Bug 797835

If you force install an RPM you'd better know what you are doing and be prepared to fix any resulting problems.  This is NOTABUG.

Hi,
This time i use "yum downgrade" to test this.and find more issues:

Issue 1:

Steps:
#rpm -qa|grep libvirt
libvirt-0.9.10-10.el6.x86_64
libvirt-client-0.9.10-10.el6.x86_64
libvirt-python-0.9.10-10.el6.x86_64
# yum downgrade libvirt libvirt-client libvirt-python
...
 Installing : libvirt-client-0.9.4-23.el6.x86_64                               
          1/6 
  Installing : libvirt-0.9.4-23.el6.x86_64                                     
           2/6 
Non-fatal POSTIN scriptlet failure in rpm package libvirt-0.9.4-23.el6.x86_64
warning: /etc/libvirt/libvirtd.conf created as
/etc/libvirt/libvirtd.conf.rpmnew
warning: /etc/libvirt/qemu.conf created as /etc/libvirt/qemu.conf.rpmnew
warning: /etc/sysconfig/libvirtd created as /etc/sysconfig/libvirtd.rpmnew
warning: %post(libvirt-0.9.4-23.el6.x86_64) scriptlet failed, exit status 1
  Installing : libvirt-python-0.9.4-23.el6.x86_64                              
           3/6 
  Cleanup    : libvirt-python-0.9.10-10.el6.x86_64                             
           4/6 
  Cleanup    : libvirt-0.9.10-10.el6.x86_64                                    
           5/6 
  Cleanup    : libvirt-client-0.9.10-10.el6.x86_64                             
           6/6 
...
# service libvirtd status
libvirtd is stopped           ------> should running.
# virsh list
error: Failed to reconnect to the hypervisor
error: no valid connection
error: Failed to connect socket to '/var/run/libvirt/libvirt-sock': No such
file or directory
# /etc/init.d/libvirtd restart
Stopping libvirtd daemon:                                  [FAILED]
Starting libvirtd daemon:                                  [  OK  ]

Issue 2:
I defined several guests when i use libvirt-0.9.10-10 and guest xml include usb
controller by default like:
...
    <controller type='usb' index='0'>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x01'
function='0x2'/>
    </controller>
...
# virsh list --all
 Id    Name                           State
----------------------------------------------------
 -     mig13                          shut off
 -     mig14                          shut off

# yum downgrade libvirt libvirt-client libvirt-python
# service libvirtd restart 
# virsh list --all
 Id Name                 State
----------------------------------   ---->guests disappear

# rpm -Uvh libvirt*0.9.10-10*
# virsh list --all
 Id    Name                           State
----------------------------------------------------
 -     mig13                          shut off
 -     mig14                          shut off   ---->guests come back

I'm not sure why downgrading libvirt stopped the deamon instead of restarting it (it probably failed to start for some reason) but AFAIK downgrading libvirt is not supported in general. If you define some domains with the new libvirt, their XML configuration may contain things that older libvirt doesn't understand (controller type='usb' in this case). Thus the older libvirt will fail to parse those XMLs and they will be ignored.

BTW, there's a related discussion on libvirt-list: https://www.redhat.com/archives/libvir-list/2012-April/msg00352.html

I've opened Bug 815503 to deal specifically with the problem of migrating from a RHEL6.3 host to a RHEL6.2 host - that problem has the same cause, but is more serious due to the absence of any reasonable workaround.

Can we reopen this bug to track the downgrade issue until there is an explicit solution in libvir-list@ ?

There isn't any guarantee that downgrading libvirt will produce defined results, so this isn't a bug even when the downgrade is done with yum.  The problem appears to be that the old vesion doesn't accept some of the XML added by the new version, and there's no way to fix that.