14:06:45 <liuyulong> #startmeeting neutron_l3
14:06:45 <openstack> Meeting started Wed Sep 18 14:06:45 2019 UTC and is due to finish in 60 minutes.  The chair is liuyulong. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:06:46 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:06:48 <openstack> The meeting name has been set to 'neutron_l3'
14:06:58 <ralonsoh> hi
14:07:07 <liuyulong> hi
14:07:30 <liuyulong> #topic Announcements
14:08:13 <liuyulong> I have no announcement today.
14:08:19 <slaweq> hi
14:08:32 <liuyulong> So if you have, please go ahead.
14:09:23 <liuyulong> OK, let's move on.
14:09:38 <liuyulong> #topic Bugs
14:10:20 <liuyulong> #link http://lists.openstack.org/pipermail/openstack-discuss/2019-September/009361.html
14:10:48 <liuyulong> Miguel sent this on Sunday.
14:12:04 <liuyulong> First one from the deputy's mail:
14:12:07 <liuyulong> #link https://bugs.launchpad.net/neutron/+bug/1843889
14:12:08 <openstack> Launchpad bug 1843889 in neutron "Windows: IPv6 tunnel endpoints" [Medium,In progress] - Assigned to Lucian Petrut (petrutlucian94)
14:12:38 <haleyb> hi, i'm in two meetings at once though, will try and lurk here
14:12:47 <ralonsoh> IMO, the patch is OK, although we can't test it in the CI
14:12:50 <ralonsoh> Linux CI
14:13:04 <ralonsoh> btw, #link https://review.opendev.org/#/c/682031/
14:13:52 <liuyulong> Yes, thank you, patch is simple. But the bug is abit out of my scope honestly. :)
14:15:46 <haleyb> i +W, looked good and they have been good at fixing this code
14:16:56 <liuyulong> You're one step ahead of me while you are attending two meetings. : )
14:17:15 <liuyulong> Next
14:17:18 <liuyulong> #link https://bugs.launchpad.net/neutron/+bug/1844168
14:17:19 <openstack> Launchpad bug 1844168 in neutron "[L3] TooManyExternalNetworks: More than one external network exists." [Low,In progress] - Assigned to LIU Yulong (dragon889)
14:17:43 <ralonsoh> one question in your patch
14:17:51 <ralonsoh> you removed the router check
14:18:01 <ralonsoh> sorry, the link
14:18:02 <liuyulong> IMO, this should be related to the external network removal.
14:18:13 <ralonsoh> #link https://review.opendev.org/#/c/682418/
14:18:25 <liuyulong> once slaweq and Miguel have done some related work.
14:18:56 <ralonsoh> in this patch, you remove the router compatible check
14:18:57 <ralonsoh> https://review.opendev.org/#/c/682418/4/neutron/agent/l3/agent.py@a639
14:19:19 <ralonsoh> but I think you should retrieve the external network list (pfor this agent)
14:19:40 <ralonsoh> and then check if the router external network is there --> compatible/not compatible then
14:20:10 <haleyb> i also can't remember removing an rpc, do we need to do that?
14:20:34 <ralonsoh> haleyb, this method should no be used anymore
14:20:57 <ralonsoh> because in case of having more than one ext netowkr, an exception will be thrown
14:21:25 <ralonsoh> but, now we should have something retrieving the ext network list
14:21:26 <haleyb> it is used by bigswitch, and i didn't know how it would affect a mixed-version environment
14:21:30 <liuyulong> hmm, I may revisit the history of this check.
14:22:27 <liuyulong> L3 agent supports muiltple external networks, this check is abit strange
14:23:22 <liuyulong> If it is not compatible, it should not be scheduled to that agent. But seems we not can't avoid that.
14:26:12 <liuyulong> More interesting thing is if a router is scheduled to L3 agent without external_gateway_info, for instance HA router creation, the compatible check does not make any sense.
14:27:13 <ralonsoh> but you can check that there is no extneral_gw_info
14:27:37 <ralonsoh> but if you have, you should check the network ID and if this ext net belongs to this L3 agent
14:28:29 <liuyulong> Then you set the external gateway info, but you may see no errors, but L3 agent just do nothing for this router update notification.
14:29:39 <slaweq> maybe we should now go quickly with smallest possible fir for this reported bug
14:29:53 <slaweq> and later we can remove (or deprecate) some not used methods
14:30:24 <ralonsoh> +1
14:32:40 <slaweq> liuyulong: is this bug happening all the time when there is more than one external network in Neutron?
14:32:43 <liuyulong> Actually the get_external_network_id just get all external networks, it does not have any relationship of L3 agent.
14:33:22 <liuyulong> slaweq, you can see the log search in the bug description.
14:33:49 <slaweq> ok, that should have IMO higher priority and we should maybe target it for rc-1
14:33:52 <slaweq> what do You think?
14:34:24 <liuyulong> I'm OK with it.
14:34:50 <slaweq> I will ask mlavalle later about it also
14:36:16 <liuyulong> slaweq, cool, IMO, it is related to this patch: https://review.opendev.org/#/c/666409/
14:36:48 <slaweq> liuyulong: yes, I'm pretty sure it is :)
14:37:05 <liuyulong> We have another merged patch, also did such removal.
