Pkg xen devel - Mar 2013 - Bug#702369: latest install report

I've had another attempt at installing Wheezy 64 bit from a template as
PV. I attach my full script below.

I've made much more progress this time, but there are still outstanding
issues:

- neither "xe console" or VNC seems to allow ALT-F[2-6] keystrokes, so
I
can't get other virtual terminals and therefore can't obtain any useful
log/status messages. This issue alone makes me feel I'm really wasting
my time as it prevents me doing any serious troubleshooting efficiently.

- when I connect with "xe console", it doesn't show me the current
content of the screen, I have to press enter - but that pushes the
installer to the next menu

- I still can't get expert mode install, it always seems to go with the
basic install (unless I do a HVM install like the other day, booting
from ISO)

- networking is really troublesome, I have to deliberately hit the
cancel button in both the IPv6 auto-config and DHCP screens, otherwise
it seems to think it has a valid IPv6 address and doesn't let me put in
the desired IPv4 address - this is partly speculation though, because
without the other virtual consoles, I can't really check which address
it picked up

- I couldn't find any convenient way to override the disk size when
using the vm-install command, and I have to delete and recreate the VDI,
this is a pain when scripting

However, my install did succeed, but it gives some errors about "g.e.
still in use":

??????????????????????? Finishing the installation ????????????????????????
? ?
? 95% ?
? ?
The system is going down NOW!ystem... ?
Sent SIGTERM to all processes ?
Sent SIGKILL to all
processes????????????????????????????????????????????????
Requesting system reboot
[ 1419.728488] WARNING: g.e. still in use!
[ 1419.728500] WARNING: g.e. still in use!
[ 1424.673842] Restarting system.


Here is the script I used to build the VM:


#!/bin/bash

set -e

NEW_HOST=wheezytest2
DISK_SIZE=5GiB
TMPL="Debian Wheezy 7.0 (64-bit)"
DEB_MIRROR=http://ftp.ch.debian.org/debian
NET0=xenbr1
SR_NAME="testSR0"


TMPL_UUID=$(xe template-list name-label="${TMPL}" --minimal)
NET0_UUID=$(xe network-list bridge=${NET0} --minimal)
SR_UUID=$(xe sr-list name-label=${SR_NAME} --minimal)

VM_UUID=$(xe vm-install template=$TMPL_UUID
new-name-label="${NEW_HOST}")

xe vm-param-set uuid=${VM_UUID}
other-config:install-repository=${DEB_MIRROR}

xe vif-create vm-uuid=${VM_UUID} network-uuid=${NET0_UUID} device=0

# the following 6 lines are needed to change the VDI size
# because there is no disk-size option for vm-install
VBD_UUID=$(xe vbd-list vm-uuid=${VM_UUID} --minimal)
VDI_UUID=$(xe vdi-list vbd-uuids=${VBD_UUID} --minimal)
xe vdi-destroy uuid=${VDI_UUID}
xe vm-disk-add vm=${VM_UUID} disk-size=${DISK_SIZE} device=0
VBD_UUID=$(xe vbd-list vm-uuid=${VM_UUID} --minimal)
xe vbd-param-set bootable=true uuid=${VBD_UUID}

echo "New VM UUID = ${VM_UUID}"

xe vm-start uuid=${VM_UUID}

On 03/09/2013 02:35 AM, Daniel Pocock wrote:
> I've made much more progress this time, but there are still outstanding
> issues:
> 
> - neither "xe console" or VNC seems to allow ALT-F[2-6]
keystrokes, so I
> can't get other virtual terminals and therefore can't obtain any
useful
> log/status messages. This issue alone makes me feel I'm really wasting
> my time as it prevents me doing any serious troubleshooting efficiently.
> 
> - when I connect with "xe console", it doesn't show me the
current
> content of the screen, I have to press enter - but that pushes the
> installer to the next menu
This is a problem with Xen, not with XAPI, IMO. Xen doesn't keep a
buffer of the tty1, which is a real annoyance. The problem is the same
with "xm console" and "xl console". But it must be possible
to start
your vm into a paused state (I haven't checked, but that might be one of
the options of "xe vm-start", just like with "xm start -p"),
attach to
its console, and THEN unpause it. This works well with xend.
> - I still can't get expert mode install, it always seems to go with the
> basic install (unless I do a HVM install like the other day, booting
> from ISO)
I believe that you can use a boot option that can be set.
> - networking is really troublesome, I have to deliberately hit the
> cancel button in both the IPv6 auto-config and DHCP screens, otherwise
> it seems to think it has a valid IPv6 address and doesn't let me put in
> the desired IPv4 address - this is partly speculation though, because
> without the other virtual consoles, I can't really check which address
> it picked up
Well, can't you go into the "Execute a shell" option of d-i to
check?
> - I couldn't find any convenient way to override the disk size when
> using the vm-install command, and I have to delete and recreate the VDI,
> this is a pain when scripting
Well, the default HDD size is with the template itself, so the only way
to change the default, is either to change the default template, or make
a new one. Making a new one is basically what you do when you do:

vm=$(xe vm-install template=$template new-name-label=debian)

Then you can resize the HDD:

xe vm-disk-list vm=$vm
xe vdi-resize uuid=<UUID-OF-THE-VDI> disk-size=10GiB

I think that's quite easy, no? This comes directly from the
README.Debian, paragraph 6.2.
> However, my install did succeed, but it gives some errors about "g.e.
> still in use":
> 
> ??????????????????????? Finishing the installation ????????????????????????
> ? ?
> ? 95% ?
> ? ?
> The system is going down NOW!ystem... ?
> Sent SIGTERM to all processes ?
> Sent SIGKILL to all
> processes????????????????????????????????????????????????
> Requesting system reboot
> [ 1419.728488] WARNING: g.e. still in use!
> [ 1419.728500] WARNING: g.e. still in use!
> [ 1424.673842] Restarting system.
That's a warning, not an error, so I believe it can be ignored.

Cheers,

Thomas