15:00:25 #startmeeting neutron_upgrades 15:00:25 Meeting started Mon Oct 3 15:00:25 2016 UTC and is due to finish in 60 minutes. The chair is ihrachys. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:26 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:27 howdy 15:00:29 The meeting name has been set to 'neutron_upgrades' 15:00:35 hello 15:00:36 hello 15:00:37 o/ 15:00:41 #link https://wiki.openstack.org/wiki/Meetings/Neutron-Upgrades-Subteam Agenda 15:01:34 I don't have specific announcements for today. everything is business as usual. 15:01:38 #topic Partial Multinode Grenade 15:01:41 o/ 15:01:46 jschwarz: around? 15:02:10 jschwarz: remember you wanted to look at linuxbridge multinode grenade job? 15:02:47 o/ 15:03:01 Hello. 15:03:04 ok I guess jschwarz is not around. 15:03:17 I will pull him in for the next meeting. 15:03:43 so far, I don't think there is any visible progress on that job, but we'll revive the effort in due course. moving on. 15:03:45 #topic Object implementation 15:04:05 a significant number of patches moving models landed lately. 15:04:09 there are more in the queue 15:04:23 we will land them in next weeks. 15:04:41 do we have all of them or just a subset though? the list does not seem complete, but I may be wrong. 15:04:52 I mean: https://review.openstack.org/#/q/status:open+project:openstack/neutron+branch:master+topic:bug/1597913 15:05:27 is there a master list of modelst aht need to be moved? 15:05:40 sorry for poor typing 15:05:49 blogan: this one -> https://docs.google.com/spreadsheets/d/1FeeQlQITsZSj_wpOXiLbS36dirb_arX0XEWBdFVPMB8/edit#gid=1434170112 15:05:52 blogan: well the easiest thing is looking into head.py file that lists all files that contain a model. 15:06:04 hi 15:06:05 blogan: the more we land those moving patches, the shorter the list 15:06:15 I just considered the models which share file with mixins 15:06:20 blogan: because everything from neutron.db.models is imported by the virtue of magic 15:06:30 ah so most, if not all, models need to be moved? 15:06:47 blogan: those other are not strictly needed. 15:06:56 blogan: we could probably go and complete that consistency? 15:07:04 electrocucaracha: so you say, the list on review is complete? 15:07:07 ihrachys: yeah 15:07:32 o/ 15:07:43 yes, I've been trying to keep it healthy 15:08:32 electrocucaracha: ok, that's good to hear. with that, we will divert our review attention to objects themselves. thanks to those who already review those, I see activity there. 15:08:50 and obviously, adoption patches are needed, if not up for review. 15:09:07 I like the approach taken by some to land smaller adoption bits where it's easily decoupled. 15:09:12 like 'adopt in tests' and such. 15:09:20 especially for those objects that are hard to adopt 15:09:31 the lesser the bits, the harder to break anything and easier to land. 15:10:32 #link https://review.openstack.org/#/q/status:open+project:openstack/neutron+branch:master+topic:bp/adopt-oslo-versioned-objects-for-db 15:10:34 good to hear is that Port and Network object landed 15:10:56 tey had 9+ months 15:10:59 oh right, did that happen the previous week? I lost that in churn around release :) 15:11:20 yeah, those were hard. there is still work to do there I believe, but we will proceed iteratively on them from now on 15:11:34 and they should not block other smaller relevant bits in the queue anymore 15:12:19 ihrachys, is kevinbenton trying to start the push-notification effort? 15:12:32 having port and network 15:12:52 korzen: I haven't seen any patches. kevinbenton as well as some other folks were consumed by release stuff lately. 15:13:03 korzen: I assume the work will revive in next month. 15:13:42 electrocucaracha: I think we wanted to respin https://review.openstack.org/#/c/352577/ but I see it hasn't happened. do we have an issue with it? 15:14:43 ihrachys: no really, I was expecting that ankur make the changes 15:14:56 ihrachys: but I can contact him later 15:15:19 electrocucaracha: ok not critical, just checking 15:16:00 ok, so expect more reviews for actual objects in next weeks. 15:16:15 moving on 15:16:30 (this meeting will be the shortest ever) 15:16:34 #topic Other patches on review 15:17:12 ihrachys https://review.openstack.org/#/c/377114/ 15:17:44 can you please review this one ? 15:17:55 ihrachys, https://review.openstack.org/#/c/353088/ (just for bookmark I'll resolve conflict riht after this meeting) 15:18:16 ihrachys: Last week we talked about this https://review.openstack.org/#/c/367810/. Did you get a chance to look at it? 15:18:17 quotas ovo (https://review.openstack.org/#/c/338625/) 15:18:46 asingh_: I will, yes. on first sight, what blogan says makes sense. objects are not supposed to communicate with db level entities (queries, db models) except in a limited set of cases. 15:19:31 dasanind_: sadly no. it may take time. now that we have blogan on board with reviews, it should go smoother. ;) 15:20:01 and long standing one https://review.openstack.org/#/c/334695/ 15:20:03 https://review.openstack.org/#/c/369744/ 15:20:13 ihrachys: will wait for your review :) 15:20:20 does this mean blogan can +2 these patches? :) 15:20:32 brianstajkowski1: no!! i do not feel comfortable at all! 15:20:55 brianstajkowski1: at least it means that there will be more people to put -1 now ;) 15:21:05 now, I guess you are all happy about it :) 15:21:18 everyone likes getting -1s! 15:21:45 manjeets: that one, I believe, should be rebased? and you planned to split? anything changed? 