Bug 676563

It's caused by the ancient qemu doesn't support "qemu -M ?", which is used to probe the machine types to construct guest capabilities, and the failure to contruct the guest capabilities then cause qemu state driver initialization failed.

Is libvirt crashing or just exiting normally?

not crashing(no coredump), but exit alnormally, it didn't release subsys lock on shutdown event, useful logs pasted following:


Feb 10 00:33:38 dhcp-65-132 libvirtd: 00:33:38.733: 5342: error : virStateInitialize:1022 : Initialization of QEMU state driver failed
Feb 10 00:33:38 dhcp-65-132 kernel: lo: Disabled Privacy Extensions
Feb 10 00:33:38 dhcp-65-132 libvirtd: 00:33:38.749: 5342: info : virDomainLoadAllConfigs:8248 : Scanning for configs in /etc/libvirt/lxc
Feb 10 00:33:38 dhcp-65-132 libvirtd: 00:33:38.755: 5342: error : main:3305 : Driver state initialization failed
Feb 10 00:33:38 dhcp-65-132 libvirtd: 00:33:38.756: 5343: warning : qemudDispatchSignalEvent:403 : Shutting down on signal 3

Sounds like we should simply be disabling the qemu driver.  What do you think?

No, the QEMU driver is expected to successfully start regardless of what QEMU version is installed, or indeed even if no QEMU is installed yet.  We need to fix the code which deals with 'qemu -M ?' since libvirt in general needs to support QEMUs without this feature.

Looks like this is a regression in commit 2211518:
2011-01-12 16:26 Eric Blake         * qemu: convert capabilities to use virCommand

Prior to that point, qemuCapsProbeMachineTypes silently ignored failure of 'qemu -M ?'; therefore, the obvious fix would be to pass a dummy status instead of NULL to virCommandRun.

Proposed upstream patch:
https://www.redhat.com/archives/libvir-list/2011-February/msg00468.html

Moving to POST:
http://post-office.corp.redhat.com/archives/rhvirt-patches/2011-February/msg00967.html

The bug is reproduced with libvirt-0.8.7-7.el6.x86_64
- kernel-2.6.32-113.el6.x86_64
- qemu-kvm-0.12.1.2-2.145.el6.x86_64

The same test step and actual result with Description

Additional patch needed to fix this issue was sent for review: http://post-office.corp.redhat.com/archives/rhvirt-patches/2011-April/msg00082.html

Verified this bug pass with libvirt-0.8.7-16.el6.x86_64
1. # virsh list --all
 Id Name                 State
----------------------------------
  - a10                  shut off
  - a11                  shut off
  - a12                  shut off
2. # rpm -ivh qemu-0.6.0-2.x86_64.rpm 
Preparing...                ########################################### [100%]
   1:qemu                   ########################################### [100%]
3. # virsh list --all
 Id Name                 State
----------------------------------
  - a10                  shut off
  - a11                  shut off
  - a12                  shut off
4. # for i in {1..10}; do service libvirtd restart; done
Stopping libvirtd daemon:                                  [  OK  ]
Starting libvirtd daemon:                                  [  OK  ]
Stopping libvirtd daemon:                                  [  OK  ]
...
5. # virsh list --all
 Id Name                 State
----------------------------------
  - a10                  shut off
  - a11                  shut off
  - a12                  shut off
6. # service libvirtd status
libvirtd (pid  8332) is running...

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHBA-2011-0596.html