15:01:12 #startmeeting neutron_upgrades 15:01:13 Meeting started Mon Jan 9 15:01:12 2017 UTC and is due to finish in 60 minutes. The chair is korzen. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:15 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:17 The meeting name has been set to 'neutron_upgrades' 15:01:19 hi all 15:01:28 Hello 15:01:33 Hello 15:01:45 hello in first team meeting in 2017 15:01:47 * dasm is sitting in corner, and watching at world in flames... 15:02:05 o/ 15:02:23 hope you had a nice holidays and recharged your batteries 15:02:45 taking a look at agenda 15:02:47 #link https://wiki.openstack.org/wiki/Meetings/Neutron-Upgrades-Subteam Agenda 15:02:51 hi 15:02:59 #topic Announcements 15:03:24 we have 3 weeks left till ocata-3 15:03:32 which is 25 Jan 15:04:04 so we better settle down today what is important for this last weeks to complete in Ocata 15:04:27 ihrachys, any other announcements? 15:04:45 I will just update everyone that I am now in PCT (West Coast) time zone 15:05:00 but I start early (6am early) so there is time to sync with Europe too 15:05:05 nothing else 15:05:11 ihrachys: welcome in "brave new world"! 15:05:19 thanks 15:05:31 good to hear for US folks, more time for us to sync with Ihar :) 15:05:39 for US 15:06:03 ok moving on 15:06:10 #topic Partial Multinode Grenade 15:06:43 linux gate is failing 15:06:56 #link http://grafana.openstack.org/dashboard/db/neutron-failure-rate?panelId=8&fullscreen 15:07:21 if anyone is interested in debugging 15:08:00 it would be nice to take a look 15:08:40 I checked logs before, nothing obvious 15:08:46 does anyone worked on testing mixed version of neutron server running? 15:08:59 so maybe local reproduction could be of use, though multinode devstack setups are not easy 15:09:29 korzen, I think manjeets was trying it out sometime back. 15:09:29 re mixed versions, I haven't looked at it, but I believe that should be a priority for the remaining Ocata time 15:10:05 ideally, we prove it works and then announce that as an experimental feature of the release 15:10:35 it would be nice if someone could step up and test in manual way 15:11:05 I have not synced with dolphm on that matter, how OSIC QA team is progressing on the gating 15:12:09 manjeets, did you have any progress on the mixed version testing? 15:13:01 I guess he is not around 15:13:08 o/ sorry for being late 15:13:19 korzen it did not try it yet 15:13:51 i setup env but was then got side tracked 15:13:59 ok, so If noone has any input, I have work for it this week 15:14:22 trying to setup the multinode devstack and try to perform some manual tests 15:14:38 I can work for it* 15:14:48 I will update you on the next meeting 15:14:54 thanks korzen 15:15:21 #action korzen to setup the mutlinode devstack env and try to test the mixed version of neutron server scenario 15:15:34 ok, moving next 15:15:43 #topic Object implementation 15:16:08 #link https://review.openstack.org/#/q/status:merged+project:openstack/neutron+branch:master+topic:bp/adopt-oslo-versioned-objects-for-db 15:16:19 what have merged since last meeting 15:16:47 so we have FlatAllocation OVO and Port OVO in ml2/db 15:17:28 there are many objects adoption in flight 15:17:50 but the most important one is PortBinding 15:18:04 #link https://review.openstack.org/#/c/407868/ Integration of (Distributed) Port Binding OVO 15:18:26 I believe https://review.openstack.org/#/c/411830/10 is its dependency? 15:18:26 not counting the merge conflict, the patch should be ready to review 15:18:36 #link https://review.openstack.org/#/c/411830/ DictOfMiscValuesField 15:18:40 yes, 15:18:52 and we proved it fixes the issue we had? 15:19:13 the new OVO field type for Port and Agent is handling full pythonic dict 15:19:43 ok I see the port binding patch was blessed by Jenkins. I assume it means it indeed helped. 15:20:22 ihrachys, yes, I have added agent and port binding test cases to 411830 15:20:45 cool, I will bubble it up in my review queue for today. 15:21:05 so, from now on, when some have to use JSON blob as field type, dictOfMiscValuesField should be used 15:21:48 DictOfStringField had this limitation that the values can be only the string, or numbers/bools converted to string 15:22:18 anyone has some important object patches to review? 