Red Hat Bugzilla – Bug 499935
virt manager install of guest doesnt work
Last modified: 2009-08-31 09:13:49 EDT
Description of problem:
I am trying to install Solaris X86 u7 (the latest) from the dvd iso file using virt manager.
virtualization method: full
initial memory: 812
OS: sun solaris x86
Disk image: /opt/image/solaris.img
Disk size: 14000
connection type: virtual network
enable audio: true.
It initializes the disk image file, then it ends up erroring out with:
Unable to complete install '<class 'libvirt.libvirtError'> internal error Cannot find QEMU binary /usr/lib/xen/bin/qemu-dm: No such file or directory
Traceback (most recent call last):
File "/usr/share/virt-manager/virtManager/create.py", line 723, in do_install
dom = guest.start_install(False, meter = meter)
File "/usr/lib/python2.5/site-packages/virtinst/Guest.py", line 732, in start_install
return self._do_install(consolecb, meter, removeOld, wait)
File "/usr/lib/python2.5/site-packages/virtinst/Guest.py", line 767, in _do_install
self.domain = self.conn.createLinux(install_xml, 0)
File "/usr/lib/python2.5/site-packages/libvirt.py", line 915, in createLinux
if ret is None:raise libvirtError('virDomainCreateLinux() failed', conn=self)
libvirtError: internal error Cannot find QEMU binary /usr/lib/xen/bin/qemu-dm: No such file or directory
This appears to me to be some path not getting set correctly in the management scripts.
ln -s /usr/bin/qemu-kvm /usr/lib/xen/bin/qemu-dm
and correcting a few selinux issues got me to the install.
booting solaris media I get:
Warning: /pci@0,0/pci8086,7020@1,2 (uhci0: No SOF interrupts have been received, this USB UHCI host controller is unusable
It appears I have no USB Solaris support for at least their installer.
Is this a bug in kvm/qemu or should this be filed with Sun?
I apologize this probably should be two separate bugs.
Please contact me if i need to file a sun bug report or if you need more info.
Could you attach ~/.virt-manager.log? Also the versions of libvirt, python-virtinst, virt-manager
/usr/lib/xen/bin/qemu-dm is relevant when you're using Xen, so it appears virt-manager/libvirt thinks this is a Xen host?
Created attachment 343524 [details]
virt manager log
the last attempt to use this, i changed the vm references from xen to kvm for the last attempt. It crapped out in the same spot. Virtual manager hangs and you have to force quit it.
It is crapping out in the install at loading java if you enable any kind of gui, and right after probing for zfs using either kvm or the zen configuration.
This is Fedora 10
[root@orc .virt-manager]# rpm -q libvirt python-virtinst virt-manager
(the gui issue was reported by someone else and filed as another bug.)
Let me know if you need more info.
deleted the /etc/libvirt/qemu/solarisx86.xml config file. restarted libvirt.
(no idea why it was using xen to begin with, and i had retried with kvm, but i wonder if i didnt restart kvm and it rewrote the original config file back?)
it got through the install interface
crapped out at:
Preparing system for Solaris install
Configuring disk (c0d0)
- Creating Fdisk partition table
ERROR: Could not create Fdisk partition table
ERROR: Could not label disks
ERROR: Could not update disks with new configuration
mkdir: "/a/var/sadm/launcher": Read-only file system
Could not update disks with new configuration
Solaris installation program exited.
Created attachment 343526 [details]
reattaching the updated virt-manager log from the install as it includes the version that crapped out at making the partitions.
note: it did NOT give me an option for picked full emulation and there wasnt a drop down for xen/kvm emulator, it autopicked xen.
the last issue I believe was selinux since i changed the name of the vm file.... doh.
This isn't making much sense to me. virt-manager clearly connected to the QEMU driver:
[Mon, 11 May 2009 21:44:39 virt-manager 4192] DEBUG (connection:372) Scheduling background open thread for qemu:///system
But it is deciding the the hypervisor type is 'Xen'
[Mon, 11 May 2009 21:47:07 virt-manager 4192] DEBUG (create:667) Creating a VM solaris10x86
OS: Sun Solaris 10
Max Memory: 512
# VCPUs: 1
Disk image: /opt/images/solaris10x86.img
Non-sparse file: True
Likewise generating XML for type='xen', which is only valid when connected to the Xen driver in libvirt - this isn't even valid for xenner, which is still type=kvm, but with an OS type of xen instead of hvm:
<disk type='file' device='disk'>
<target dev='hda' bus='ide'/>
<disk type='file' device='cdrom'>
<target dev='hdc' bus='ide'/>
<input type='mouse' bus='ps2'/>
<graphics type='vnc' port='-1' keymap='en-us'/>
Can you attach the output of
# virsh -c qemu:///system capabilities
Created attachment 343649 [details]
This is the requested virtsh output.
It installed and =seems= to be working (rather slowly), but working nonetheless.
The shutdown button (if you force shutdown it works) and the OHCI (USB) driver/controller issue.
It did crash when networking was off and the gui of course doesnt work correctly. (not something I particularly care about but it would be nice to have a better video card..)
The shutdown buttons -does- actually get the machine to power down sometimes but not consistantly.
The NEW thing is I am trying to allocate a second cpu to the system. I tried to add a second processor to it through vm manager software it said it saved it and the new vm still comes up as a single processor. I edited the .xml config file so
<vcpu>2</vcpu> for that system and it still only is showing 1 processor after doing a force shutdown and a restart.
I shut vm down, vm manager down, and libvirtd, and checked the vcpu settings. turned it back on the vcpu is now showing two, but the memory setting reverted back to the old smaller value. After waiting a good half an hour, it doesnt appear solaris got much past loading the kernel.
Okay. there is too much going on in this bug.
Reporter, if you are seeing multiple issues, please file them as separate bugs. Any issues you encounter inside the VM should probably be filed against qemu and can be triaged from there. Let's use this bug to track the original report: virt-manager trying to install a xen guest on a qemu connection. Can you still reproduce this issue?
(In reply to comment #9)
> The NEW thing is I am trying to allocate a second cpu to the system. I tried to
> add a second processor to it through vm manager software it said it saved it
> and the new vm still comes up as a single processor. I edited the .xml config
> file so
> <vcpu>2</vcpu> for that system and it still only is showing 1 processor after
> doing a force shutdown and a restart.
> I shut vm down, vm manager down, and libvirtd, and checked the vcpu settings.
> turned it back on the vcpu is now showing two, but the memory setting reverted
> back to the old smaller value.
I think this particular issue is bug #503784
Agree with Cole, though - please file separate bug reports for separate issues
This bug has been inactive for a while, and since I can't reproduce the original issue, I'm closing as WORKSFORME.
Reporter (solanum), if you are still seeing this issue, please reopen this bug, again posting ~/.virt-manager/virt-manager.log. Any other issues should be filed as separate reports. Thanks.