<div dir="ltr"><div>Many thanks.</div><div>Please keep in touch.</div><div>Ignazio<br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">Il giorno mer 29 apr 2020 alle ore 16:55 Sean Mooney <<a href="mailto:smooney@redhat.com">smooney@redhat.com</a>> ha scritto:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">so bing pragmatic i think the simplest path forward given my other patches have not laned<br>
in almost 2 years is to quickly add a workaround config option to disable mulitple port bindign<br>
which we can backport and then we can try and work on the actual fix after.<br>
acording to <a href="https://bugs.launchpad.net/neutron/+bug/1815989" rel="noreferrer" target="_blank">https://bugs.launchpad.net/neutron/+bug/1815989</a> that shoudl serve as a workaround<br>
for thos that hav this issue but its a regression in functionality.<br>
<br>
i can create a patch that will do that in an hour or so and submit a followup DNM patch to enabel the<br>
workaound in one of the gate jobs that tests live migration.<br>
i have a meeting in 10 mins and need to finish the pacht im currently updating but ill submit a poc once that is done.<br>
<br>
im not sure if i will be able to spend time on the actul fix which i proposed last year but ill see what i can do.<br>
<br>
<br>
On Wed, 2020-04-29 at 16:37 +0200, Ignazio Cassano wrote:<br>
> PS<br>
> I have testing environment on queens,rocky and stein and I can make test<br>
> as you need.<br>
> Ignazio<br>
> <br>
> Il giorno mer 29 apr 2020 alle ore 16:19 Ignazio Cassano <<br>
> <a href="mailto:ignaziocassano@gmail.com" target="_blank">ignaziocassano@gmail.com</a>> ha scritto:<br>
> <br>
> > Hello Sean,<br>
> > the following is the configuration on my compute nodes:<br>
> > [root@podiscsivc-kvm01 network-scripts]# rpm -qa|grep libvirt<br>
> > libvirt-daemon-driver-storage-iscsi-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-kvm-4.5.0-33.el7.x86_64<br>
> > libvirt-libs-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-network-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-nodedev-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-storage-gluster-4.5.0-33.el7.x86_64<br>
> > libvirt-client-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-storage-core-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-storage-logical-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-secret-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-nwfilter-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-storage-scsi-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-storage-rbd-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-config-nwfilter-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-storage-disk-4.5.0-33.el7.x86_64<br>
> > libvirt-bash-completion-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-qemu-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-storage-4.5.0-33.el7.x86_64<br>
> > libvirt-python-4.5.0-1.el7.x86_64<br>
> > libvirt-daemon-driver-interface-4.5.0-33.el7.x86_64<br>
> > libvirt-daemon-driver-storage-mpath-4.5.0-33.el7.x86_64<br>
> > [root@podiscsivc-kvm01 network-scripts]# rpm -qa|grep qemu<br>
> > qemu-kvm-common-ev-2.12.0-44.1.el7_8.1.x86_64<br>
> > qemu-kvm-ev-2.12.0-44.1.el7_8.1.x86_64<br>
> > libvirt-daemon-driver-qemu-4.5.0-33.el7.x86_64<br>
> > centos-release-qemu-ev-1.0-4.el7.centos.noarch<br>
> > ipxe-roms-qemu-20180825-2.git133f4c.el7.noarch<br>
> > qemu-img-ev-2.12.0-44.1.el7_8.1.x86_64<br>
> > <br>
> > <br>
> > As far as firewall driver /etc/neutron/plugins/ml2/openvswitch_agent.ini:<br>
> > <br>
> > firewall_driver = iptables_hybrid<br>
> > <br>
> > I have same libvirt/qemu version on queens, on rocky and on stein testing<br>
> > environment and the<br>
> > same firewall driver.<br>
> > Live migration on provider network on queens works fine.<br>
> > It does not work fine on rocky and stein (vm lost connection after it is<br>
> > migrated and start to respond only when the vm send a network packet , for<br>
> > example when chrony pools the time server).<br>
> > <br>
> > Ignazio<br>
> > <br>
> > <br>
> > <br>
> > Il giorno mer 29 apr 2020 alle ore 14:36 Sean Mooney <<a href="mailto:smooney@redhat.com" target="_blank">smooney@redhat.com</a>><br>
> > ha scritto:<br>
> > <br>
> > > On Wed, 2020-04-29 at 10:39 +0200, Ignazio Cassano wrote:<br>
> > > > Hello, some updated about this issue.<br>
> > > > I read someone has got same issue as reported here:<br>
> > > > <br>
> > > > <a href="https://bugs.launchpad.net/neutron/+bug/1866139" rel="noreferrer" target="_blank">https://bugs.launchpad.net/neutron/+bug/1866139</a><br>
