14:00:39 #startmeeting neutron_drivers 14:00:39 Meeting started Fri Oct 28 14:00:39 2022 UTC and is due to finish in 60 minutes. The chair is ralonsoh. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:39 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:39 The meeting name has been set to 'neutron_drivers' 14:00:40 o/ 14:00:42 hello all 14:00:50 hi 14:01:12 Hi 14:01:37 o/ 14:02:00 I think we have quorum 14:02:02 let's start 14:02:07 #link https://bugs.launchpad.net/neutron/+bug/1994137 14:02:14 [RFE] Specify the precedence of port routes if multiple ports attached to a VM 14:02:50 there are many proposals for this RFE 14:02:57 not incompatible between them 14:03:02 hi 14:03:29 the first one is adding the router metric into the cloud-init 14:03:35 as commented by Sean 14:04:09 but I don't know if we can provide this info via dnsmasq 14:04:27 if we need this feature, then we'll need to use config drive 14:04:32 (that's what I think) 14:05:25 a second alternative (not incompatible with the other one): https://bugs.launchpad.net/neutron/+bug/1717560 14:05:39 to disable the subnet default route 14:06:14 the dnsmasq won't provide the default GW, that means the kernel won't provide a default route for this subnet 14:06:17 the above one is an old RFE, wasn't that implemented? 14:06:29 or it was just forgotten? 14:06:35 no, just abandoned 14:06:41 ok 14:06:59 that requires to set a dnsmasq option 14:07:29 (let me find it again) 14:07:51 dhcp-option=3 14:08:01 that disables the default route 14:08:23 of course, that will be for any port on this subnet 14:09:18 any opinion? 14:10:39 Just out of curiosity, how would it work in OVN without dnsmasq? 14:10:54 if I understand well the first one to add metric to cloud init is more a workaround, or am I wrong? 14:11:16 njohnsto_, no, that feature will need to be implemented in the internal dhcp server 14:12:05 for the second option it says "API extension for a new 'default_route' attribute on the subnet resource" - can't this be for a port rather than subnet? 14:12:07 lajoskatona, not really, if you add a metric, you can have multiple default routes with different metrics 14:12:17 (not very useful) but works 14:12:41 obondarev, that will depend on the DHCP service 14:12:50 agree 14:12:59 obondarev: that would overwrite the gateway on subnet? 14:13:04 for dnsmasq, for example, this is for all ports in the subnet 14:13:10 because this is a config in the service 14:13:26 for OVN could be implemented independently (but just guessing, no idea) 14:13:32 lajoskatona: I think it's more a flag rather than actual value 14:14:29 To me, passing the metric to cloud-unit seems like the best and least intrusive solution on first look 14:14:30 obondarev: yeah, that's true 14:15:47 njohnsto_, that implies not using dhcp service 14:15:51 should get more input from chateaulav? 14:16:15 (the person who brought this up in the pain points etherpad) 14:17:56 mlavalle: all for providing more info 14:18:21 good to see you here... 14:18:38 did you have an approach in mind when you brought this up? 14:18:40 chateaulav: welcome :-) 14:19:11 always been here. just not a full-time openstack dev so always pulled 5 ways to the wind 14:19:15 or could you provide more detail? 14:19:17 lol 14:21:15 chateaulav, so do you have an implementation in mind when you proposed this RFE? 14:22:31 (am I connected?) 14:22:43 ``` 14:22:43 root@hack-the-power-de-ctf-dev-1-powerplant-sel351:~# ip route 14:22:43 default via 10.30.20.1 dev ens4 proto dhcp src 10.30.20.5 metric 100 14:22:43 default via 192.168.20.1 dev ens3 proto dhcp src 192.168.20.113 metric 100 14:22:43 10.30.20.0/24 dev ens4 proto kernel scope link src 10.30.20.5 14:22:44 ralonsoh: yes 14:22:45 169.254.169.254 via 10.30.20.1 dev ens4 proto dhcp src 10.30.20.5 metric 100 14:22:45 169.254.169.254 via 192.168.20.1 dev ens3 proto dhcp src 192.168.20.113 metric 100 14:22:47 192.168.20.0/24 dev ens3 proto kernel scope link src 192.168.20.113 14:22:47 ``` 14:22:49 this is an example when both subnets have dhcp enabled. 14:23:23 i know we discussed the option of potentially passing through a route metric 14:24:12 mulitple df gw's arent bad, they just cant all be 100. and correct me if im wrong or missing something as far as being able to accomodate this now 14:26:29 no, I can't create two default routes, using two different interfaces with the same metric 14:27:08 but the goal is just the opposite, in any case 14:27:28 how do you suggest to do this? via a config drive? 14:29:30 hello? 14:29:41 we are all here 14:29:59 probabble chateaulav has a bad connection 14:30:33 We can also ask for more details in the RFE, ot understand better the possibilities 14:30:47 yes, that exactly what I was going to say 14:31:01 I don't want you to spend more time on this 14:31:12 I'll update the launchpad bug requesting more information 14:31:23 if we have feedback, we'll continue with this 14:31:26 of for you? 14:31:29 ok for you? 14:31:32 yeap 14:32:25 +1 14:32:42 +1 14:32:48 +1 14:32:57 +1 14:33:06 ok, sorry for the lack of progress in this meeting, next time I'll try to have the owners of the requests before starting 14:33:14 have a nice weekend, see you 14:33:22 o/ 14:33:22 #endmeeting