Hi,
I''ve suddenly been told I need to qualify the current Xen platform
I''m
building on some other hardware for our internal use and I''m having a
hard time. I''d really appreciate your help.
The platform is an HP DL-180 server (Quad core Xeon E5620 @2.4GHz) and
the machine I''ve been asked to show-horn the install into is a Dell
R210
(Quad core Xeon X3430 @2.4GHz). I know the X3430 isn''t as hot as the
E5620 particularly in the cache department. They both appear to have
VT-x and VT-d but I don''t think the X3430 has Vpro. I''m not
sure if the
chipset on the Dell is likely to be deficient in any other way.
The install is an Ubuntu 12.04 64-bit Xen 4.2 hypervisor with a 12.04
64-bit guest and a Windows Server 2008 R2-64 guest. There is a single
ext2 partition (sda1) on the main drive that contains the host''s /boot
and all the other partitions are contained in an LVM VG (sda5). A single
PCI card is passed through to the Windows guest using VT-d.
The first of my woes was that I couldn''t put the Clonezilla image onto
the Dell as it contains an LVM volume group, drive on the Dell is
smaller than the HP and Clonezilla says it can''t resize the volume. I
did a dd copy of each of the individual partitions and LVM volumes on
one of the HP boxes. I created a partition table on the Dell and made an
sda1 and sda5. I created all the LVs and dd''d the contents into them.
I did find I''d miscalculated the size of sda1 and had trimmed off the
end (which should have been empty) so used cp -a to move the files out,
reformatted the FS then put them back. I booted the machine on a boot
cd, chroot mounted the host fs and installed grub. The machine booted
but I found I had to play with the udev network rules and grub.cfg to
make everything mount and for the network devices to be recognised. I
also doctored all the guest conf files to reference the correct PCI card
and remove any references to individual CPU bindings.
The machine looks like it boots OK and the host and Linux guest start
fine. The Windows VM steadfastly refuses to boot and if I do an xl
create on it, I get the following messages:
libxl: error: libxl_create.c:420:libxl_domain_make: domain creation fail
libxl: error: libxl_create.c:646:initiate_domain_create: cannot make
domain: -3
libxl: error: libxl.c:1392:libxl__destroy_domid: non-existant domain -1
libxl: error: libxl.c:1356:domain_destrpy_callback: unable to
destroy guest with domid 4294967295
libxl: error: libxl_create.c:1152:domcreate_destruction_cb: unable
to destroy domain 4294967295 following failed creation
These errors don''t seem to contain any useful information to me (the
original error displayed concerned failing to allocate memory but I put
more RAM in the Dell and that went away)
I can''t find any /var/log/xen entries for the Windows guest indicating
why it was destroyed. There''s nothing in kern.log either indicating a
problem. I can mount the Windows guest''s file system on the host. I see
files and I don''t see anything that is obviously corrupt.
Has anyone else tried to move installs between hardware please? I''d
really appreciate any help you could give me in working out why this
isn''t working.
Thanks,
Paul.
_______________________________________________
Xen-users mailing list
Xen-users@lists.xen.org
http://lists.xen.org/xen-users
On Fri, 2013-02-01 at 17:14 +0000, Paul Stimpson wrote:> The machine looks like it boots OK and the host and Linux guest start > fine. The Windows VM steadfastly refuses to boot and if I do an xl > create on it, I get the following messages: > > libxl: error: libxl_create.c:420:libxl_domain_make: domain creation fail > libxl: error: libxl_create.c:646:initiate_domain_create: cannot make domain: -3This seems to correspond to -ESRCH from xc_domain_create.> libxl: error: libxl.c:1392:libxl__destroy_domid: non-existant domain -1 > libxl: error: libxl.c:1356:domain_destrpy_callback: unable to destroy guest with domid 4294967295 > libxl: error: libxl_create.c:1152:domcreate_destruction_cb: unable to destroy domain 4294967295 following failed creation > > These errors don''t seem to contain any useful information to me"xl -vvv create ..." might give you some more information, or you might see some useful messages in "xl dmesg". It might be useful to see the complete logs in case something interesting is hidden there. What does your guest configuration file contain? Ian.
