Created attachment 715892 [details] vdsm, engine, libvirt logs Description of problem: Trying to shutdown a vm using using ShutdownVM command fails due to: Thread-188981::ERROR::2013-03-25 09:21:05,057::guestIF::269::vm.Vm::(desktopShutdown) vmId=`3bfd8c39-acc2-4cb3-9e6a-36b4b386e2f4`::desktopShutdown failed Traceback (most recent call last): File "/usr/share/vdsm/guestIF.py", line 267, in desktopShutdown self._forward('shutdown', {'timeout': timeout, 'message': msg}) File "/usr/share/vdsm/guestIF.py", line 135, in _forward self._sock.send(message) File "/usr/lib64/python2.6/socket.py", line 167, in _dummy raise error(EBADF, 'Bad file descriptor') error: [Errno 9] Bad file descriptor in vdsm. The scenario used is described below Version-Release number of selected component (if applicable): vdsm-4.10.2-11.0.el6ev.x86_64 How reproducible: 100% Steps to Reproduce: 1. Have 2 vms with guest-agent and RHEL installed 2. Plug in the same shareable disk (in addition to each VMs OS disk) to both vms 3. Power vms on 4. Deactivate disk from one of the vms 5. attempt to shutdown vm that still has disk attached Actual results: Shutdown fails - command is never sent to guest-agent Expected results: VM should shutdown gracefully Additional info:
This sounds related to the funny call to Vm._cleanup() on hotunplug: http://gerrit.ovirt.org/12564 Could you comment-out that call in vdsm, and retry the scenario?
On a disk hot unplug VDSM closes the communication channel to the guest agent and this is a bug, that's for sure (reproduced and will be fixed) But Gadi, are you sure that VDSM also closes the channel to the agent on the VM, from where a disk is *not* detached? I can not reproduce this particular case in both cases (2 VMs on the same hypervizor and 2 VMs on 2 hypervizors, detach disc from one VM and shut down other VM — all works OK for the second VM)
(In reply to comment #4) > On a disk hot unplug VDSM closes the communication channel to the guest > agent and this is a bug, that's for sure (reproduced and will be fixed) > > But Gadi, are you sure that VDSM also closes the channel to the agent on the > VM, from where a disk is *not* detached? I can not reproduce this particular > case in both cases (2 VMs on the same hypervizor and 2 VMs on 2 hypervizors, > detach disc from one VM and shut down other VM — all works OK for the second > VM) You are right Peter - the scenario actually shuts down the VM that had the disk unplugged first.
Great, thanks.
Merged as ovirt:435568c89df238f24e39b74d2441f925581d3c87 Preparing the backport
Merged as rhev:4b7a7b882d965c3b35f7f44cb42422ca0be671d6
Verified on SI13.1 - vdsm-4.10.2-15.0.el6ev.x86_64 - ran automated test that reproduced scenario and shutdownVm worked successfully.
(In reply to comment #10) > Verified on SF13.1 - vdsm-4.10.2-15.0.el6ev.x86_64 - ran automated test > that reproduced scenario and shutdownVm worked successfully. ** SF13.1
Doc text provided, see the field
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. http://rhn.redhat.com/errata/RHSA-2013-0886.html