DescriptionIdith Tal-Kohen
2013-09-03 14:38:43 UTC
+++ This bug was initially created as a clone of Bug #1001001 +++
Description of problem:
Running vm in "runOnce" mode prevent to unlink vnic from vm with error message:
"Error while executing action Edit VM Interface properties: Unexpected exception"
Version-Release number of selected component (if applicable):
is11
host-vdsm-4.12.0-61.git8178ec2.el6ev, libvirt-0.10.2-18.el6_4.9
How reproducible:
Always
Steps to Reproduce:
1. instal vm on host with network and disk
2. start vm in runOnce mode
3. Try to change link state of one of vnics to down
Actual results:
error message:
"Error while executing action Edit VM Interface properties: Unexpected exception"
Expected results:
vnic change link state to down
Additional info:
In run mode all works fine, also the same thing is relevant for sf20
boot order hd, cdrom, pixe
--- Additional comment from Dan Kenigsberg on 2013-08-27 16:31:08 EDT ---
domxml is created with
<interface type="bridge">
<mac address="00:1a:4a:23:a1:a9"/>
<model type="virtio"/>
<source bridge="rhevm"/>
<filterref filter="vdsm-no-mac-spoofing"/>
<link state="up"/>
<boot order="2"/>
</interface>
but calling updateDevice with
<interface type="bridge">
<address bus="0x00" domain="0x0000" function="0x0" slot="0x03" type="pci"/>
<mac address="00:1a:4a:23:a1:a9"/>
<model type="virtio"/>
<source bridge="rhevm"/>
<filterref filter="vdsm-no-mac-spoofing"/>
<link state="down"/>
<boot order="2"/>
</interface>
fails with
libvirtError: this function is not supported by the connection driver: cannot modify network device boot index setting
I am guessing that vdsm should simply not pass the <boot order="2"/> element on updateDevice. The interesting question is: when was this introduced?
--- Additional comment from Dan Kenigsberg on 2013-08-27 16:51:30 EDT ---
--- Additional comment from Dan Kenigsberg on 2013-09-02 08:11:42 EDT ---
libvirt bug 895294 should have fixed this issue in libvirt-0.10.2-19.el6. Vdsm for el6 should only require this libvirt version.
Artyom, would you verify that with the proper libvirt, the issue disappears?
--- Additional comment from Dan Kenigsberg on 2013-09-02 17:44:27 EDT ---
Until a libvirt-side fix is ready, we may want to hack it on vdsm side with the like of the yet-untested http://gerrit.ovirt.org/18796 .