Hi Ian, Thank you for your reply. On 04/02/13 10:07, Ian Campbell wrote:> On Fri, 2013-02-01 at 17:14 +0000, Paul Stimpson wrote: > >> The machine looks like it boots OK and the host and Linux guest start >> fine. The Windows VM steadfastly refuses to boot and if I do an xl >> create on it, I get the following messages: >> >> libxl: error: libxl_create.c:420:libxl_domain_make: domain creation fail >> libxl: error: libxl_create.c:646:initiate_domain_create: cannot make domain: -3 > This seems to correspond to -ESRCH from xc_domain_create. > >> libxl: error: libxl.c:1392:libxl__destroy_domid: non-existant domain -1 >> libxl: error: libxl.c:1356:domain_destrpy_callback: unable to destroy guest with domid 4294967295 >> libxl: error: libxl_create.c:1152:domcreate_destruction_cb: unable to destroy domain 4294967295 following failed creation >> >> These errors don''t seem to contain any useful information to me > "xl -vvv create ..." might give you some more information, or you might > see some useful messages in "xl dmesg". It might be useful to see the > complete logs in case something interesting is hidden there.It seems to go south fairly early on:> libxl: debug: libxl_create.c:1173:do_domain_create: ao 0x10b49a0: > create: how=(nil) callback=(nil) poller=0x10b4920 > libxl: error: libxl_create.c:420:libxl__domain_make: domain creation fail > libxl: error: libxl_create.c:646:initiate_domain_create: cannot make > domain: -3 > libxl: error: libxl.c:1392:libxl__destroy_domid: non-existant domain -1 > libxl: error: libxl.c:1356:domain_destroy_callback: unable to destroy > guest with domid 4294967295 > libxl: error: libxl_create.c:1152:domcreate_destruction_cb: unable to > destroy domain 4294967295 following failed creation > libxl: debug: libxl_event.c:1497:libxl__ao_complete: ao 0x10b49a0: > complete, rc=-3 > libxl: debug: libxl_create.c:1186:do_domain_create: ao 0x10b49a0: > inprogress: poller=0x10b4920, flags=ic > libxl: debug: libxl_event.c:1469:libxl__ao__destroy: ao 0x10b49a0: destroy > xc: debug: hypercall buffer: total allocations:5 total releases:5 > xc: debug: hypercall buffer: current allocations:0 maximum allocations:2 > xc: debug: hypercall buffer: cache current size:2 > xc: debug: hypercall buffer: cache hits:3 misses:2 toobig:0> What does your guest configuration file contain? >> builder = "hvm" > name = "GUEST-Windows" > memory = 2048 > #vcpus = 7 > #cpus = "1,2,3,4,5,6,7" > disk = > [''phy:/dev/VolGroup00/LogVol03,hda,w'',''phy:/dev/DataVG/WindowsData,hdb,w''] > vif = [''bridge=xenbr0,ip=192.168.73.20''] > pci = [''01:00.00''] > usbdevice = "tablet" > vnc = 1 > vnclisten = "0.0.0.0" > vncpassword = "REDACTED"I commented the CPU allocations in the conf files and in /etc/default/grub out (and did an update-grub) since the recipient machine doesn''t have hyperthreading and therefore only seems to have half the number of cores the source machine had. I also tried modding the conf file to make a new one with the Windows DVD mounted to try a reinstall. That also failed to launch. Comparison of the CPUs is here: http://www.cpu-world.com/Compare/825/Intel_Xeon_E5620_vs_Intel_Xeon_X3430.html The source machine has the E5620 and the destination the X3430. The source is an HP DL180G6 and the destination an original Dell Poweredge R210. My thoughts are that possible explanations could be: * The CPU or the chipset on the Dell is missing some feature that is expected. * The number of cores is coded somewhere I haven''t found (I didn''t build the image) and the box only having 4 cores is preventing the whole thing from starting. * Something about Windows is self-destructing milliseconds after launch (but that version of Windows from that DVD is known to run on the Dell native). * Some driver the Dell needs is missing. * Compiling Xen 4.2.0 on the source machine to the stock Ubuntu 12.04-x64 kernel and moving it produces a kernel that is somehow unacceptable on the Dell. Thanks again for your help. Paul.