15:01:31 <korzen> #startmeeting neutron_upgrades 15:01:32 <openstack> Meeting started Mon Aug 22 15:01:31 2016 UTC and is due to finish in 60 minutes. The chair is korzen. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:33 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:36 <openstack> The meeting name has been set to 'neutron_upgrades' 15:01:37 <korzen> Hi all 15:01:44 <dasanind_> HI 15:01:56 <ankur-gupta-f> morning all 15:01:57 <korzen> are there anyone from upgrades team? 15:01:58 <john-davidge> hi 15:02:06 <korzen> oh, I see there are some :) 15:02:30 <asingh> hey 15:02:31 <korzen> so today I will chair, since Ihar reported sick 15:02:49 <johndperkins> o/ 15:02:58 <korzen> let's start with agenda 15:03:08 <korzen> #topic Announcements 15:03:27 <korzen> so I hope that everyone returned home safely after midcycle 15:04:03 <ankur-gupta-f> korzen: any big updates from the midcycle? We saw the presentation. Anything else 15:04:08 <korzen> in this section I would like to share the midcycle update 15:04:16 <korzen> yeas 15:04:32 <korzen> so Ihar has presented the slides 15:04:41 <korzen> #link https://docs.google.com/presentation/d/18sqjQhNU_SnzNPy4WGJ4pH9-kaPvzJUVnYOo0Vc9AwU/edit?usp=sharing midcycle slide deck for upgrades 15:05:07 <korzen> the intent of this was to introduce the upgrades topic to other core reviewers 15:05:14 <manjeets__> o/ 15:05:30 <korzen> and generally ask/response to a question - why even care about upgrades 15:05:51 <korzen> does it make any sense 15:06:13 <korzen> the ultimate goal is to have the zero downtime in data and control plane 15:06:59 <korzen> in order to do so, we need to introduce the OVOs and allow to run concurrently different version of Neutron API server 15:07:22 <korzen> current effort is to have the OVOs in place 15:07:39 <korzen> for newton phase, the network, port and subnet have to be ready and merged 15:07:58 <korzen> it is also the requirement for push notification to be done also in Newton 15:08:22 <korzen> no integration code for above objects are required to be merged 15:08:47 <korzen> it would be safer to leave the integration code to be merged at the beginning of Ocata 15:09:15 <manjeets__> korzen: what would be their use without integration ? 15:09:27 <korzen> manjeets__, for push notification 15:09:41 <manjeets__> afaik we can't have microversioning until it integrates 15:09:44 <korzen> push notification is dependent on sending the objects with RPC callbacj 15:09:55 <manjeets__> ok 15:10:18 <korzen> #link https://etherpad.openstack.org/p/newton-neutron-midcycle-workitems the Objects section 15:10:57 <korzen> in the link, there is what we have discussed and action items at the lines 89-96 15:11:34 <korzen> we would also be mergind the relocate mode stuff, which was stacked in patch queue by 4 15:11:50 <korzen> no to create too much git conflicts while merging 15:12:05 <korzen> relocate model* 15:12:47 <manjeets__> korzen: so they won't merge until they are conflicting so many other patches ? 15:13:33 <korzen> so the plan is to merge it very carefully like if their will not break anything critical from release point of view 15:14:17 <ankur-gupta-f> korzen: don't we only have a week or two till the n-3 cutoff? Will network & ports be done by then? 15:14:41 <manjeets__> port is close i guess 15:14:59 <korzen> ankur-gupta-f, we hope to have them merged 15:15:12 <korzen> please review 15:15:35 <korzen> also other core reviewers are doing so 15:15:37 <ankur-gupta-f> ports: https://review.openstack.org/#/c/351368/ networks: https://review.openstack.org/#/c/269658/ 15:16:12 <korzen> the model relocation, OVO implementation and their integration is what we have in the scope right now 15:16:40 <korzen> the next step is to write down the spec how we should proceed with no API downtime 15:16:56 <korzen> it should be written down before BArcelona summit 15:17:18 <korzen> to agree on final design and try to implement it during Ocata 15:17:44 <korzen> fortunately there was johnthetubaguy present in the room 15:17:45 <korzen> , he shared his experience with Nova upgrade story 15:18:00 <manjeets__> korzen: a general question , is