thr3ads.net - Xen users - Xen live migration + Openvswitch give longer downtime [May 2013]

If this information is useful, please help other people find it:
Share via:

agya naila

2013-May-02 10:02 UTC

Xen live migration + Openvswitch give longer downtime

Hi All,

I have setup my Xen 4.2.1 from sources in Ubuntu 12.04 as domu with XM
toolstack. I used Opensuse 12.3 as the DomU and create the bridge using
Openvswitch. Firstly I hacked the vif-bridge of Xen in order to support
Openvswitch since I am using xm tollstack. I am able to to run the DomU and
Live migration using openvswitch. However, I got longer downtime (8-15s) as
the result. Using basic bridge from brctl-utils the downtime is in ms. I
see the log it seem VM spent a lot of time waiting for network resources.

Since I am not sure my code hacking in vif-script have the bad impact then
I decided to use brcompat module to be able using brctl command and restore
vif-bridge back to original. I have configured and test the scenarios using
brcompat module, its improved by still the down time is too long. I got
5-6s downtime when failover occurs on my scenarios. The problem still
remains the same, the VM spent a lot of time waiting for the resource.

[2013-05-02 11:32:23 1625] DEBUG (XendDomainInfo:3015)
XendDomainInfo.completeRestore
[2013-05-02 11:32:23 1625] DEBUG (XendDomainInfo:1795) Storing domain
details: {''console/ring-ref'': ''357365'',
''image/entry'':
''18446744071562076160'', ''console/port'':
''2'', ''store/ring-ref'':
''357366'',
''image/loader'': ''generic'',
''vm'':
''/vm/26a013d3-fb9e-481a-0a49-6dabcfcb261a'',
''control/platform-feature-multiprocessor-suspend'':
''1'',
''image/features/auto-translated-physmap'':
''1'', ''image/guest-os'':
''linux'',
''image/features/writable-descriptor-tables'':
''1'', ''image/virt-base'':
''18446744071562067968'', ''memory/target'':
''1048576'', ''image/guest-version'':
''2.6'',
''image/features/supervisor-mode-kernel'':
''1'',
''control/platform-feature-xs_reset_watches'':
''1'', ''description'': '''',
''console/limit'': ''1048576'',
''image/paddr-offset'': ''0'',
''image/hypercall-page'':
''18446744071562080256'',
''image/suspend-cancel'':
''1'', ''cpu/0/availability'':
''online'',
''image/features/writable-page-tables'':
''1'', ''console/type'':
''xenconsoled'', ''image/supported-features'':
''2063'',
''image/mod-start-pfn'': ''1'',
''name'': ''opensuse'',
''domid'': ''6'',
''image/xen-version'': ''xen-3.0'',
''image/init-p2m'': ''18446719884453740544'',
''store/port'': ''1''}
[2013-05-02 11:32:23 1625] DEBUG (XendDomainInfo:3027)
XendDomainInfo.completeRestore done
[2013-05-02 11:32:23 1625] DEBUG (XendDomainInfo:3028) GRE Tunnel Switching
OK
[2013-05-02 11:32:24 1625] DEBUG (XendDomainInfo:1881)
XendDomainInfo.handleShutdownWatch
[2013-05-02 11:32:29 1625] DEBUG (DevController:139) Waiting for devices
tap2.
[2013-05-02 11:32:29 1625] DEBUG (DevController:139) Waiting for
devices vif.
[2013-05-02 11:32:29 1625] DEBUG (DevController:144) Waiting for 0.
[2013-05-02 11:32:29 1625] DEBUG
(DevController:628) hotplugStatusCallback
/local/domain/0/backend/vif/6/0/hotplug-status.
[2013-05-02 11:32:29 1625] DEBUG
(DevController:642) hotplugStatusCallback 1.
[2013-05-02 11:32:29 1625] DEBUG (DevController:139) Waiting for
devices vkbd.
[2013-05-02 11:32:29 1625] DEBUG (DevController:139) Waiting
for devices ioports.

As we can see above xend complete restore at 11:32:24 but got the result
11:32:29.

Anybody have experince on it? any clues would be very appreciated.

