Hello,
I have been trying xen for a couple of days with no success (what
has permited me to learn much more about it ;-). I cannot boot any
user domain with any kernel I have tried. I explain the problem below.
1 - I build xen from the debian sources:
apt-get build-deps xen; apt-get -b source xen
That's because I'm in Sarge and it wants to upgrade
my libc6 which scares me (the source package is in
unstable, that's why I need to compile it).
2 - I get and install this:
root@backupsid:~/xen# ls *deb
kernel-patch-xen_2.0.6-1_i386.deb libxen-dev_2.0.6-1_i386.deb
xen_2.0.6-1_i386.deb
libxen2.0_2.0.6-1_i386.deb libxen-python_2.0.6-1_i386.deb
xen-docs_2.0.6-1_all.deb
3 - Then I aptitude install the kernel-source for 2.6.11 (btw since
2.6.12 it's called linux-source)
4 - I copy the .config of the kernel I'm using to kernel-source-2.6.11
5 - I edit it and add by hand this, otherwise I miss lots of
already configured options:
#
CONFIG_XEN=y
CONFIG_ARCH_XEN=y
CONFIG_NO_IDLE_HZ=y
#
# XEN
#
CONFIG_XEN_PRIVILEGED_GUEST=y
CONFIG_XEN_PHYSDEV_ACCESS=y
CONFIG_XEN_BLKDEV_BACKEND=y
CONFIG_XEN_NETDEV_BACKEND=y
# CONFIG_XEN_BLKDEV_FRONTEND is not set
# CONFIG_XEN_NETDEV_FRONTEND is not set
CONFIG_XEN_WRITABLE_PAGETABLES=y
CONFIG_XEN_SCRUB_PAGES=y
CONFIG_X86=y
# CONFIG_X86_64 is not set
CONFIG_HAVE_ARCH_DEV_ALLOC_SKB=y
It happens the same to this guy:
http://www.xmlvalidation.com/xen_howto_sarge.0.html
6 - I compile the kernel for dom0 with the kernel-patch-xen
installed and with this command:
$ cd /usr/src/kernel-source-2.6.11
$ PATCH_THE_KERNEL=YES make-kpkg --config=menuconfig --arch=xen
--revision=1 --append-to-version=myxen0 kernel_image
7 - I do the same after a 'make mrpropper' and go back to the step 5
but this time I configure the options for what the Xen people call
the 'user domain'.
8 - I install the myxen0 package and myxenU package that I have created
in steps 6 and 7.
9 - I configure Grub:
root@backupsid:/usr/src/kernel-source-2.6.11# tail
-5 /boot/grub/menu.lst
title Xen 2.0 / XenLinux 2.6.11myxen0
kernel /boot/xen.gz dom0_mem=400000
root (hd0,0)
module /boot/xen-linux-2.6.11myxen0 root=/dev/md0 ro console=tty0
10 - I boot with this kernel
11 - I have this domain configured:
kernel = "/boot/vmlinuz-2.6.11.12-xenU"
memory = 300
name = "pau"
vif = [ 'mac=aa:00:00:00:00:11, bridge=xen-br0' ]
disk = [ 'phy:md1,md1,w' ]
root = "/dev/md1 ro"
extra = "4"
I use raid1 but this is not the problem, I happens exactly the
same in my laptop.
11 - I start xend: /etc/init.d/xen start
root@backupsid:/home/ramon# xm list
Name Id Mem(MB) CPU State Time(s) Console
Domain-0 0 385 0 r---- 13.3
12 - I start the domain I configured in step 11:
root@backupsid:/home/ramon# xm create -c pau.conf
Using config file "/etc/xen/pau.conf".
Started domain pau, console on port 9602
************ REMOTE CONSOLE: CTRL-] TO QUIT ********
Linux version 2.6.11.12-xenU (xenod@xxxxxxxxxxxxxxxxxxxxx) (gcc version
3.3.3 20040412 (Red Hat Linux 3.3.3-7)) #1 Thu Aug 4 00:48:08 BST 2005
BIOS-provided physical RAM map:
Xen: 0000000000000000 - 0000000012c00000 (usable)
300MB LOWMEM available.
(....)
It gets hanged here, at another console I do xm list and see
only the main domain (dom0).
