr 01, 2005 at 12:26:13PM -0700, Ted Hilts wrote:
> So, theoretically speaking, if I were to use /dev/hde (the fith (5th)
> drive -- in the sequence a, b, c, d, e, f, if there were no gaps) I
> would have /hde1 and if on hde I allocated an extended partion with 64
> logical partitions (numbered 1 to 64) I could install a bootable (using
> Grub) linux distribution in each of these logical paritions.

Almost, but you can't use quite that many partitions (see below).

> But I don't think this variability is
> the case with Grub and the designations are reserved, as in (hd4, n)
> which always refers to the 5th hard drive and a partition.

Yep. Linux names drives based on how they are connected, e.g., /dev/sda
for the first SCSI drive, /dev/hdd for the second IDE drive on the
second channel, and so on.

Linux uses partition numbers 1 to 4 for primary partitions, and then
counts all the logical partitions from 5 upwards, to a maximum of 63 for
IDE disks (so you can only have 59 logical partitions) and of 15 for
SCSI disks (so you can only have 11 logical partitions).

http://lxr.linux.no/source/Documentation/devices.txt#L192
http://lxr.linux.no/source/Documentation/devices.txt#L272

GrUB allocates drive numbers in the order the BIOS presents them,
regardless of whether they're SCSI or IDE, so they may not necessarily
be in the same order that linux sees them in. I believe if you run the
grub command in a running linux it will try to figure out the mapping
for you.

GrUB uses partition numbers 0 to 3 for primary partitions, and then
counts all the logical partitions from 4 upwards.

http://www.gnu.org/software/grub/manual/html_node/Naming-convention.html#Naming%20convention

> Continuing theoretically: If that drive contained just logical
> partitions I would have on hde (hde1, hde2, hde3...hden) where n=1 to 64
> -- I believe 0 refers to the extended partition itself but I am not sure
> as it is not used just assumed.
> And if this were translated into GRUB terminology I would have ((hd4,0),
> (hd4,1), (hd4,2)...(hdn-1,n-1) where n =1 to 64

If you have 59 logical partitions on the 5th IDE drive (and that drive
is the master device on the third controller, and you don't have any
non-IDE drives...), GrUB should call the logical partitions (hd4,4) to
(hd4,62). Linux should call them /dev/hde5 to /dev/hde63.

Have you thought about using LVM? :)

HTH,

Tim.

--
Tim Deegan (My opinions, not the University's)
Systems Research Group
University of Cambridge Computer Laboratory

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
Re: Basic questions - numbered /New String -- REPOSITORY ,SERVER and VIRTUAL OWNERSHIP [ In reply to ]
Tim
That clarification was very helpful. Thankyou. You are far more patient
about elementary issues than some people would be. It is a privilege to
know people such as yourself. Thanks again.

Your question "Have you thought about using LVM?"
Yes, but right now I am just studying issues where I am behind everyone
else from an insight standpoint. I want to be sure that I properly
understand everything communicated. So I have a ways to go in that respect.

Once I have my satellite connection and static IP address I will
probably step into LVM.

BTW: Regarding my satellite connection. I have offered a repository and
server (operating out of my LAN) to serve Dom 0 and Dom U build
contributions -- if maintained by the contributor(s). My satellite ISP
indicated that there are bandwidth limitations and if you people get
involved and we find they are inadequate I will try to find money to
extend that bandwidth. This server will guarantee that the latest stable
packages are available on an on-going basis. I indicated that Xen
developers would have authority to determine and influence other issues
such as the placement of test and unstable builds on the server as it
suited their purposes. In other words, if desirable, part of the server
operation could be dedicated to Xen development issues that require a
server. All these considerations impact the choice of the server and
necessitate set up by a knowledgeable web master. But for the general
public I indicated that only stable build contributions would be served.
Every build contribution to the repository would have an associated web
page maintained by the remote web master. I generated quite a bit of
email on this subject -- I don't know if you followed the thread as it
got confused with another. Anyway, this server would operate under Xen
development authority and would have a remote web master and
administrator far more qualified than myself (hopefully approved by the
development team) -- I have requested a volunteer(s) to work remotely in
this capacity (with me) and define and implement web server operation
details. I also offered to the contributors (including developers)
remote access to the LAN (if they found it advantageous) so these
packages could be exercised without downloading them (saving internet
bandwidth). I offered gobs of hard drive and this is probably where the
administrator would choose to implement LVM. I would simply be a
provider and responsible for maintaining the physical resources
involved. I do not want to represent myself as something I am not and
feel that this service could be extremely beneficial to everyone
involved. See the email threads.

