15:01:34 <ihrachys> #startmeeting neutron_upgrades 15:01:38 <openstack> Meeting started Mon Jan 16 15:01:34 2017 UTC and is due to finish in 60 minutes. The chair is ihrachys. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:39 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:41 <openstack> The meeting name has been set to 'neutron_upgrades' 15:01:44 <ihrachys> #link https://wiki.openstack.org/wiki/Meetings/Neutron-Upgrades-Subteam 15:02:12 <electrocucaracha> o/ 15:03:05 <korzen> hello 15:03:07 <ihrachys> today is MLK day in some US offices, that's why I believe we have lower presence 15:03:17 <korzen> what is MLK? 15:03:24 <ihrachys> Martin Luter King day 15:03:30 <korzen> oh, ok 15:03:42 <electrocucaracha> Yeah , it's 15:03:44 <korzen> we do not have this holiday in Europe ;) 15:03:49 <ihrachys> electrocucaracha: Intel celebrating? 15:04:14 <korzen> Intel US szhould have free day 15:04:29 <ihrachys> I see. well, I am in RedHat US, and I don't have a holiday. ;) 15:04:29 <electrocucaracha> Yes, but Rackspace no 15:04:47 <ihrachys> gotcha. ok, let's run it quick. 15:04:55 <ihrachys> #topic Announcements 15:05:31 <ihrachys> as you know, there is PTG next month 15:05:43 <ihrachys> armax sent an email with a link to etherpad: http://lists.openstack.org/pipermail/openstack-dev/2017-January/110040.html 15:06:07 <ihrachys> it would probably make sense to brain storm which upgrade topics may be worth covering there 15:06:12 <ihrachys> and filling in the etherpad 15:06:33 <electrocucaracha> I won't be able to be there, but I'll try to follow you in etherpad 15:06:49 <ihrachys> I was thinking we can try to make progress on mixed server version gating, but that probably belongs to cross-project sessions, not neutron 15:07:21 <korzen> I have made some small list: 15:07:28 <korzen> API versioning 15:07:33 <korzen> New mechanism for blocking contract migrations 15:07:39 <korzen> Online Data Migration 15:08:06 <ihrachys> korzen: what's the 'new' mechanism? 15:08:26 <korzen> ihrachys, I mean to change that UT 15:08:49 <korzen> maybe it is not so critical 15:10:08 <ihrachys> I guess it's high detail. as for other two, I think both are good. can you put them into the pad so that we can start expanding on them this week? 15:10:30 <korzen> ok 15:12:05 <ihrachys> ok cool 15:12:11 <korzen> do you think about something more to be discuss? 15:13:24 <ihrachys> I think we are otherwise good; push-notifications progress may also be relevant to us, but I believe it will be raised in other context. 15:15:06 <ihrachys> ok, don't hesitate to fill in the pad with more topics if you come up with any 15:15:08 <ihrachys> #topic Partial Multinode Grenade 15:15:44 <ihrachys> linuxbridge job still not feeling well. I also saw tempest (not grenade) job failing at 30%+ rate previous week. 15:16:03 <ihrachys> so I was not really sure if it's grenade specific 15:16:13 <ihrachys> now it seems the tempest flavour is back to normal 15:16:17 <ihrachys> http://grafana.openstack.org/dashboard/db/neutron-failure-rate?panelId=8&fullscreen 15:16:48 <ihrachys> gotta look at it closer this week 15:17:21 <ihrachys> reminder: it's already in check queue, but not-voting so far. 15:17:35 <ihrachys> let's move to objects 15:17:36 <ihrachys> #topic Object implementation 15:18:17 <ihrachys> we landed DictOfMiscValuesField: https://review.openstack.org/411830 15:18:29 <korzen> :) 15:18:46 <ihrachys> we also landed tenant_id -> project_id switch: https://review.openstack.org/#/c/382659/ 15:18:58 <korzen> good work dasm 15:19:44 <ihrachys> one of the more critical bits up for review is port binding adoption: https://review.openstack.org/#/c/407868/ 15:19:57 <ihrachys> I reviewed it the previous week; it's currently in conflict 15:20:04 <ihrachys> would be nice if we can rebase it sooner 15:20:17 <korzen> I will ping Lujin 15:20:24 <ihrachys> thanks 15:20:51 <korzen> I was about to rebase and address comments but waited for path author to do it first 15:21:12 <ihrachys> there is nothing major there now, it's just a matter of code cleanup 15:23:37 <korzen> I wonder if subnet, port and network OVO adoption patch are relevant? 