A 'top' reveals that a python process is eating the CPU:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
2657 root 25 0 7796 4328 2120 R 99.9 1.1 3:03.42 python
It is the xm create process.
a strace to it shows me that:
recv(3, "", 1024, 0) = 0
write(1, "", 0) = 0
recv(3, "", 1024, 0) = 0
And never ends.
The log shows me that It starts and dies immediatly after
connecting to the console:
[2005-08-16 13:15:43 xend] DEBUG (XendDomainInfo:1107) Creating vif
dom=2 vif=0 mac=aa:00:00:00:00:11
[2005-08-16 13:15:43 xend] INFO (XendRoot:113) EVENT>
xend.console.create [13, 2, 9602]
[2005-08-16 13:15:44 xend] INFO (XendRoot:113) EVENT> xend.domain.create
['pau', '2']
[2005-08-16 13:15:44 xend] INFO (XendRoot:113) EVENT>
xend.domain.unpause ['pau', '2']
[2005-08-16 13:15:44 xend] INFO (console:44) Console connected 13
127.0.0.1 1458
[2005-08-16 13:15:44 xend] INFO (XendRoot:113) EVENT>
xend.console.connect [13, '127.0.0.1', 1458]
[2005-08-16 13:15:44 xend] DEBUG (blkif:203) Connecting blkif to event
channel <BlkifBackendInterface 2 0> ports=14:3
[2005-08-16 13:15:44 xend] DEBUG (XendDomain:244) XendDomain>reap>
domain died name=pau id=2
[2005-08-16 13:15:44 xend] INFO (XendDomain:568) Destroying domain:
name=pau
[2005-08-16 13:15:44 xend] DEBUG (XendDomainInfo:665) Destroying vifs
for domain 2
[2005-08-16 13:15:44 xend] DEBUG (netif:305) Destroying vif domain=2
vif=0
[2005-08-16 13:15:44 xend] DEBUG (XendDomainInfo:674) Destroying vbds
for domain 2
[2005-08-16 13:15:44 xend] DEBUG (blkif:552) Destroying blkif domain=2
[2005-08-16 13:15:44 xend] DEBUG (blkif:408) Destroying vbd domain=2
idx=0
[2005-08-16 13:15:44 xend] DEBUG (XendDomainInfo:634) Closing console,
domain 2
[2005-08-16 13:15:44 xend] DEBUG (XendDomainInfo:622) Closing channel to
domain 2
[2005-08-16 13:15:44 xend] INFO (XendRoot:113) EVENT> xend.virq 4
[2005-08-16 13:15:44 xend] DEBUG (blkif:363) Unbinding vbd (type phy)
from md1
[2005-08-16 13:15:44 xend] INFO (console:58) Console disconnected 13
127.0.0.1 1458
[2005-08-16 13:15:44 xend] INFO (XendRoot:113) EVENT> xend.domain.exit
['pau', '2', 'crash']
[2005-08-16 13:15:44 xend] INFO (XendRoot:113) EVENT>
xend.domain.destroy ['pau', '2']
[2005-08-16 13:15:44 xend] INFO (XendRoot:113) EVENT>
xend.console.disconnect [13, '127.0.0.1', 1458]
[2005-08-16 13:15:45 xend] INFO (XendRoot:113) EVENT> xend.domain.died
['pau', '2']
I have installed python 2.3 and python2.4 with same results. I have
installed the 2.0.7 from the sources as well with same results.
I have seen in the list people with the same problem but with no
answer:
http://lists.xensource.com/archives/html/xen-users/2005-07/msg00526.html
http://lists.xensource.com/archives/html/xen-users/2005-07/msg00403.html
http://lists.xensource.com/archives/html/xen-users/2005-07/msg00124.html
I'm stalled right now. If the listers of the messages above got to solve
it, it would be great that they explain how did they do it (I'll do the
same).
I must say that I have tried already compiled kernels from SuSE and the
result was the same.
I'll let you know my progress!
Thanks for your time.
Ramon
El lun, 15-08-2005 a las 18:00 +0100, Michael Moritz escribió:
> Hi Ramon
>
> the same moment i sent my msg about the qemu to you I started playing around
> with Xen. Since I had to go to the collo center anyway today I installed it
> on the backup server and got currently the backup server running. I havent
> put a second machine on it yet (so only dom0).