there any testing plan to check running different versions of neutron server concurrently 15:18:26 <korzen> manjeets__, that is a good question 15:18:44 <korzen> we do not have that one 15:18:53 <korzen> but nova aslo does not have it 15:18:56 <manjeets__> korzen: you can update etherpad then 15:19:21 <manjeets__> may be i can give it a try later once it OVO is close to complete 15:20:04 <korzen> it is like QA work in my opinion, to have the API server started in 2 nodes 15:20:13 <korzen> etc, more of the cross project effort 15:21:48 <korzen> so sometime in the Ocata timeframe, we should take a bullet and forbid the contract migration 15:22:12 <korzen> and see how it will be working for us 15:23:14 <korzen> some PoC of that should be prepared before we can go to the community and say: contract migration are now forbidden 15:23:43 <manjeets__> korzen: if you know (does other projects like nova also forbid contract migration) ? 15:24:06 <korzen> yes, manjeets__ nova has only additive changes to the schema 15:24:32 <korzen> but they do not have mixed version of software talking to the DB right now 15:24:57 <manjeets__> ok that's less prone to breakage 15:25:01 <korzen> you have some small window of API downtime, where the nova-conductor and nov-api has to be restarted 15:25:56 <korzen> other problem with neutron are plugins 15:26:01 <korzen> and extensions 15:26:29 <korzen> I will take a look how right now the neutron plugins are using the DB and API layer 15:27:06 <korzen> the problem here is the arbitrary values additions to neutron resources 15:27:19 <korzen> and access to the DB model instead of OVO 15:27:38 <manjeets__> yea I had seen that issue 15:27:42 <korzen> API filtering is also kind of free to use with arbitrary strings 15:28:28 <korzen> we should prepare backward compatible dropping off the random filters at API level 15:28:50 <korzen> not to break the plugins that are relying to do so 15:29:53 <korzen> so the plugin API need to be revisited 15:30:18 <korzen> it is also question for operators, what type of plugins they are using 15:30:42 <korzen> refering the official openstack user survey, most of them are using OVS 15:31:51 <korzen> for doing the no api downtime, we would also require having the API microversioning mechanism 15:32:15 <korzen> but that is also not so clear how to be handled for neutron 15:32:46 <manjeets__> korzen: cinder handles it pretty well but their resources are very less as compared to neutron 15:33:17 <korzen> so and the last thing is that NeutronDbObject should be sometime moved into neutron-lib 15:33:40 <korzen> so that any other networking projects can use it 15:33:59 <manjeets__> so would base object api also be moved ? 15:34:12 <korzen> yes, 15:34:44 <korzen> it wont happen like now, but in the long run it should land in neutron-lib 15:35:08 <manjeets__> ok 15:35:18 <korzen> ok, that is what I think is the most important update 15:35:27 <korzen> any questions? 15:35:43 <ankur-gupta-f> korzen: Question along similar lines of priority. This comment from Carl Baldwin on Relocate Tag DB model: "Given that this bug is marked wishlist and is not targeted at Newton, maybe it is best to hold off until Ocata is open for development. Could we hold off on merging this unless the drivers team targets it to Newton?" 15:35:59 <ankur-gupta-f> Should we put -1 workflows on all the Relocate DB model patches? 15:36:36 <korzen> ankur-gupta-f, it is not required for you to put -1 on workflow 15:37:01 <korzen> it is the neutron core reviewer responsibility to look at the conficting patches 15:37:17 <korzen> and merge only that pieces that are safe to Newton deliverables 15:37:46 <korzen> ankur-gupta-f, which patch has carl_baldwin comment? 15:37:49 <ankur-gupta-f> Carl's comment pertains to all the db models relocation. So how should we respond? 15:37:55 <ankur-gupta-f> https://review.openstack.org/#/c/355536/ 15:38:04 <ankur-gupta-f> after ihar and Akihiro had already +2ed it 15:39:06 <korzen> I guess the right thing to do to is to wait till the drivers meeting and ask them ;) 15:39:35 <korzen> after the meeting, if the decision is to -1 all the relocate patches, then we should do it 15:39:55 <korzen> ankur-gupta-f, can you add this item on agenda? 