14:03:12 <liuyulong> #startmeeting neutron_l3
14:03:13 <openstack> Meeting started Wed Mar 25 14:03:12 2020 UTC and is due to finish in 60 minutes.  The chair is liuyulong. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:03:14 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:03:16 <openstack> The meeting name has been set to 'neutron_l3'
14:03:48 <haleyb> hi
14:03:55 <liuyulong> hi
14:04:04 <ralonsoh> hi
14:05:01 <liuyulong> OK, let's start.
14:05:05 <liuyulong> #topic Announcements
14:05:08 <slaweq> hi
14:05:41 <liuyulong> #link http://eavesdrop.openstack.org/meetings/networking/2020/networking.2020-03-24-14.00.log.html#l-11
14:05:51 <liuyulong> Recall the team Announcements
14:07:49 <liuyulong> Since we are an online community at most time, a virtual PTG should work for our community members.
14:07:54 <liuyulong> #link https://www.openstack.org/events/opendev-ptg-2020/
14:08:27 <liuyulong> So the Vancouver PTG is officially cancelled.
14:10:00 <liuyulong> I've been working from home about 2 months.
14:11:05 <liuyulong> And more days may still be needed.
14:11:11 <liuyulong> So, take care of yourself guys.
14:11:35 <slaweq> liuyulong: thx, I'm working from home for last 2 years already ;)
14:11:56 <liuyulong> slaweq, cool, good to know that.
14:12:32 <liuyulong> slaweq, you are really safe now and then.
14:12:45 <liuyulong> OK, according to the bug deputy email
14:12:47 <liuyulong> #link http://lists.openstack.org/pipermail/openstack-discuss/2020-March/013510.html
14:12:57 <liuyulong> IMO, no L3 bugs last week.
14:13:48 <liuyulong> So, anyone has bug to discuss here? Or let's just go next topic.
14:14:14 <liuyulong> OK
14:14:20 <liuyulong> #topic OVN_L3
14:14:58 <liuyulong> #link https://bugs.launchpad.net/neutron/+bug/1867869
14:14:59 <openstack> Launchpad bug 1867869 in neutron "[OVN] Remove dependency on fip_object" [Low,In progress] - Assigned to zhanghao (zhanghao2)
14:15:12 <ralonsoh> this is still ongoing
14:15:27 <liuyulong> One bug from OVN
14:15:36 <ralonsoh> we need to know if this will be needed in QoS FIP
14:16:02 <ralonsoh> so, IMO, I'll wait to remove this dependency
14:17:52 <liuyulong> ralonsoh, sure, thank you for the information. Mostly are TODOs from lucasagomes, IMO, he may also have a deep review.
14:18:59 <liuyulong> He is already in the list.
14:19:31 <liuyulong> Next should be the patch:
14:19:33 <liuyulong> #link https://review.opendev.org/#/c/705660/
14:19:57 <liuyulong> I run it locally last week, it works. But it now needs a rebase.
14:20:14 <ralonsoh> maciejjozefczyk, ^
14:21:22 <ralonsoh> I'll ping him in gerrit or IRC
14:22:17 <maciejjozefczyk> ralonsoh, thanks for testing!
14:22:21 <liuyulong> Sure, I will run the rebased code again if I have enough time. : )
14:22:49 <maciejjozefczyk> I'll rebase it
14:23:07 <liuyulong> #link https://review.opendev.org/#/c/710881/2/doc/source/ovn/gaps.rst
14:23:58 <liuyulong> This is from the gaps from between ML2/OVS and OVN, most of them looks like are related to L3.
14:24:45 <maciejjozefczyk> Ok, so for 'QoS for Layer 3 IPs': This week I tested OVN QoS on FIP - there is a possibility. Its only a matter of proper match in QoS row.
14:25:35 <maciejjozefczyk> #link: https://mail.openvswitch.org/pipermail/ovs-discuss/2020-March/049857.html
14:25:48 <liuyulong> I want to say that "QoS for Layer 3 IPs" has the higher priority from the perspective of cloud provider.
14:26:43 <liuyulong> Datacenter could not ensure that every external IPs to reach the max bandwidth at the same time.
14:31:42 <liuyulong> maciejjozefczyk, cool, so just logical flows should be added from L3_ovn plugin to north DB, and then ovn-controller, gateway chassis should do the meters installation.
14:33:09 <maciejjozefczyk> liuyulong, yes. For now we're working on refactor of adding of QoS rows in NBDB table. From this table the Logical Flows are generated.
14:33:36 <maciejjozefczyk> After a refactor that would be possible to add specific QoS row that will match a FIP
14:33:55 <maciejjozefczyk> from Core OVN perspective its possible and I tested it manually, N/S traffic is shaped, while E/W not :)
14:36:23 <liuyulong> So the OVN DB schema should be refactored first at core OVN?
14:37:52 <liuyulong> OK, I found the answer. https://mail.openvswitch.org/pipermail/ovs-discuss/2020-March/049863.html
14:38:31 <maciejjozefczyk> liuyulong, There is no need to change anything in Core OVN - in relation to QoS on FIP.
14:40:34 <liuyulong> "inport == \"96242e8f-602f-4077-984a-5659e84a8c4f\" && ip4.src == && is_chassis_resident(\"cr-lrp-96242e8f-602f-4077-984a-5659e84a8c4f\")"
14:41:17 <liuyulong> I'm not quite sure about this, so in gateway chassis, the packet has already be NATed (source IP) to floating IP.
14:41:20 <maciejjozefczyk> liuyulong, yes, Neutron, while creating the QoS row should define match similar to this oen
14:41:46 <liuyulong> But, what about the ingress from outside world, the destination IP is floating IP.
14:42:22 <ralonsoh> that will be another rule
14:42:28 <ralonsoh> depending on the rule direction
14:42:43 <ralonsoh> the OVN qos rules will match the Neutron QoS ones
14:43:01 <ralonsoh> and will add the needed "match" field in the NB
14:43:15 <ralonsoh> (matching src/dst IP or whatever is needed)
14:43:19 <maciejjozefczyk> liuyulong, yes, the match field in the rule will look differently
14:44:10 <liuyulong> So, this line is only for egress (from the perspective of VM) to outside world.
14:44:29 <ralonsoh> yes
14:44:44 <liuyulong> ralonsoh, maciejjozefczyk, thank you guys, : )
14:45:48 <liuyulong> OK, no more OVN related things from me then.
14:45:58 <liuyulong> Any updates?
14:46:17 <liuyulong> OK, let's move on.
14:46:21 <liuyulong> #topic On demand agenda
14:46:45 <liuyulong> Something related to IPv6 again.
14:47:09 <liuyulong> I rebased the patch https://review.opendev.org/#/c/662111/ from Swaminathan, hope we could continue the work.
14:49:34 <haleyb> liuyulong: yes, that would be good to move forward. We should think about OVN as well
14:51:57 <liuyulong> haleyb, indeed, we should not introduce another gap. : )
14:53:55 <haleyb> i know, but knowing it exists and adding it to the list is good, that way it can be worked on.
14:55:17 <liuyulong> +1 make sense
14:56:36 <liuyulong> I will test Swaminathan's patch more deeply locally, and comment the feedback.
14:57:14 <liuyulong> IMO, it is pretty close now.
14:57:38 <liuyulong> Alright, no more topics from me now.
15:00:17 <liuyulong> Time is up.
15:00:26 <liuyulong> Lets end here.
15:00:33 <liuyulong> #endmeeting