> What processwas hanging when you tried and did it bring down the whole
> machine? Are you using same version (2.0.6)?
>
> mimo
>
> On Monday 15 August 2005 14:27, you wrote:
> > Hi Mimo!
> >
> > This is great, have you tried to launch more than one at the
> > same time? I guess you have configured the network, have you
> > used a bridge? How did you do it? Are the listening ports of
> > the emulated machine reachable? I've been spending some
> > evenings with Xen. I haven't succeed, I have had some
> > troubles with a python process which hangs the whole program.
> > The good thing is that I have learnt much more than if it
> > would have worked out of the box.
> >
> > See you and keep leting me know! :-)
> >
> > Ramon
> >
> > El dom, 14-08-2005 a las 16:59 +0100, Michael Moritz escribió:
> > > Hi Ramon
> > >
> > > having spend a weekend with qemu I have now managed to get it working so
> > > it could theoretically do what a backup server would need. Have a look at
> > >
> > > http://mimo.gn.apc.org/index.php/Running Debian inside qemu using console
> > >
> > > and tell me waht you think.
> > > Hope you are well,
> > >
> > > mimo
> > >
> > > On Monday 08 August 2005 11:56, you wrote:
> > > > Hi there!
> > > >
> > > > El dom, 07-08-2005 a las 13:25 +0100, Michael Moritz escribió:
> > > > > Yes, I see what you mean. Our high availability server is currently
> > > > > there to replace any of ~10 machines if necessary. So maybe not
> > > > > perfect, since you and us would rely on it. Luckily, we never had to
> > > > > use it but you never know. Just found this: http://www.rsnapshot.org
> > > >
> > > > Interesting but too much code IMHO, I studied it some months ago and
> > > > it's basically inspired in this article:
> > > > http://www.mikerubel.org/computers/rsync_snapshots/
> > > >
> > > > I was inspired in this document when I decided to implement
> > > > and document this:
> > > > http://cargol.net/~ramon/snapshots but in Spanish...
> > > >
> > > > What I do for HA is documented here:
> > > > http://cargol.net/~ramon/ad and the syncronization here:
> > > > http://cargol.net/~ramon/ad#AEN43 The problem is that is written
> > > > in Spanish :-( The scripts are very easy though.
> > > >
> > > > > I'd still like to explore the possibilities of running a backuped
> > > > > machine inside the high availability machine. So I'll do my
> > > > > experiments.
> > > >
> > > > Sure! I have already done some exploration with UML a year ago but
> > > > I stopped for lack of time. If you get to use qemu, xen or uml
> > > > to start a server and bring up a IP in which the virtualised server
> > > > can listen let me know!
> > > >
> > > > If we finally decide to collaborate I'd like to help in configuring
> > > > and document a good and simple HA system.
> > > >
> > > > > > What we noticed about the activities we both do and how
> > > > > > similar they are is very interesting as well. Think about
> > > > > > sharing the costs of the servers you use right now... This
> > > > > > approach has pros and cons, we probably should analyse it
> > > > > > further, what do you think? I'll comment it to the people
> > > > > > at Pangea, we usually get sensible conclusions after the
> > > > > > meetings :-)
> > > > >
> > > > > Same here. Let's see what possibilities we can come up with.
> > > >
> > > > Yes, we are meeting in September here. I have proposed three
> > > > aproaches:
> > > >
> > > > * Co-locate and nothing else.
> > > > * Co-locate and share backups and HA
> > > > * Share web, mail, backups and HA servers.
> > > >
> > > > We'll discuss about every of them, sure you too! :-)
> > > >
> > > > > > If we'd decide to buy and co-locate the boxes there I'd
> > > > > > fly to London to install and configure them, which I guess
> > > > > > is cheaper, easier and more secure than configuring them
> > > > > > here and shipping them to there.
> > > > >
> > > > > Makes sense, our matress shall be your home ;)
> > > >
> > > > Great, thanks! I'll have still my friend's matress as well but a
> > > > backup is allways good!
> > > >
> > > > See you at yahoo messenger.
> > > >
> > > > Ramon
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
Home