15:23:42 <korzen> for Ocata release 15:23:44 <ihrachys> NetworkSegment adoption: https://review.openstack.org/#/c/385178 also seems close 15:23:52 <ihrachys> korzen: it may or it may not, depending on readines 15:23:57 <ihrachys> *readiness 15:24:13 <ihrachys> electrocucaracha: anything that blocks us from respinning ^ ? 15:24:37 <electrocucaracha> I don't remember the last issue 15:24:47 <ihrachys> electrocucaracha: I will look at the UT failure you mentioned for .db_obj 15:24:59 <ihrachys> apart from it, it seems close 15:25:05 <electrocucaracha> Ohh that one 15:25:36 <electrocucaracha> I can take a look but that will be tomorrow 15:26:27 <ihrachys> electrocucaracha: sure, no need to spoil the holiday ;) 15:26:47 <ihrachys> korzen: are any of those adoptions close? 15:27:05 <korzen> subnet is 15:27:17 <korzen> port and network are non-existing 15:27:33 <korzen> subnet is implemented 15:27:46 <korzen> but I had issue with db_obj being detached from session 15:28:06 <korzen> and I'm waiting for Ann to get it fixed with new engine facade 15:28:19 <ihrachys> I see. will new engine fix it? 15:28:45 <korzen> yes, because the new engine is adding the object to active session 15:28:49 <korzen> db_obj 15:29:32 <korzen> well, I need to rebase subnet adoption patch on top of Ann's patch 15:29:47 <korzen> but in theory it should work 15:29:57 <ihrachys> korzen: I guess it's https://review.openstack.org/#/c/402750 that is needed for subnet? 15:30:13 <korzen> yes, that one 15:31:11 <ihrachys> it's mostly very simple 15:31:16 <ihrachys> I will take a look 15:31:21 <ihrachys> in the meantime, rebase would be nice 15:31:21 <electrocucaracha> What about the Flavor and service Profile patch it's also close to be completed 15:31:27 <korzen> It is also a matter of push-notification 15:32:36 <ihrachys> electrocucaracha: that one, I just +2d it again, let's find a second opinion. rossella_s, https://review.openstack.org/#/c/306685/ maybe? 15:32:59 <rossella_s> ihrachys, ack, will look at it 15:33:03 <ihrachys> korzen: I am not sure push-notification is related, but we will cover it in a separate section 15:33:10 <electrocucaracha> Thanks rossella_s 15:33:21 <rossella_s> electrocucaracha, my pleasure 15:34:12 <ihrachys> ok, let's move on to next section 15:34:17 <ihrachys> #topic Other patches to review 15:34:23 <electrocucaracha> Just FYI, I included a new status on the spreadsheet that can help to focus on those OVo patches that I consider ready 15:34:36 <ihrachys> electrocucaracha: thanks 15:34:38 <ihrachys> first topic is, multiple port binding 15:35:02 <ihrachys> spec being https://review.openstack.org/#/c/309416/ 15:35:11 <ihrachys> I think it's very close, next respin should do 15:35:46 <ihrachys> there is also a patch adding host to primary key in the binding table: https://review.openstack.org/#/c/404293/ 15:36:32 <ihrachys> that one is also very close, I just need to have a look at replies to my comments. 15:37:13 <ihrachys> with that, we may get the PK expansion in Ocata BEFORE nova or neutron starts writing into it, which should simplify the online db upgrade matters once we get to landing actual extension 15:37:48 <ihrachys> which prolly won't happen this time 15:38:19 <ihrachys> any questions on port binding? 15:38:25 <korzen> so the whole feature port binding for LM wont happen in Ocata 15:38:27 <korzen> ? 15:38:34 <ihrachys> korzen: it looks that way, yes. 15:39:02 <korzen> I was thinking at implementation patch, if we also need to update the OVO? 15:39:07 <korzen> for port binding 15:39:11 <ihrachys> I think back in Barcelona, there was agreement that it may as well happen, so it shouldn't be news for e.g. nova 15:40:14 <ihrachys> korzen: if we land db expansion this cycle only, then no need to update it in Ocata; we will add new field in Pike 15:40:43 <ihrachys> for status 15:41:03 <ihrachys> or you mean that now primary key is expanded, so we need to reflect it in the object? 15:41:39 <korzen> if we are expending, then it should not matter 15:41:47 <korzen> expanding* 15:42:12 <korzen> until API and backend code is ready to consume it 15:42:17 <ihrachys> korzen: well, for one, get_object should allow only filters that result in unique result? 