> > > > <br>
> > > > If you read the discussion, someone tells that the garp must be sent by<br>
> > > > qemu during live miration.<br>
> > > > If this is true, this means on rocky/stein the qemu/libvirt are bugged.<br>
> > > <br>
> > > it is not correct.<br>
> > > qemu/libvir thas alsway used RARP which predates GARP to serve as its mac<br>
> > > learning frames<br>
> > > instead <a href="https://en.wikipedia.org/wiki/Reverse_Address_Resolution_Protocol" rel="noreferrer" target="_blank">https://en.wikipedia.org/wiki/Reverse_Address_Resolution_Protocol</a><br>
> > > <a href="https://lists.gnu.org/archive/html/qemu-devel/2009-10/msg01457.html" rel="noreferrer" target="_blank">https://lists.gnu.org/archive/html/qemu-devel/2009-10/msg01457.html</a><br>
> > > however it looks like this was broken in 2016 in qemu 2.6.0<br>
> > > <a href="https://lists.gnu.org/archive/html/qemu-devel/2016-07/msg04645.html" rel="noreferrer" target="_blank">https://lists.gnu.org/archive/html/qemu-devel/2016-07/msg04645.html</a><br>
> > > but was fixed by<br>
> > > <a href="https://github.com/qemu/qemu/commit/ca1ee3d6b546e841a1b9db413eb8fa09f13a061b" rel="noreferrer" target="_blank">https://github.com/qemu/qemu/commit/ca1ee3d6b546e841a1b9db413eb8fa09f13a061b</a><br>
> > > can you confirm you are not using the broken 2.6.0 release and are using<br>
> > > 2.7 or newer or 2.4 and older.<br>
> > > <br>
> > > <br>
> > > > So I tried to use stein and rocky with the same version of libvirt/qemu<br>
> > > > packages I installed on queens (I updated compute and controllers node<br>
> > > <br>
> > > on<br>
> > > > queens for obtaining same libvirt/qemu version deployed on rocky and<br>
> > > <br>
> > > stein).<br>
> > > > <br>
> > > > On queens live migration on provider network continues to work fine.<br>
> > > > On rocky and stein not, so I think the issue is related to openstack<br>
> > > > components .<br>
> > > <br>
> > > on queens we have only a singel prot binding and nova blindly assumes<br>
> > > that the port binding details wont<br>
> > > change when it does a live migration and does not update the xml for the<br>
> > > netwrok interfaces.<br>
> > > <br>
> > > the port binding is updated after the migration is complete in<br>
> > > post_livemigration<br>
> > > in rocky+ neutron optionally uses the multiple port bindings flow to<br>
> > > prebind the port to the destiatnion<br>
> > > so it can update the xml if needed and if post copy live migration is<br>
> > > enable it will asyconsly activate teh dest port<br>
> > > binding before post_livemigration shortenting the downtime.<br>
> > > <br>
> > > if you are using the iptables firewall os-vif will have precreated the<br>
> > > ovs port and intermediate linux bridge before the<br>
> > > migration started which will allow neutron to wire it up (put it on the<br>
> > > correct vlan and install security groups) before<br>
> > > the vm completes the migraton.<br>
> > > <br>
> > > if you are using the ovs firewall os-vif still precreates teh ovs port<br>
> > > but libvirt deletes it and recreats it too.<br>
> > > as a result there is a race when using openvswitch firewall that can<br>
> > > result in the RARP packets being lost.<br>
> > > <br>
> > > > <br>
> > > > Best Regards<br>
> > > > Ignazio Cassano<br>
> > > > <br>
> > > > <br>
> > > > <br>
> > > > <br>
> > > > Il giorno lun 27 apr 2020 alle ore 19:50 Sean Mooney <<br>
> > > <br>
> > > <a href="mailto:smooney@redhat.com" target="_blank">smooney@redhat.com</a>><br>
> > > > ha scritto:<br>
> > > > <br>
> > > > > On Mon, 2020-04-27 at 18:19 +0200, Ignazio Cassano wrote:<br>
> > > > > > Hello, I have this problem with rocky or newer with iptables_hybrid<br>
> > > > > > firewall.<br>
> > > > > > So, can I solve using post copy live migration ???<br>
> > > > > <br>
> > > > > so this behavior has always been how nova worked but rocky the<br>
> > > > > <br>
> > > > > <br>
> > > <br>
> > > <a href="https://specs.openstack.org/openstack/nova-specs/specs/rocky/implemented/neutron-new-port-binding-api.html" rel="noreferrer" target="_blank">https://specs.openstack.org/openstack/nova-specs/specs/rocky/implemented/neutron-new-port-binding-api.html</a><br>
> > > > > spec intoduced teh ablity to shorten the outage by pre biding the<br>
> > > <br>
> > > port and<br>
> > > > > activating it when<br>
> > > > > the vm is resumed on the destiation host before we get to pos live<br>
> > > <br>
> > > migrate.<br>
> > > > > <br>