15:21:45 from last bits of subnet https://review.openstack.org/#/c/377112/ 15:22:34 yea I can do both ways either this way or split, I guess to be make it less conflictive spliting would be a better idea 15:23:08 subnet ovo integration patch is still failing due to detached db_obj, so no progress there 15:23:09 manjeets: yeah, the lesser the bits the easier to get it merged. 15:23:26 manjeets: note f.e. that Ann or Henry are happy to land model moves but maybe not objects. 15:23:39 ok I'll split by EOD 15:23:47 ihrachys, I noticed that from reviews 15:25:09 #topic Open discussion 15:25:44 I want to optimize my involvement in the effort, particularly spend a tad more time on reviews, so I got rid of some project roles lately, hopefully it will add velocity. 15:26:11 just fyi 15:26:15 ihrachys, indeed it will fast the pace 15:26:19 anything else to share? 15:26:32 clarification, for getting context, we need to pass it from calling methods rather than importing right? 15:26:34 ihrachys, I have one not specifically upgrades 15:26:48 manjeets: if it's not upgrades it's not the venue :) 15:27:02 ok i'll ask on neutron channel 15:27:07 np 15:27:29 sshank: context is passed to us from api layer I believe, so it's then up to us to stick to its constraints. 15:27:43 ideally, this week will have all the relocation patches merged right? 15:27:45 sshank: meaning, you probably can't reconstruct it from inside the plugin 15:27:53 blogan: do I make sense re context building? 15:28:11 electrocucaracha: ideal doesn't happen. never. but yea. 15:28:18 ihrachys: yes, but sometimes the original code is importing context and getting admin context, in which case i dont see why we'd change that 15:28:28 probably woudln't even be able to get the api context 15:29:03 we change definition ideal all the times .... electrocucaracha 15:29:24 yeah, in those limited cases we may decide that admin context is the right one (e.g. if later we apply tenant based filtering on our own) but in general case, we are supposed to stick to whatever is passed to us from api layer. 15:29:36 ihrachys: agreed 15:29:50 ihrachys, I've seen that armax has posted a Barcelona session ideas, are you planning to add upgrades on the list? 15:30:02 korzen, +1 15:30:10 +1 15:30:38 korzen: do we have anything to discuss in the room full of people? 15:30:51 korzen: I mean, looks like we have a lot of work to just act on ready decisions 15:31:05 no API downtime ideas, stopping contracting migration during Ocata cycle 15:31:20 and I am still due to post a detailed description of upgrade process, but that could be in scope of a spec. 15:31:49 ihrachys, +1 15:32:04 rossella_s: define what you are +1ing :) 15:32:11 in my opinion we should repeat what has been told in mid-cycle 15:32:11 ihrachys, spec 15:32:24 rossella_s: and as for summit plans?.. 15:32:34 ihrachys, I don't think we need a session 15:33:17 korzen: what would be the goal of such a session? 15:33:31 korzen: raising awareness? seeking for comments on a proposal? 15:33:37 inform operators 15:33:39 yes 15:34:13 maybe it wont take a whole session 15:34:15 well it seems like we don't have much to *inform* them about. I mean, we have not delivered. 15:34:23 that sounds more like a main summit presentation on neutron upgrades 15:34:29 but at least it should be mentioned somewhere 15:34:30 once its finished 15:34:50 blogan: right. would be cool to deliver AND have a presentation on the fact :) 15:35:10 :) 15:35:12 korzen: ok, maybe I should sync with HenryG on his db related plans for the summit 15:35:13 ok, as long as we deliver it, I'm fine with it 15:35:22 I know that we're not measuring the performance impact about implementing OVO but AFAIK osic-qa will do 15:35:41 just a fyi 15:35:50 we can plan testing right after it is implemented 15:35:55 electrocucaracha: cool. I think we also have some efforts around RDO scheduled for this cycle that should help to track it too 15:36:13 manjeets: electrocucaracha: we are looking more into continuous perf impact validation 15:36:20 with trends per patch and such 15:36:26 we should put attention on new features that are delivered in Ocata to have OVO as DB access 15:36:50 korzen: right. the last cycle it did not work in some cases. 15:37:02 korzen: though vlan-aware-vms was notably objectified :) 15:37:12 so it's mixed, not a total disaster in this regard. 15:37:22 I have spotted for example SubnetServiceType feature that was deliced 'old-style' 15:37:46 yes, I'm happy seeing the VLAN-aware to ne OVO'ed 15:37:51 s/ne/be/ 15:38:27 s/deliced/delivered 15:38:33 :) 15:38:41 now I decoded it 15:39:02 ;) 15:39:13 SubnetServiceType had lice? 15:39:18 :0 15:39:41 that's what you get when you don't OVO 15:39:44 I had OVOed it :) 15:39:53 https://review.openstack.org/#/c/375536/ 15:40:03 korzen: I noticed :P 15:40:14 korzen: mmm nice 15:40:40 that's quite small, should be easy to land :) 15:41:06 dasm has -1, but do not bother 15:41:41 ok, so tl;dr from summit discussion: I will talk to HenryG about database related sessions and how to raise upgrade topics there. in the meantime, you can propose your ideas in the etherpad. 15:43:00 ok cool to see you all, I start feeling we add steam to relevant reviews lately, we should be able to land more in next weeks. 15:43:07 * ihrachys calls it a day 15:43:11 #endmeeting