15:42:52 <ihrachys> and unless we add the host to PK in object, it will happily accept just port_id 15:43:00 <korzen> so you mean when we will have extra PK, ovo get_object won't work? 15:43:53 <ihrachys> I think it may then fail if there are multiple matching records in db 15:44:20 <korzen> get_object would return .first() 15:44:25 <ihrachys> hm, we implement get_object as .first 15:44:28 <ihrachys> yea 15:44:54 <korzen> adding new PK is backward compatible 15:44:58 <korzen> right? 15:45:21 <ihrachys> if the code does not assume a single record (like calling .one) then yes, it's ok 15:45:22 <korzen> there is no new code that will insert port_id twice 15:45:38 <ihrachys> btw I checked Newton code and haven't found any place were we would assume that 15:46:08 <ihrachys> there is no such code now, but there will be in Pike, so Ocata code should be ready for that. 15:46:45 <korzen> good point 15:46:50 <korzen> I will take another look 15:47:46 <ihrachys> ok, let's move on to push notifications 15:47:48 <korzen> so, I guess that merging the port binding ovo adoption in Ocata would be helpful 15:47:57 <korzen> one last question ;) 15:47:59 <ihrachys> korzen: absolutely! 15:48:04 <korzen> ok 15:48:13 <ihrachys> korzen: was that your question? 15:48:21 <korzen> that one 15:48:36 <ihrachys> ok. yes, it would be very helpful. I want to land it asap. 15:48:36 <korzen> merging port_bindgin adoption 15:48:40 <korzen> ok 15:49:02 <ihrachys> speaking of push-notifications, kevinbenton made some progress 15:49:13 <ihrachys> a patch to mention is server side notifier: https://review.openstack.org/#/c/388939/ 15:49:34 <ihrachys> the notifier subscribes to AFTER_* events, re-fetches the relevant object, and pushes it on the wire 15:49:44 <ihrachys> note it re-fetches, not reuses the same object 15:49:51 <ihrachys> that was used during update of the database 15:50:19 <ihrachys> that's what I meant saying that landing adoption patches is not strictly relevant to push-notifications now 15:50:51 <ihrachys> is my thinking correct? korzen 15:52:01 <korzen> it looks like 15:52:39 <korzen> it would affect only the RPC 15:52:46 <korzen> and not REST API 15:52:57 <ihrachys> right. as long as we can construct the object from ID, it should do 15:53:34 <ihrachys> I don't know of any patches that currently consume the objects on agent side, but I believe kevinbenton is on it 15:53:49 <ihrachys> we may want to help with reviews there too 15:54:34 <korzen> I was looking at some patches but did not have enough time to jump into details 15:54:49 <ihrachys> korzen: so far there is not much to review there 15:54:52 <ihrachys> ok let's look at the list of patches with UpgradeImpact: https://review.openstack.org/#/q/status:open+message:%22UpgradeImpact%22+project:openstack/neutron 15:55:14 <ihrachys> ok, first one will require a spec/write-up, so we can skip that one 15:55:55 <ihrachys> second is, not allowing to start macvtap agent with wrong bridge configuration: https://review.openstack.org/#/c/323398/ 15:56:00 <ihrachys> sounds like no-brainer to me 15:56:15 <ihrachys> and we do it for other agents already 15:57:03 <ihrachys> finally, the patch that claims to solve another network disruption in ovs agent: https://review.openstack.org/#/c/305724/ 15:57:13 <ihrachys> I thought we solved all of those like 2 cycles ago? :) 15:57:42 <korzen> in Liberty 15:57:47 <ihrachys> seems like it's l2pop only, at least that's what the patch touches 15:57:47 <korzen> and in Mitaka 15:57:52 <ihrachys> so corner case 15:57:55 <ihrachys> korzen: and in Newton! :) 15:58:21 <ihrachys> if that's l2pop only, we may look into expanding the scenarios list in fullstack connectivity tests. 15:58:28 <korzen> which one was in Newton? 15:58:33 <ihrachys> I believe we have some tests there that restart agents and ping in the meantime 15:58:38 <ihrachys> korzen: I am just kiddin' 15:59:03 <korzen> :) 15:59:15 <ihrachys> ok, let's do the last quick 15:59:18 <ihrachys> #topic Open discussion 15:59:22 <ihrachys> there is not much time 15:59:31 <korzen> https://review.openstack.org/#/c/336518/ 15:59:33 <ihrachys> but if you have something to raise and move to #openstack-neutron please do 15:59:37 <korzen> ihrachys, ^ 15:59:42 <ihrachys> sigh 15:59:42 <korzen> please review ;) 15:59:43 <ihrachys> yeah 15:59:55 <electrocucaracha> Thanks 16:00:09 <korzen> ok, thanks :) 16:00:10 <ihrachys> ok that's it, we need to free the floor 16:00:12 <ihrachys> #endmeeting