Running SLES 10 as pv guest in Fedora 7

[Felix Schwarz]

Hi,

I tried to run a SLES 10 installation as (pv) guest in Fedora 7 but the guest 
does not start.

First, I installed SLES using HVM which was somewhat slow (due to Suse's 
graphics mode) but worked. After that I installed kernel-xen from the SLES 
install media.

I found Gerd Hoffmann's very helpful page 
http://et.redhat.com/~kraxel/xen/suse-guest.html and tried to incorporate his 
suggestions regarding the domUloader.

My current config:
-----------------------------------------------------------------
name = "sles10_64_test"
memory = "512"

disk = [ 'tap:aio:/var/lib/xen/images/sles10_64_test,hda,w', ]

vif = [ 'mac=00:16:3e:4a:f5:f0, bridge=eth0', ]

vfb = [ 'type=vnc,vncunused=1' ]

uuid = "b1fd65e84128ced2dc435bb9d2e1861e"
bootloader="/tmp/sles-install/domUloader.py"
bootentry = "hda2:/boot/vmlinuz-xen,/boot/initrd-xen"
vcpus=1
on_reboot   = 'restart'
on_crash    = 'restart'
-----------------------------------------------------------------

/tmp/sles-install/domUloader.py exists and is executable, SElinux is configured 
as permissive.

I guess something is wrong with the bootloader, from /var/log/xen/xend.log:
-----------------------------------------------------------------

2007-07-26 19:28:47 4590] DEBUG (XendDomainInfo:1463) XendDomainInfo.constructDomain
[2007-07-26 19:28:47 4590] DEBUG (balloon:113) Balloon: 538932 KiB free; need 
2048; done.
[2007-07-26 19:28:47 4590] DEBUG (XendDomain:443) Adding Domain: 33
[2007-07-26 19:28:47 4590] DEBUG (XendDomainInfo:1517) 
XendDomainInfo.initDomain: 33 256
[2007-07-26 19:28:47 25532] DEBUG (XendBootloader:100) Launching bootloader as 
['/tmp/sles-install/domUloader.py
', '--output=/var/run/xend/boot/xenbl.24430', '-q', 
'/var/lib/xen/images/sles10_64_test'].
(xm command hangs, nothing happens anymore, have to kill the guest)
-----------------------------------------------------------------

Did anyone manage to run SLES {9,10} as pv guest?

thank you for reading :-)
fs

[Gerd Hoffmann]

Felix Schwarz wrote:
> Hi,
>
> I tried to run a SLES 10 installation as (pv) guest in Fedora 7 but the
> guest does not start.
> I found Gerd Hoffmann's very helpful page
> http://et.redhat.com/~kraxel/xen/suse-guest.html and tried to
> incorporate his suggestions regarding the domUloader.

Hmm, looks like the page needs some updating ...

Some patches by the solaris guys for pygrub made booting sles (and
opensuse) easier in xen 3.1 (I think 3.0.4 too).  You can simply put
these lines into your config file ...

kernel=/boot/vmlinux-xen
   ramdisk=/boot/initrd-xen
   bootloader=/usr/bin/pygrub

... and pygrub will go and pull kernel+ramdisk from the guest disk
without looking for a config file on the guest disk.  No need for
domUloader any more.

HTH,
  Gerd

PS: all IIRC, havn't actually tried that for some time.

[Andrew Cathrow]

On Mon, 2007-08-06 at 11:04 +0200, Gerd Hoffmann wrote:
> Felix Schwarz wrote:
> > Hi,
> >
> > I tried to run a SLES 10 installation as (pv) guest in Fedora 7 but the
> > guest does not start.
> > I found Gerd Hoffmann's very helpful page
> > http://et.redhat.com/~kraxel/xen/suse-guest.html and tried to
> > incorporate his suggestions regarding the domUloader.
>
> Hmm, looks like the page needs some updating ...
>
> Some patches by the solaris guys for pygrub made booting sles (and
> opensuse) easier in xen 3.1 (I think 3.0.4 too).  You can simply put
> these lines into your config file ...
>
> kernel=/boot/vmlinux-xen
>    ramdisk=/boot/initrd-xen
>    bootloader=/usr/bin/pygrub
>
> ... and pygrub will go and pull kernel+ramdisk from the guest disk
> without looking for a config file on the guest disk.  No need for
> domUloader any more.