Cheers,

Agya


_______________________________________________
Xen-users mailing list
Xen-users@lists.xen.org
http://lists.xen.org/xen-users

agya naila

2013-May-02 11:23 UTC

head link

Fwd: Xen live migration + Openvswitch give longer downtime

Hi All,

I have setup my Xen 4.2.1 from sources in Ubuntu 12.04 as domu with XM
toolstack. I used Opensuse 12.3 as the DomU and create the bridge using
Openvswitch. Firstly I hacked the vif-bridge of Xen in order to support
Openvswitch since I am using xm tollstack. I am able to to run the DomU and
Live migration using openvswitch. However, I got longer downtime (8-15s) as
the result. Using basic bridge from brctl-utils the downtime is in ms. I
see the log it seem VM spent a lot of time waiting for network resources.

Since I am not sure my code hacking in vif-script have the bad impact then
I decided to use brcompat module to be able using brctl command and restore
vif-bridge back to original. I have configured and test the scenarios using
brcompat module, its improved by still the down time is too long. I got
5-6s downtime when failover occurs on my scenarios. The problem still
remains the same, the VM spent a lot of time waiting for the resource.

[2013-05-02 11:32:23 1625] DEBUG (XendDomainInfo:3015)
XendDomainInfo.completeRestore
[2013-05-02 11:32:23 1625] DEBUG (XendDomainInfo:1795) Storing domain
details: {''console/ring-ref'': ''357365'',
''image/entry'':
''18446744071562076160'', ''console/port'':
''2'', ''store/ring-ref'':
''357366'',
''image/loader'': ''generic'',
''vm'':
''/vm/26a013d3-fb9e-481a-0a49-6dabcfcb261a'',
''control/platform-feature-multiprocessor-suspend'':
''1'',
''image/features/auto-translated-physmap'':
''1'', ''image/guest-os'':
''linux'',
''image/features/writable-descriptor-tables'':
''1'', ''image/virt-base'':
''18446744071562067968'', ''memory/target'':
''1048576'', ''image/guest-version'':
''2.6'',
''image/features/supervisor-mode-kernel'':
''1'',
''control/platform-feature-xs_reset_watches'':
''1'', ''description'': '''',
''console/limit'': ''1048576'',
''image/paddr-offset'': ''0'',
''image/hypercall-page'':
''18446744071562080256'',
''image/suspend-cancel'':
''1'', ''cpu/0/availability'':
''online'',
''image/features/writable-page-tables'':
''1'', ''console/type'':
''xenconsoled'', ''image/supported-features'':
''2063'',
''image/mod-start-pfn'': ''1'',
''name'': ''opensuse'',
''domid'': ''6'',
''image/xen-version'': ''xen-3.0'',
''image/init-p2m'': ''18446719884453740544'',
''store/port'': ''1''}
[2013-05-02 11:32:23 1625] DEBUG (XendDomainInfo:3027)
XendDomainInfo.completeRestore done
[2013-05-02 11:32:23 1625] DEBUG (XendDomainInfo:3028) GRE Tunnel Switching
OK
[2013-05-02 11:32:24 1625] DEBUG (XendDomainInfo:1881)
XendDomainInfo.handleShutdownWatch
[2013-05-02 11:32:29 1625] DEBUG (DevController:139) Waiting for devices
tap2.
[2013-05-02 11:32:29 1625] DEBUG (DevController:139) Waiting for
devices vif.
[2013-05-02 11:32:29 1625] DEBUG (DevController:144) Waiting for 0.
[2013-05-02 11:32:29 1625] DEBUG
(DevController:628) hotplugStatusCallback
/local/domain/0/backend/vif/6/0/hotplug-status.
[2013-05-02 11:32:29 1625] DEBUG
(DevController:642) hotplugStatusCallback 1.
[2013-05-02 11:32:29 1625] DEBUG (DevController:139) Waiting for
devices vkbd.
[2013-05-02 11:32:29 1625] DEBUG (DevController:139) Waiting
for devices ioports.

As we can see above xend complete restore at 11:32:24 but got the result
11:32:29.

Anybody have experince on it? any clues would be very appreciated.

Cheers,

Agya


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Xen users - May 2013 - Xen live migration + Openvswitch give longer downtime

Xen live migration + Openvswitch give longer downtime

Fwd: Xen live migration + Openvswitch give longer downtime