> > > > > this reduces the outage time although i cant be fully elimiated as<br>
> > > <br>
> > > some<br>
> > > > > level of packet loss is<br>
> > > > > always expected when you live migrate.<br>
> > > > > <br>
> > > > > so yes enabliy post copy live migration should help but be aware that<br>
> > > <br>
> > > if a<br>
> > > > > network partion happens<br>
> > > > > during a post copy live migration the vm will crash and need to be<br>
> > > > > restarted.<br>
> > > > > it is generally safe to use and will imporve the migration performace<br>
> > > <br>
> > > but<br>
> > > > > unlike pre copy migration if<br>
> > > > > the guess resumes on the dest and the mempry page has not been copied<br>
> > > <br>
> > > yet<br>
> > > > > then it must wait for it to be copied<br>
> > > > > and retrive it form the souce host. if the connection too the souce<br>
> > > <br>
> > > host<br>
> > > > > is intrupted then the vm cant<br>
> > > > > do that and the migration will fail and the instance will crash. if<br>
> > > <br>
> > > you<br>
> > > > > are using precopy migration<br>
> > > > > if there is a network partaion during the migration the migration will<br>
> > > > > fail but the instance will continue<br>
> > > > > to run on the source host.<br>
> > > > > <br>
> > > > > so while i would still recommend using it, i it just good to be aware<br>
> > > <br>
> > > of<br>
> > > > > that behavior change.<br>
> > > > > <br>
> > > > > > Thanks<br>
> > > > > > Ignazio<br>
> > > > > > <br>
> > > > > > Il Lun 27 Apr 2020, 17:57 Sean Mooney <<a href="mailto:smooney@redhat.com" target="_blank">smooney@redhat.com</a>> ha<br>
> > > <br>
> > > scritto:<br>
> > > > > > <br>
> > > > > > > On Mon, 2020-04-27 at 17:06 +0200, Ignazio Cassano wrote:<br>
> > > > > > > > Hello, I have a problem on stein neutron. When a vm migrate<br>
> > > <br>
> > > from one<br>
> > > > > <br>
> > > > > node<br>
> > > > > > > > to another I cannot ping it for several minutes. If in the vm I<br>
> > > <br>
> > > put a<br>
> > > > > > > > script that ping the gateway continously, the live migration<br>
> > > <br>
> > > works<br>
> > > > > <br>
> > > > > fine<br>
> > > > > > > <br>
> > > > > > > and<br>
> > > > > > > > I can ping it. Why this happens ? I read something about<br>
> > > <br>
> > > gratuitous<br>
> > > > > <br>
> > > > > arp.<br>
> > > > > > > <br>
> > > > > > > qemu does not use gratuitous arp but instead uses an older<br>
> > > <br>
> > > protocal<br>
> > > > > <br>
> > > > > called<br>
> > > > > > > RARP<br>
> > > > > > > to do mac address learning.<br>
> > > > > > > <br>
> > > > > > > what release of openstack are you using. and are you using<br>
> > > <br>
> > > iptables<br>
> > > > > > > firewall of openvswitch firewall.<br>
> > > > > > > <br>
> > > > > > > if you are using openvswtich there is is nothing we can do until<br>
> > > <br>
> > > we<br>
> > > > > > > finally delegate vif pluging to os-vif.<br>
> > > > > > > currently libvirt handels interface plugging for kernel ovs when<br>
> > > <br>
> > > using<br>
> > > > > <br>
> > > > > the<br>
> > > > > > > openvswitch firewall driver<br>
> > > > > > > <a href="https://review.opendev.org/#/c/602432/" rel="noreferrer" target="_blank">https://review.opendev.org/#/c/602432/</a> would adress that but it<br>
> > > <br>
> > > and<br>
> > > > > <br>
> > > > > the<br>
> > > > > > > neutron patch are<br>
> > > > > > > <a href="https://review.opendev.org/#/c/640258" rel="noreferrer" target="_blank">https://review.opendev.org/#/c/640258</a> rather out dated. while<br>
> > > <br>
> > > libvirt<br>
> > > > > <br>
> > > > > is<br>
> > > > > > > pluging the vif there will always be<br>
> > > > > > > a race condition where the RARP packets sent by qemu and then mac<br>
> > > > > <br>
> > > > > learning<br>
> > > > > > > packets will be lost.<br>
> > > > > > > <br>
> > > > > > > if you are using the iptables firewall and you have opnestack<br>
> > > <br>
> > > rock or<br>
> > > > > > > later then if you enable post copy live migration<br>
> > > > > > > it should reduce the downtime. in this conficution we do not have<br>
> > > <br>
> > > the<br>
> > > > > <br>
> > > > > race<br>
> > > > > > > betwen neutron and libvirt so the rarp<br>
> > > > > > > packets should not be lost.<br>
> > > > > > > <br>
> > > > > > > <br>
> > > > > > > > Please, help me ?<br>
> > > > > > > > Any workaround , please ?<br>
> > > > > > > > <br>
> > > > > > > > Best Regards<br>
> > > > > > > > Ignazio<br>
> > > > > > > <br>
> > > > > > > <br>
> > > > > <br>
> > > > > <br>
> > > <br>
> > > <br>
<br>
</blockquote></div>