15:00:08 <mlavalle> #startmeeting neutron_l3
15:00:09 <openstack> Meeting started Thu May 25 15:00:08 2017 UTC and is due to finish in 60 minutes.  The chair is mlavalle. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:10 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:12 <openstack> The meeting name has been set to 'neutron_l3'
15:00:18 <mlavalle> #chair haleyb
15:00:19 <openstack> Current chairs: haleyb mlavalle
15:00:29 <haleyb> hi
15:00:43 <haleyb> Swami will not be attending today
15:00:49 <mlavalle> swami pinged me earlier today. He won't make it to today's meeting
15:00:56 <mlavalle> LOL
15:01:01 <haleyb> jinx
15:01:20 <mlavalle> #topic Announcements
15:02:08 <mlavalle> Pike 2 is almost here, June 5 - 9.
15:02:25 <mlavalle> Less than 2 weeks from now
15:02:38 <mlavalle> #link https://releases.openstack.org/pike/schedule.html
15:03:07 <mlavalle> any other annoucements
15:03:09 <mlavalle> ?
15:03:32 <haleyb> not from me
15:03:43 <mlavalle> ok.....
15:03:49 <mlavalle> #topic Bugs
15:04:20 <mlavalle> haleyb: do you have dvr bugs to discuss today?
15:05:14 <haleyb> there are no new bugs, let me look quickly at older ones
15:05:40 <mlavalle> cool
15:06:21 <haleyb> no, just the RFE/bug swami is working on for centralized
15:06:37 <mlavalle> ok
15:07:17 <mlavalle> on the l3-dhcp-ipam side:
15:07:23 <mlavalle> https://bugs.launchpad.net/neutron/+bug/1684682
15:07:27 <openstack> Launchpad bug 1684682 in neutron "DHCP namespace doesn't have IPv6 default route" [High,In progress] - Assigned to Brian Haley (brian-haley)
15:07:39 <mlavalle> I reviewed the patchset earlier this week
15:07:54 <mlavalle> Nice! I enjoyed reviewing it
15:08:13 <haleyb> mlavalle: yes, thanks for that, i will ping kevinbenton for review
15:08:26 <mlavalle> cool
15:08:46 <mlavalle> Next one is https://bugs.launchpad.net/neutron/+bug/1655567
15:08:47 <openstack> Launchpad bug 1655567 in neutron "DHCPv6 failures with "address already allocated in subnet" error" [High,Confirmed] - Assigned to Miguel Lavalle (minsel)
15:09:04 <mlavalle> I made progress on this one
15:09:29 <mlavalle> I found it is a race condition between two requests
15:10:23 <haleyb> two requests asking for the same IP?
15:10:42 <mlavalle> not exactly
15:11:06 <mlavalle> first one attempts to create a dhcpv6-stateless subnet
15:11:48 <mlavalle> the second one updates the dhcp port for this subnet and allocates an ip adress succesfully
15:12:16 <mlavalle> so when the first one executes https://github.com/openstack/neutron/blob/master/neutron/db/ipam_pluggable_backend.py#L449
15:12:32 <mlavalle> it fails adding ip address to the DHCP port
15:14:40 <mlavalle> haleyb: so, should the DHCP agent request be trying to add an ip address to the port on a dhcpv6-stateless subnet?
15:15:18 <haleyb> i was just looking at that code for another bug
15:15:37 <mlavalle> isn't that an autoaddress that should be added by the method above^^^
15:15:40 <mlavalle> ?
15:15:41 <haleyb> it requests a port on all subnets where dhcp is enabled
15:15:57 <haleyb> so maybe there is a bug?
15:16:05 <mlavalle> I am certain there is a bug
15:16:23 <mlavalle> either the agent shouldn't be requesting the ip address in the second request
15:16:40 <mlavalle> or the method above should skip dhcp ports
15:17:01 <mlavalle> right?
