I just tried installing Fedora 13 release candidate as Xen PV guests/domUs:
- F13 rc 32bit PAE Xen PV guest (kernel 18.104.22.168-85.fc13.i686.PAE)
- F13 rc 64bit Xen PV guest (kernel 22.214.171.124-85.fc13.x86_64)
Xen dom0 for these tests was RHEL 5.5 x86_64.
I installed F13 guests using virt-manager and http fedora mirror.
Both installed OK, no problems found. Absolute mouse pointer worked OK
on the graphical console.
RHEL 5.5 has pygrub support for ext4, so having F13 guest
/boot on ext4 (the default for F13) is not a problem anymore.
I also tested save/restore functionality:
- F13 64bit guest with 1 and 4 vcpus: save/restore OK.
- F13 32bit PAE guest with 1 and 4 vcpus: save/restore OK.
I've got a F13 kvm VM running under CentOS 5.4 and I'm seeing the clock
suddenly jump forward by about 10 hours. I have no idea what is up.
Perhaps confusion about whether the rtc clock is utc or not? (though
I'm in MDT which I think is 6 hours diff). Any ideas?
A few issues I hit trying to build/install the latest xen from the
myoung dom0 repo:
1) [root@librescu SPECS]# rpmbuild -bb xen.spec kernel.spec
error: Failed build dependencies:
/usr/include/gnu/stubs-32.h is needed by xen-4.0.0-0.7.fc12.src
The trick here seems to be glibc-devel is required, but the 32-bit
version. I was building on x86_64. Installing the 32-bit version gets
it to build, but is that the right thing to do? If so, can that be
specified as a buildrequires?
pod2man --section=1 --center=" " --release=" " qemu.pod > qemu.1
/bin/sh: /usr/local/bin/pod2man: /usr/local/bin/perl: bad interpreter:
No such file or directory
A simple symlink from /usr/bin/perl to /usr/local/bin/perl does the
trick there. Fedora's been shipping a usable perl for at least several
releases. :) I presume a script needs to be patched to use system perl.
3) During xen RPM install:
Stopping xend daemon: [ OK ]
Starting xend daemon: [FAILED]
warning: %post(xen-4.0.0-0.7.fc12.x86_64) scriptlet failed, exit status 1
[2010-05-04 12:41:32 23773] ERROR (SrvDaemon:356) Exception starting
xend ((38, 'Function not implemented'))
Traceback (most recent call last):
342, in run
xinfo = xc.xeninfo()
Error: (38, 'Function not implemented')
It looks like a xend restart can't happen from 3->4, so perhaps just
handle the upgrade scenario gracefully? xend starts fine after a
reboot, but we shouldn't kick out in %post.
4) grub not properly updated
I'm not sure why or if this is a "kernel-xen" problem, but grub got
hosed. I previously had a kernel and xen kernel installed, each with
grub entries. I updated the kernel and xen with "rpm -Uhv" and:
a) reboot hung on the 'stand by for reboot' message.
b) grub.conf got overwritten with only the new entry for the new
xen-aware kernel, but as a standard kernel entry, not with 'kernel
/xen.gz' and 'module' for kernel & initramfs.
c) the other kernels got erased
d) grub dropped into its shell on reboot
I brought the kernel up as a regular kernel (just from muscle-memory, I
forgot to kernel xen, but nice that it works both ways) and ran dracut
and grub-install, hand-fixed menu.lst and then all was well.
5) qemu-dm moved from /usr/lib64 to /usr/lib - dunno if this is correct,
if I have the right version, etc.
------ end of issues ------
Those said, my dom0 appears to be up fine now. I just installed a
Nexenta DomU on hvm successfully, working on getting pvm working. Thanks
for all the good work.
Bill McGonigle, Owner
BFC Computing, LLC
Email, IM, VOIP: bill(a)bfccomputing.com
Social networks: bill_mcgonigle/bill.mcgonigle
using F12 x86_64 + rawvirt repo I'm experiencing this kind of problem inside
vm is a rh el 5.5 x86 os
vm has defined default virtual network (nat; tried also creating a new
virtuial network nat-to-eth0; I only have eth0 on my host)
vm with virtio network card defined (tried also with rtl8139 model)
during boot of vm in linux: prompt I type:
I can see in messages of host (and in console 3 of guest) that it gets an
ip, but then I get error about unable to retrieve ks file...
on the host where I have httpd running if I run:
I can see dnsmasq process for both default virbr0 and the other one (bounded
to 192.168.101.x network, and accordingly changing the ks line in guest of
Tried both to use NetworkManager and standard network init script to manage
my network card at host side but with the same results...
I tried both of them because when vnet started I got an error from
NetworkManger in messages about "unknown device (or dirver), ignoring..." so
I presumed it was part of the problem...
About network card definition in guest: when I try rtl8139 model, I first
delete the other one and then create a new one, so the guest has always only
one network card defined.
I noticed that with virtio model in guest I have a "no link" message in
console 3 but AFTER it successfully got an ip via dhcp by dnsmaq related
Instead with rtl8139 I get the link up also after the dhcp get of ip, but
the same way I get the error about not being able to get the kickstart
Did I miss anything? I seem to remember in the past I was able to do so...
Thanks in advance
My objective is to have four FC12 VMs, with preferable Xen as the
So far, I installed a vanilla FC12 from the Live-CD, and I got KVM
working. I used virt-manager to install a FC12 VM from that same
Live-CD, but that failed. It blocks with message : XT4-fs error (device
dm-0): __ext4_get_inode_loc: unable to read inode block - inode=29634,
Concerning Xen, and as part as "yum install xen-kernel xen
virt-manager" , I got in the /boot directory the files xen-3.4.2.gz and
xen-syms-3.4.2. However I have no idea what to put in /etc/grub.conf to
boot this kernel.
I hope that when I get the xen kernel booted that I will be able to
install a FC12 VM from the Live-CD.
Any advice on above most welcome.