Regardless, I will be setting up such a server. But I am giving you guys
first crack at usage (part or full) of this server. If you guys are
involved it would be a much more professional server set up than what I
would do for myself. I thought you guys could do a much more
professional job and so my email deliberations for such a hosting
situation have been put forth as a "proposal", the character of which
require that your development group has to agree to be involved at the
very minimum as a regulatory body. I am perfectly willing to see this
set up so you people have as much control over this repository as you
feel you want. You would be calling the shots, not me. This way you
don't find yourselves subjected to someone else's whims and authority
and your authority is not split off into other directions. You are even
welcome to assume the role as prime remote web master over the parts of
my lan you may find useful and request additional resources. All this
comes FREE with no legal obligations in any form on your part so as not
to compromise your project and other associations in any way. This
should be considered by you as an opportunity which you may or may not
want to accept for your own reasons. Anyway, you have whatever time you
need to deliberate this issue and state your terms -- if any. I fully
understand if you feel you have to reject this proposal. I just want to
ensure you have first crack at this.

No hard feelings either way regarding any of your decisions. I am just
trying to be helpful in the ways that I feel I can be helpful and learn
at the same time. Not every idea is in the right place at the right time
and so that is your decision and not mine. If you accept the
opportunity development will have "VIRTUAL" ownership. I did not send
any email to [Xen-devel], you can do that if you see fit.

Thanks -- Ted


Tim Deegan wrote:

>Hi Ted,
>
>On Fri, Apr 01, 2005 at 12:26:13PM -0700, Ted Hilts wrote:
>
>
>>So, theoretically speaking, if I were to use /dev/hde (the fith (5th)
>>drive -- in the sequence a, b, c, d, e, f, if there were no gaps) I
>>would have /hde1 and if on hde I allocated an extended partion with 64
>>logical partitions (numbered 1 to 64) I could install a bootable (using
>>Grub) linux distribution in each of these logical paritions.
>>
>>
>
>Almost, but you can't use quite that many partitions (see below).
>
>
>
>>But I don't think this variability is
>>the case with Grub and the designations are reserved, as in (hd4, n)
>>which always refers to the 5th hard drive and a partition.
>>
>>
>
>Yep. Linux names drives based on how they are connected, e.g., /dev/sda
>for the first SCSI drive, /dev/hdd for the second IDE drive on the
>second channel, and so on.
>
>Linux uses partition numbers 1 to 4 for primary partitions, and then
>counts all the logical partitions from 5 upwards, to a maximum of 63 for
>IDE disks (so you can only have 59 logical partitions) and of 15 for
>SCSI disks (so you can only have 11 logical partitions).
>
>http://lxr.linux.no/source/Documentation/devices.txt#L192
>http://lxr.linux.no/source/Documentation/devices.txt#L272
>
>GrUB allocates drive numbers in the order the BIOS presents them,
>regardless of whether they're SCSI or IDE, so they may not necessarily
>be in the same order that linux sees them in. I believe if you run the
>grub command in a running linux it will try to figure out the mapping
>for you.
>
>GrUB uses partition numbers 0 to 3 for primary partitions, and then
>counts all the logical partitions from 4 upwards.
>
>http://www.gnu.org/software/grub/manual/html_node/Naming-convention.html#Naming%20convention
>
>
>
>>Continuing theoretically: If that drive contained just logical
>>partitions I would have on hde (hde1, hde2, hde3...hden) where n=1 to 64
>>-- I believe 0 refers to the extended partition itself but I am not sure
>>as it is not used just assumed.
>>And if this were translated into GRUB terminology I would have ((hd4,0),
>>(hd4,1), (hd4,2)...(hdn-1,n-1) where n =1 to 64
>>
>>
>
>If you have 59 logical partitions on the 5th IDE drive (and that drive
>is the master device on the third controller, and you don't have any
>non-IDE drives...), GrUB should call the logical partitions (hd4,4) to
>(hd4,62). Linux should call them /dev/hde5 to /dev/hde63.
>
>Have you thought about using LVM? :)
>
>HTH,
>
>Tim.
>
>
>