I know that SLES10SP1 will run on RHEL5.1beta - I'm running a PV guest.
I ended up copying the vmlinux-xenpae and initrd-xenpae outside of the
guest - on the dom0 filesystem.
I didn't debug too deeply but using reiserfs was probably by first
mistake ,,,,,,,
> HTH,
>   Gerd
>
> PS: all IIRC, havn't actually tried that for some time.
>
> --
> Fedora-xen mailing list
> [e-mail]
> https://www.redhat.com/mailman/listinfo/fedora-xen

[Felix Schwarz]

Andrew Cathrow schrieb:
> I ended up copying the vmlinux-xenpae and initrd-xenpae outside of the
> guest - on the dom0 filesystem.

I think, I do not need vmlinux-xenpae etc. if I use 64 Bit Dom0/DomU, correct?
> I didn't debug too deeply but using reiserfs was probably by first
> mistake ,,,,,,,

At least that is no problem to me: I never chose reiserfs as I don't trust this 
fs. .-)

fs

[Andrew Cathrow]

On Mon, 2007-08-06 at 16:00 +0200, Felix Schwarz wrote:
> Andrew Cathrow schrieb:
> > I ended up copying the vmlinux-xenpae and initrd-xenpae outside of the
> > guest - on the dom0 filesystem.
>
> I think, I do not need vmlinux-xenpae etc. if I use 64 Bit Dom0/DomU, correct?

Correct
> > I didn't debug too deeply but using reiserfs was probably by first
> > mistake ,,,,,,,
>
> At least that is no problem to me: I never chose reiserfs as I don't trust this 
> fs. .-)

Yeah - comes from be hitting next, next, next
> fs
>
> --
> Fedora-xen mailing list
> [e-mail]
> https://www.redhat.com/mailman/listinfo/fedora-xen

[Felix Schwarz]

Hi,

Gerd Hoffmann schrieb:
> Hmm, looks like the page needs some updating ...
>
> Some patches by the solaris guys for pygrub made booting sles (and
> opensuse) easier in xen 3.1 (I think 3.0.4 too).  You can simply put
> these lines into your config file ...
>
> kernel=/boot/vmlinux-xen
>    ramdisk=/boot/initrd-xen
>    bootloader=/usr/bin/pygrub
>
> ... and pygrub will go and pull kernel+ramdisk from the guest disk
> without looking for a config file on the guest disk.  No need for
> domUloader any more.

Thanks for your help. I finally got it working, so maybe this may interest some 
people:
------------------------------------------------------------------------
name = "sles10_64_test"
memory = "512"

disk = [ 'tap:aio:/var/lib/xen/images/sles10_64_test,hda,w', ]
vif = [ 'mac=00:16:3e:4a:f5:f0, bridge=eth0', ]
vfb = [ 'type=vnc,vncunused=1' ]
uuid = "b1fd65e84128ced2dc435bb9d2e1861e"
bootloader = "/usr/bin/pygrub"
kernel = "/boot/vmlinuz-xen"
ramdisk = "/boot/initrd-xen"

vcpus=1
on_reboot   = 'restart'
on_crash    = 'restart'
------------------------------------------------------------------------

The only thing which is missing is the VNC console. Does SLES 10 not support it?

fs

LUN to VM

[Reed, Tameika N]

Hi All,

I have some questions about presenting LUNS to a virtual machine.  My
configuration is as follows:

1 HP EVA 8000 that is being presented to a blade bl 456c running Rhel 5.  The
kernel is basic 2.6.18 xen kernel.  I can see the LUNS on the blade, by using
multipath.  Below is a snippet of mulitpath.

mpath10 (3600508b4001069430001d000017d0000) dm-17 HP,HSV200
[size=100G][features=1 queue_if_no_path][hwhandler=0]
_ round-robin 0 [prio=0][active]
 _ 0:0:0:11 sdk 8:160 [active][undef]
mpath3 (3600508b4001069430001d00001960000) dm-5 HP,HSV200
[size=250G][features=1 queue_if_no_path][hwhandler=0]
_ round-robin 0 [prio=0][active]
 _ 0:0:0:4  sdd 8:48  [active][undef]

Now as you can see the devices are showing up under /dev/sd* and
/dev/mapper/mpath*.  I can mount the LUNS to  blade without any problems. I
now want these devices to be seen by the virtual machine created on the san.
Below is a my configuration that I have set up for one of my VM's.

