Ubuntu
libvirt package

ISST-LTE:Ubuntu16.04.2KVM: Migration of guest fails between Ubuntu16.04.2 KVM hosts with error: load of migration failed: Invalid argument

Bug #1654150 reported by bugproxy on 2017-01-05

This bug affects 1 person

Affects		Status	Importance	Assigned to	Milestone
	libvirt (Ubuntu)	Fix Released	Medium	Unassigned

Bug Description

== Comment: #0 - Manjunatha H R <email address hidden> - 2016-12-26 05:45:38 ==
Migrating Ubuntu16.04.2 guest between two Ubuntu16.04.2 KVM hosts fails :

Migration error:
--------------------
root@lotkvm:~# virsh migrate --live --domain lotg1 qemu+ssh://10.33.32.222/system --verbose --undefinesource --persistent --timeout 30 --copy-storage-all --unsafe 2>&1
error: monitor socket did not show up: No such file or directory

OR sometimes it fails with following error:
---------------------------
Linux lotkvm 4.8.0-32-generic #34~16.04.1-Ubuntu SMP Tue Dec 13 17:01:57 UTC 2016 ppc64le ppc64le ppc64le GNU/Linux
root@lotkvm:~# virsh migrate --live --domain lotg1 qemu+ssh://10.33.32.222/system --verbose --undefinesource --persistent --copy-storage-all --timeout 60 --unsafe 2>&1

Migration: [100 %]error: internal error: early end of file from monitor, possible problem: 2016-12-26T11:09:42.030260Z qemu-system-ppc64: Unknown savevm section or instance 'pci@800000020000000:06.0/ohci' 0
2016-12-26T11:09:42.031574Z qemu-system-ppc64: load of migration failed: Invalid argument

Qemu logs show following traces:
---------------------
2016-12-26T09:35:35.709182Z qemu-system-ppc64: Unknown savevm section or instance 'pci@800000020000000:06.0/ohci' 0
2016-12-26T09:35:35.710089Z qemu-system-ppc64: load of migration failed: Invalid argument

Host build:
--------------
uname -a
Linux lotkvm 4.8.0-32-generic #34~16.04.1-Ubuntu SMP Tue Dec 13 17:01:57 UTC 2016 ppc64le ppc64le ppc64le GNU/Linux

Guest build:
----------------
# uname -a
Linux lotg1 4.8.0-32-generic #34~16.04.1-Ubuntu SMP Tue Dec 13 17:01:57 UTC 2016 ppc64le ppc64le ppc64le GNU/Linux

Issue is seen in shared storage and non-shared storages scenaris (i.e with or wothout --copy-storage-all option).

== Comment: #8 - Shivaprasad G. Bhat <email address hidden> - 2016-12-27 03:28:16 ==
This is a known issue and the patch proposed to fix the issue was rejected upstream and later decided to live with a workaround as discussed in https://www.redhat.com/archives/libvir-list/2016-July/msg01055.html and https://bugzilla.redhat.com/show_bug.cgi?id=1357468

Let me know if the workarounds suggested works or not, if not may be we need to pull some patches.

== Comment: #9 - Manjunatha H R <email address hidden> - 2016-12-27 05:52:01 ==
(In reply to comment #8)
> This is a known issue and the patch proposed to fix the issue was rejected
> upstream and later decided to live with a workaround as discussed in
> https://www.redhat.com/archives/libvir-list/2016-July/msg01055.html and
> https://bugzilla.redhat.com/show_bug.cgi?id=1357468
>
> Let me know if the workarounds suggested works or not, if not may be we need
> to pull some patches.

Thanks Shiva !!

With around of adding : model='pci-ohci' to usb controller helps to resolve migration error.

Before applying workaround:
----------------
root@lotkvm:~# virsh dumpxml lotg9|grep usb
<controller type='usb' index='0'>
<alias name='usb'/>
root@lotkvm:~#

root@lotkvm:~# virsh migrate --live --domain lotg9 qemu+ssh://10.33.32.222/system --verbose --undefinesource --persistent --timeout 60 --unsafe 2>&1
Migration: [100 %]error: operation failed: job: unexpectedly failed

After applying work around:
-------------------------
root@lotkvm:~# virsh dumpxml lotg9|grep usb
<controller type='usb' index='0' model='pci-ohci'>
<alias name='usb'/>
root@lotkvm:~#

root@lotkvm:~# virsh migrate --live --domain lotg9 qemu+ssh://10.33.32.222/system --verbose --undefinesource --persistent --timeout 60 --unsafe 2>&1
Migration: [100 %]
root@lotkvm:~#

Does this need to be documented in Ubuntu release notes ?

Thanks,
Manju

== Comment: #11 - Shivaprasad G. Bhat <email address hidden> - 2017-01-02 02:31:00 ==
Hi Manju, Chandan,

Yes, I think this should be documented as its likely to happen and because this is seen only on PPC without a documentation it is hard to figure out why its happening.

Thanks,
Shiva

Tags: