[ovirt-users] live migration with openvswitch

Fri Sep 16 12:41:32 UTC 2016

> On 16 Sep 2016, at 14:36, Michal Skrivanek <michal.skrivanek at redhat.com> wrote:
> 
> 
>> On 15 Sep 2016, at 21:46, Edward Haas <ehaas at redhat.com <mailto:ehaas at redhat.com>> wrote:
>> 
>> 
>> 
>> On Thu, Sep 15, 2016 at 1:30 PM, Michal Skrivanek <michal.skrivanek at redhat.com <mailto:michal.skrivanek at redhat.com>> wrote:
>> 
>> > On 15 Sep 2016, at 10:11, Dan Kenigsberg <danken at redhat.com <mailto:danken at redhat.com>> wrote:
>> >
>> > On Wed, Sep 14, 2016 at 03:04:14PM +0200, Michal Skrivanek wrote:
>> >>
>> >>> On 09 Sep 2016, at 13:09, Edward Haas <ehaas at redhat.com <mailto:ehaas at redhat.com>> wrote:
>> >>>
>> >>>
>> >>>
>> >>> On Thu, Sep 8, 2016 at 11:27 AM, Pavel Levshin <lpk at 581.spb.su <mailto:lpk at 581.spb.su> <mailto:lpk at 581.spb.su <mailto:lpk at 581.spb.su>>> wrote:
>> >>> Hi.
>> >>>
>> >>> I'm trying to learn Ovirt 4 and have a problem with it.
>> >>>
>> >>> My cluster consists of 3 nodes. I use Openvswitch for network connectivity. I have a HostedEngine and one additional VM in the cluster.
>> >>>
>> >>> When I try to migrate the VM to another node, it fails. From vdsm and libvirtd logs I see that proper network interface on destination node cannot be found. Libvirt tries to find Openvswitch bridge with name like "vdsmbr_AOYiPtcT". It exists on source node, but it is unique on every node, because it contains random part. Additionally, it changes on every reboot.
>> >>>
>> >>> How this is supposed to work?
>> >>>
>> >>> --
>> >>> Pavel Levshin
>> >>>
>> >>>
>> >>>
>> >>> Hi Pavel,
>> >>>
>> >>> VM migration is supported on the master branch, however it has not been ported to 4.0 yet.
>> >>
>> >>> You can either build VDSM from source (from master branch) or try to apply this patch on what you have:
>> >>> https://gerrit.ovirt.org/#/c/59645 <https://gerrit.ovirt.org/#/c/59645> <https://gerrit.ovirt.org/#/c/59645 <https://gerrit.ovirt.org/#/c/59645>>
>> >>
>> >> That’s quite a horrible solution right now. I certainly would not like to see it in 4.0 (given the hacks around display).
>> 
>> What is horrible exactly?
>> It's not too late to propose other solutions.
> 
> if OVS is the next great feature it should fit into the code accordingly. I.e. using hooks only when it’s absolutely necessary and as a temporary measure only until the respective proper RFEs are implemented and available. E.g. when there is a libvirt support missing we can add a qemu command line parameter ourselves bypassing libvirt but we always should have a clear plan (i.e. a bug) to move away from there as soon as the support is there(requested back then when we went with the hack)
> 
> Such things should be reviewed as soon as we get to a similar area, so while modifying libvirt-hook.sh we can see the original reason for the hook is not valid anymore as everything is addressed and the hacky code should have been removed
> It was easy to see that because there is a clear comment about dependent bugs and issues (though missed by all the reviewers, unfortunately!)
> Your new code doesn’t have anything like that and I have no idea what kind of API or behavior we actually need, whether appropriate requests has been filed on e.g. libvirt. That makes it very hard to revisit in the future by the next random person.
> 
>> 
>> Display uses libvirt to resolve a network name to an IP address for it to bound to. But that works only for linux bridges.
>> That is limiting, especially now that we do not have a Linux bridge, but something else.
> 
> that’s ok, whatever needs to be done. But then please make sure you’re not breaking existing features, at least again not without a plan(==bug) to fix it.
> 
>> 
>> >> Do we have a bug/plan to improve it?
>> >
>> > We have Bug 1362495 - [OVS] - Add support for live migration
>> > to track that.

oh, and yes, that’s exactly the tracking I wanted to make sure exists. There’s just no link in the gerrit commit itself so I didn’t find it (but I wasn’t really looking hard either;-)

Thanks,
michal

>> >
>> > I'm afraid that we are not yet ready to backport it to 4.0 - we found
>> > out that as it is, it break migration for vmfex and external network
>> > providers; it also breaks when a buggy Engine db does not send a
>> > displayNetwork. But we plan to fix these issues quite soon.
> 
> which “buggy” engine? There were changes in parameters, most of these issues are not relevant anymore since we ditched <3.6 though.
> Again it’s ok as long as it is clearly mentioned like "3.6 engine sends it in such and such parameter, we can drop it once we support 4.0+"
> 
>> >
>> > The hacks arround display are an actual imporovement. For "legacy"
>> > switchType, we maintain an on-host libvirt-side database of all networks
>> > only to keep libvirt happy. Having a database copy has all the known
>> > troubles of mismatches and being out of sync. For "ovs" switchType, we
>> > do not (we don't use a bridge, but a port group so there's no natural
>> > way to define our network in libvirt). Modifying the listening address
>> > on destination is the flexible and quick way to do it - I wish we had
>> > the libvirt migrate hook years ago.
>> 
>> doesn’t it prevent seamless virti-viewer console connection?
>> 
>> The end result is the same, we listen on the address of a specific network.
>> Previously it contained a network name and libvirt converted it to the correct IP it should bind to, now vdsm resolves it.
> 
> so did we request libvirt to be able to do that itself or requested an API to do that cleanly without mangling xml on the fly?
> 
>> 
>> also the “TODO” in the code about multiple graphics is worrying (we fully support it and are considering to make it a default)
>> 
>> Supported where? virt networking code in VDSM which creates an interface for domxml does not support it at the moment.
>> Or am I missing something?
> 
> supported in oVirt. Multiple graphics VM is a regular oVirt feature. The hook doesn’t handle it correctly hence I suppose the migration would not really work correctly for these VMs.
> Again, that might be ok temporarily with a clear documentation and a plan to fix it.
> 
>> 
>> If we have an idea of what API would work well? we should raise or contribute that to libvirt. Surely it takes time but it is the only way how to improve the code eventually.
>> 
>> If using libvirt can allow us to drop some persisted data and logic from vdsm, then it makes sense, but I do not think this is the case.
>> As it stands today, depending on libvirt persisted data is limiting us, at least in the networking area. I also do not see the advantage of using it as a DB.
> 
> I’m not saying you should depend on libvirt data. But you should let libvirt community know if it is a useless feature, use a clean way how not to use that feature(if it gets in your way), ideally discuss and propose how such a feature can be improved so oVirt and all other virtualization projects can benefit from that. If it’s a truly useless feature I’m sure libvirt would be happy to drop maintaining it
> 
> Thanks,
> michal
> 
>> 
>> Thaks,
>> Edy.
>>  
>> 
>> Thanks,
>> michal
>> 
>> 
>> _______________________________________________
>> Users mailing list
>> Users at ovirt.org <mailto:Users at ovirt.org>
>> http://lists.ovirt.org/mailman/listinfo/users
> 
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20160916/1da5ac56/attachment-0001.html>