15:40:01 <korzen> on drivers meeting 15:40:16 <korzen> I'm not sure when the nearest driver meeting is.. 15:40:48 <johndperkins> thursday 4pm MST 15:40:58 <ankur-gupta-f> okay I will put it on the agenda 15:41:08 <korzen> ok, thanks ankur-gupta-f 15:41:32 <korzen> ok, any other questions? 15:42:10 <korzen> if not then: #topic Object implementation 15:42:12 <korzen> #topic Object implementation 15:42:58 <korzen> last week Ihar and I had run over the port and network patches 15:43:18 <korzen> so hope they will be merged in Newton 15:43:43 <korzen> I have prepared the multi foreign keys support in NeutronDbObject 15:43:54 <korzen> #link https://review.openstack.org/#/c/357207 objects: add support for per parent type foreign keys 15:44:31 <korzen> anyone has a patch that want to discuss? 15:45:08 <ankur-gupta-f> none from me 15:45:47 <johndperkins> I'm good 15:45:48 <asingh> https://review.openstack.org/#/c/356825 15:46:34 <ankur-gupta-f> korzen: I created a Gerrit Dash to keep track of all the OVO stuff if you want me to share the link 15:46:46 <manjeets__> asingh: what is the question on that patch ? 15:46:54 <korzen> ankur-gupta-f, yes please 15:47:20 <asingh> kozen: standard_attr_id is biginteger, Couldn't find a filed in oslo.versionedobjects which is for BigInteger. Used IntegerField instead 15:47:28 <ankur-gupta-f> https://review.openstack.org/#/dashboard/?foreach=%28%28project%3Aopenstack%2Fneutron%29+branch%3Amaster+status%3Aopen+NOT+status%3Amerged&title=Neutron+OVO&All+My+Patche=%28owner%3Aself+status%3Aopen%29&My+Object+Patches=%28topic%3Abp%2Fadopt%2Doslo%2Dversioned%2Dobjects%2Dfor%2Ddb+owner%3Aself%29&My+DB+Patches=%28topic%3Abug%2F1597913+owner%3Aself%29&Objects+Need+Review=%28topic%3Abp%2Fadopt%2Doslo%2Dversioned%2Dobjects%2Dfor%2Ddb+status% 15:47:42 <johndperkins> nice 15:48:14 <korzen> asingh, the IntegerField is fine 15:48:16 <asingh> krozen: does it seems right? 15:48:23 <manjeets__> korzen: to answer that question oslo.versioned object is using python (int) 15:48:39 <manjeets__> and python's int of arbitary length 15:48:50 <manjeets__> so won't be issue according to me 15:49:07 <korzen> manjeets__, yes, so it is okay to use Integer 15:49:24 <johndperkins> ankur-gupta-f: url got clipped, can you use a shortener? 15:49:41 <korzen> ankur-gupta-f, yes, it is not working for me 15:49:42 <asingh> thanks korzen, manjeets_ 15:50:06 <ankur-gupta-f> https://goo.gl/v8Q4RC 15:50:10 <johndperkins> thnx 15:50:41 <korzen> ankur-gupta-f, yes, now it is ok :) 15:51:06 <korzen> nice, thanks 15:51:58 <korzen> ok, anything else to discuss? 15:52:11 <korzen> #topic Open discussion 15:52:14 <manjeets__> korzen: so until n3 priority is to review port and network ? 15:52:53 <korzen> manjeets__, yes 15:53:05 <johndperkins> I'm going to push more field test patches for ovo like https://review.openstack.org/#/c/357186/ so that should help the later objects, keep an eye out for those 15:53:06 <manjeets__> If drivers meeting decides to hold on relocating models then I think we should not send any more patches until n3 15:53:11 <korzen> but we are free to continue to work on OVO adoption 15:53:57 <manjeets__> ok, for complete adoption those are mandatory 15:55:28 <korzen> ok, so the last thing is that I have limited time to do the review this week, and I will be on PTO till 12th September 15:55:41 <dasanind> manjeets_: korzen_: so the new OVO patches will not be merged then correct? 15:56:14 <manjeets__> new relocation of model patches not for sure if driver meetings decided to hold em 15:56:41 <manjeets__> I guess we can then send introduce OVO patches without integration 15:56:52 <korzen> dasanind, I think that OVO patch wont be priority right now 15:57:24 <korzen> but we should be prepared to fast merged them after Ocata is openede 15:57:29 <korzen> opened* 15:57:58 <dasanind> manjeets_: korzen: thank you 15:58:06 <korzen> ok, we are running out of time 15:58:21 <manjeets__> thanks bye 15:58:28 <korzen> thanks to all of you 15:58:30 <johndperkins> thanks 15:58:31 <korzen> #endmeeting