VM Config file:

# Automatically generated xen config file
name = "REDDISK"
builder = "hvm"
memory = "2000"
disk = [ 'file:/opt/reddiskd0,hda,w','phy:/dev/mapper/mpath0,sda1,w', ]
vif = [ 'type=ioemu, mac=00:16:3e:4d:d2:ea, bridge=xenbr0', ]
uuid = "57ad8f41-cd50-5e89-985f-6462aa5a5f9b"
device_model = "/usr/lib64/xen/bin/qemu-dm"
kernel = "/usr/lib/xen/boot/hvmloader"
vnc=1
vncunused=1
apic=1
acpi=1
pae=1

vcpus=1
serial = "pty" # enable serial console
on_reboot   = 'restart'
on_crash    = 'restart'
**************************************************************

I can also run the following command from cli and see the output in my
configuration file once I enter this command:

command 1: xm block-attach TXWD_COURT /block/sda sda w

command 2: xm list REDDISK --long | less

results from command 2

domain
    (domid 20)
    (uuid 57ad8f41-cd50-5e89-985f-6462aa5a5f9b)
    (vcpus 1)
    (cpu_weight 1.0)
    (memory 2000)
    (shadow_memory 9)
    (maxmem 2000)
    (features )
    (name REDDISK)
    (on_poweroff destroy)
    (on_reboot restart)
    (on_crash restart)
    (image
        (hvm
            (kernel /usr/lib/xen/boot/hvmloader)
            (device_model /usr/lib64/xen/bin/qemu-dm)
            (pae 1)
            (vcpus 1)
            (boot c)
            (serial pty)
            (vnc 1)
            (vncunused 1)
            (display localhost:1003.0)
            (xauthority /root/.xauthJl2aRE)
            (acpi 1)
            (apic 1)
            (vncpasswd 'None')
        )
    )
    (device
        (vif
            (backend 0)
            (script vif-bridge)
            (bridge xenbr0)
            (mac 00:16:3e:4d:d2:ea)
            (type ioemu)
        )
    )
    (device
        (vbd (backend 0) (dev hda:disk) (uname file:/opt/reddiskd0) (mode w))
    )
    (device (vbd (backend 0) (dev sda:disk) (mode w)))
    (device (vkbd (backend 0)))
    (device
        (vfb
            (backend 0)
            (type vnc)
            (vncunused 1)
            (display localhost:1003.0)
            (xauthority /root/.xauthJl2aRE)
        )
    )
    (state -b----)
    (shutdown_reason poweroff)
    (cpu_time 116.031996341)
    (online_vcpus 1)
    (up_time 3220.21273112)
    (start_time 1186604696.56)
    (store_mfn 61791)
)

This tells me that the command was ok an that it should be see as sda on my
virtual machine ( i think)?
*************************************************************************

So sum up all my questions:

1.  Is my configuration file correct to present LUNS to the VM's ?
disk = [ 'file:/opt/reddiskd0,hda,w','phy:/dev/mapper/mpath0,sda1,w', ]

2.  Do I need to create a scsi block device in order for this work?
ie mknod or MAKEDEV. Reason being none of the scsi partitions show up at all
I checked cat /proc/partitions and /dev

3.Am I presenting this information wrong to the VM?

4.  I have also explored setting up WWID inside the VM but if it can't see
the SAN then it does me no good for right now.

I know that this may see confusing but if you need me to clarify then I can
do that.

ANY IDEAS OR SUGGESTIONS ?

Thanks in advance,

Tameika

LUN to VM

[Reed, Tameika N]

Hi All,

So sum up all my answers to questions

1.  This line is correct!
disk = [ 'file:/opt/reddiskd0,hda,w', 'phy:/dev/mapper/mpath0,hdb,w',
'phy:/dev/mapper/mpath4,hdc,w', ]

2.  SCSCi device create is not needed futhermore once you create it, it will
not be there upon rebooting the server.

3.The way this works is as long as the LUN is not mounted on the blade in
which the guest vm is running on. Then the VM can mount the LUN.  The VM can
mount the LUN even if it is presented and mounted to a different blade.  The
one key point here is as long as the LUN is not mounted to blade where the
virtual machine is running from, it should be able to mount it.

4.  WWID is only necessary if I want to use multipath inside VM.  For the
most part I rebooted the blades and the VM's serveral time and they come back
with the same WWID. so far...
I will continue to look at this post back if any significant changes.

Thanks

Tameika