15:17:37 <mlavalle> since this is an autoaddress subnet, I lean towards the bug being in the agent (the second request)
15:18:12 <mlavalle> That is why we have the method above^^^ to allocate from autoaddress subnets
15:18:28 <haleyb> i'm thinking the agent should be the one to request it
15:19:13 <haleyb> in the ipv4 case it's the agent that does this
15:19:39 <mlavalle> ok, in that case, we should refine this query https://github.com/openstack/neutron/blob/master/neutron/db/ipam_pluggable_backend.py#L461
15:19:53 <mlavalle> to exclude dhcp owned ports
15:21:17 <haleyb> mlavalle: that's probably worth a try
15:21:37 <mlavalle> haleyb: ok, let's follow that path
15:21:41 <haleyb> we probably haven't thought of something which the tests should shake out
15:22:16 <mlavalle> here's the plan: 1) I will spend some time in kibana with other ocurrences of the bug, to make sure I see the same thing
15:22:17 <haleyb> mlavalle: the other way is to have the agent look for the mode flags
15:22:29 <mlavalle> yeah
15:23:06 <mlavalle> and if the mode flag is dhcpv6-stateless, let the server take care of the address, right?
15:24:22 <haleyb> right, that might be more invasive and wouldn't cover other drivers that use the same logic as the linux one
15:25:16 <mlavalle> but it is conceptually beter, IMO. since the subnet is dhcpv6-stateless, dhcp shouldn't be handing out addresses
15:25:18 <haleyb> the linux driver should have tried to use an existing port, and only creates a new one if it doesn't exist, so guess that's the race
15:26:21 <haleyb> mlavalle: for dhcpv6-stateless the driver will hand-out addresses, but instance must learn other info from the router advertisement
15:26:40 <mlavalle> ah ok, that was my question above
15:27:20 <haleyb> i should have a link, but the code is in setup_dhcp_port()
15:27:35 <mlavalle> so, as I said, i'll confirm in kibana with other instances of the bug, to make sure I see the same thing
15:27:43 <mlavalle> and then I will propose a fix
15:27:49 <mlavalle> sounds like a plan?
15:28:07 <haleyb> +1
15:28:15 <mlavalle> cool
15:28:54 <mlavalle> Last one I have today is https://bugs.launchpad.net/neutron/+bug/1683227
15:28:55 <openstack> Launchpad bug 1683227 in neutron "test_connection_from_same_address_scope failed with: Cannot find device "qg-74372988-7f"" [High,Confirmed]
15:29:21 <mlavalle> This one is still searching for an owner
15:29:46 <mlavalle> if one doesn't show up, I'll take it next and give it some TLC
15:30:52 <mlavalle> any other bugs we should discuss today?
15:30:58 <haleyb> ok, thanks.  i will try and look as well
15:32:42 <mlavalle> take that as a no
15:32:53 <haleyb> sorry, no more bugs
15:32:59 <mlavalle> #topic DNS extension
15:33:54 <mlavalle> I've been working on implementing https://bugs.launchpad.net/neutron/+bug/1650678
15:33:55 <openstack> Launchpad bug 1650678 in neutron "[RFE] Allow specifying dns_domain when creating a port" [Wishlist,In progress] - Assigned to Miguel Lavalle (minsel)
15:34:15 <mlavalle> I have already 3 patchsets in gerrit
15:34:43 <mlavalle> and I just finished implementing the logic. I will push that patchset over the next few days
15:34:52 <mlavalle> the logic is working fine
15:35:03 <haleyb> :) i'll add myself if i haven't already
15:35:15 <mlavalle> it doesn't break any current tests
15:35:28 <mlavalle> I need to add tests for the new functionality, though
15:35:52 <mlavalle> haleyb: no need to review just yet. I'll let you know when the entire series is ready
15:36:31 <haleyb> mlavalle: ok, i just like getting gerrit emails :)
15:36:31 <mlavalle> haleyb: should we add this to the team's charter for Pike?
15:37:19 <haleyb> mlavalle: yes, think we can add if it's this far along already
15:37:34 <mlavalle> ok, i'll do it
15:38:13 <mlavalle> #topic Open Agenda
15:38:26 <mlavalle> any other topics we should discuss today?
15:39:00 <haleyb> mlavalle: i know you answered a few questions for hjensas
15:39:19 <haleyb> but i think he got some good answers from you and kevin
15:39:32 <mlavalle> yeah, I intend to follow up with him and I have the bug he filed in my radar screen
15:40:25 <haleyb> mlavalle: great, thanks.  i had asked him to add an agenda item if there were more questions, but think kevin answered them last night
15:41:32 <mlavalle> any other topics?
15:44:29 <mlavalle> ok, have a nice rest of your week
15:44:42 <mlavalle> #endmeeting