14:39:17 <liuyulong> OK, next one
14:39:44 <liuyulong> #link https://bugs.launchpad.net/neutron/+bug/1844171
14:39:45 <openstack> Launchpad bug 1844171 in neutron "Configuration parameter missing at "Configure the layer-3 agent"" [Low,In progress] - Assigned to YAMAMOTO Takashi (yamamoto)
14:40:12 <liuyulong> Patch just get a +W from me. : )
14:40:14 <ralonsoh> trivial patch IMO, +2
14:40:37 <ralonsoh> #link https://review.opendev.org/#/c/682645/ (just in case)
14:41:31 <liuyulong> Next two are all related to IPv6:
14:41:33 <slaweq> I also gave +2 :D
14:41:35 <liuyulong> #link https://bugs.launchpad.net/neutron/+bug/1844097
14:41:36 <openstack> Launchpad bug 1844097 in neutron "Redundant ipv6 address(SLAAC/DHCPv6 stateless) created for port" [Undecided,New]
14:41:41 <liuyulong> #link https://bugs.launchpad.net/neutron/+bug/1844171
14:41:42 <openstack> Launchpad bug 1844171 in neutron "Configuration parameter missing at "Configure the layer-3 agent"" [Low,In progress] - Assigned to YAMAMOTO Takashi (yamamoto)
14:41:57 <liuyulong> #undo
14:41:58 <openstack> Removing item from minutes: #link https://bugs.launchpad.net/neutron/+bug/1844171
14:42:10 <liuyulong> #link https://bugs.launchpad.net/neutron/+bug/1844123
14:42:10 <openstack> Launchpad bug 1844123 in neutron "Unable to trigger IPv6 Prefix Delegation" [Undecided,New]
14:42:30 <liuyulong> One can not get IPv6 address, but another just get two...
14:44:03 <liuyulong> During our local testing, we tested the port creation in network with IPv4 and IPv6 subnets.
14:44:25 <liuyulong> It works fine, the port will have only one IPv4 addr and one IPv6 addr.
14:44:59 <ralonsoh> I don't understand https://bugs.launchpad.net/neutron/+bug/1844097
14:45:00 <openstack> Launchpad bug 1844097 in neutron "Redundant ipv6 address(SLAAC/DHCPv6 stateless) created for port" [Undecided,New]
14:45:30 <liuyulong> From the bug description, seems they do not need that IPv6 address...
14:46:06 <ralonsoh> but they are creating a port in a network with two subnets
14:46:11 <ralonsoh> ipv4 and ipv6
14:46:20 <liuyulong> They think the IPv6 address is "redundant".
14:46:43 <ralonsoh> what they are doing is defining the IPv4 one, but the ipv6 will be assigned automatically
14:46:54 <ralonsoh> I'll review this and comment in the bug
14:47:20 <liuyulong> Yes, maybe we can mark it as invalid. It is an feature.
14:48:35 <liuyulong> OK, I remember another bug.
14:51:16 <liuyulong> This https://bugs.launchpad.net/neutron/+bug/1752903
14:51:18 <openstack> Launchpad bug 1752903 in neutron "Floating IPs should not allocate IPv6 addresses" [Medium,New]
14:52:13 <liuyulong> And I had a proposal for the port creation API once related to this: https://bugs.launchpad.net/neutron/+bug/1797140
14:52:14 <openstack> Launchpad bug 1797140 in neutron "[RFE] create port by providing parameters subnet_id only" [Wishlist,Won't fix] - Assigned to LIU Yulong (dragon889)
14:53:30 <liuyulong> OK, that's all from me.
14:54:14 <liuyulong> Time is running out...
14:54:51 <liuyulong> #topic On demand agenda
14:55:06 <liuyulong> OK, continue with the IPv6.
14:56:50 <liuyulong> NetworkManager seems not work for the VM NICs, the configured IPv6 address is not same to the port IPv6 address.
14:57:32 <liuyulong> Maybe bad settings.
14:57:53 <slaweq> liuyulong: what do You mean exactly?
14:58:38 <slaweq> do You mean that instance have got different IPv6 configured with SLAAC/DHCPv6 instead of what is in Neutron's DB?
14:58:40 <liuyulong> CentOS7 image's default NetworkManger is mis-configured. IMO
14:58:49 <liuyulong> slaweq, Yes
14:58:58 <slaweq> https://developer.gnome.org/NetworkManager/unstable/settings-ipv6.html
14:59:03 <slaweq> it's "privacy extension"
14:59:38 <slaweq> basically IPv6 is given to instance by router based on link local IP which is based on MAC address
15:00:08 <slaweq> if this extension is enabled, request is done with some "fake" mac address thus router responds with wrong IP address
15:00:25 <slaweq> if You change this setting in NetworkManager, it will work as expected
15:00:38 <liuyulong> Yes, we are tunning the settings
15:02:04 <liuyulong> OK, let's stop here.
15:02:15 <liuyulong> #endmeeting