15:22:41 i would like to see someone looking at this: https://review.openstack.org/#/c/382659/ 15:22:51 it's layer between keystone v3 and ovo 15:22:59 yes, that one 15:23:11 quotas ovo 15:23:27 RouterL3AgentBinding https://review.openstack.org/#/c/360908/ 15:23:29 https://review.openstack.org/#/c/338625/ 15:24:05 https://review.openstack.org/#/c/416352/ to break circular import issue for quotas 15:24:27 until lib release this patch https://review.openstack.org/#/c/416364/ 15:24:57 so as far as our goal for Ocata, unless some feature is not modifying the DB model, the OVO is not critical 15:25:36 like for PortBinding, which will be adding new PK and 'status' field 15:26:01 and I believe that's the only feature that could reasonably land in Ocata right? 15:26:20 yes, I guess that is true 15:26:53 the last thing that we are missing is online data migration framework 15:27:14 but PortBinding for LM is not affected by not having the online data migration 15:27:28 * ihrachys leaves the audience early to attend another meeting, sorry 15:27:52 since, we will add new field that is by default empty string does not have to be online migrated 15:28:24 korzen: and ihrachys : for LM data model changes I will need some review on this patch https://review.openstack.org/#/c/404293/ 15:28:43 this is what I have came up when thinking about data migration 15:29:25 ok dasanind, I will take a look 15:29:37 korzen: thank you 15:29:58 korzen, any ideas on the port binding level synthetic field issue I mentioned in the comments of that patch? 15:30:08 dasanind, It would be nice to test your change in mixed version env 15:30:36 sshank, which one? 15:30:38 korzen: yeah that's right 15:30:48 korzen, https://review.openstack.org/#/c/382037/ 15:31:31 sshank, I did not have time to look at it 15:31:44 Will take a look ASAP 15:31:58 korzen, for mixed versions test how many network nodes will be running 15:31:59 korzen, Okay. Thanks. 15:32:02 2 or 3 ? 15:32:21 manjeets, network nodes should be 2 15:32:47 but the neutron-server instances will be at least 2 15:33:05 and the neutron-server is the most important here 15:33:41 beacause we need to test the API nodes 15:33:47 in mixed version 15:33:54 okay master and newton version of server 15:34:12 the network node should be already tested in grenade multinode 15:34:36 ok, moving next 15:34:44 #topic Other patches on review 15:35:11 since we already touched some of patches, port binding for LM 15:35:28 I sent some links earlier would be great if can get some reviews on those 15:35:35 anyone has some interesting patches to share? 15:35:58 manjeets, yes I will take them in my queue 15:36:16 thank you korzen 15:36:43 I would like to remind you of 15:36:44 #link https://review.openstack.org/336518 devref about OVO in neutron 15:37:09 thanks for all of the reviews, I would need to ping ihrachys to review ^ :) 15:37:34 korzen: ihrachys : https://review.openstack.org/#/c/360799/ 15:38:23 It's almost ready will fix the functional test today:) 15:38:41 ndahiwade, good, added to the queue :) 15:39:17 I wanted also to talk about methods in OVO classes 15:39:20 korzen, thanks:) 15:39:24 Methods in OVO class returning non-OVO output 15:39:35 like here: ○ https://review.openstack.org/#/c/381209/17/neutron/objects/router.py@103 15:39:59 My opinion is that classmethod in OVO class should return OVO 15:40:28 do you have any drawbacks of that assumption? 15:41:05 korzen, i have one quotas 15:41:31 korzen, In that example, I moved it inside the class because it had a complex join. 15:42:23 sshank, yes, and moving it to OVO class is a good idea 15:42:29 korzen, Right now it is returning a list of Router models but since router OVO is not ready yet, it is not returning OVOs 15:42:55 sshank, ok 15:43:01 https://review.openstack.org/#/c/338625/23/neutron/objects/quota.py@94 15:43:24 sshank, so the example is not so good :) 15:43:31 but the main point stays the same 15:43:50 manjeets, good one 15:44:16 manjeets, I was thinking of returning reservation OVO 15:44:29 korzen, I think this is returning non OVO: https://review.openstack.org/#/c/377074/35/neutron/objects/l3agent.py@42 15:44:42 korzen: my understanding is if the class method needs to return OVO we need to translate the model returned by the query to object is that correct? 15:45:26 sshank, yes, we can have some special methods returing ids 15:45:29 sshank, it is fine 15:45:36 korzen, i would trying doing may be more refactoring of code if returns anything instead of dictionary 15:45:37 it is named get_..._ids 15:46:28 dasanind, yes, we need to evaluate query inside OVO classmethod and call load_object on the result 15:46:58 my main point is to not to return query sqlalchemy object 15:47:16 korzen: do you have an example? 15:48:08 dasanind, would need to look for, but it should be smth like: return [self.load_object(x) for x in query)] 15:49:00 korzen: thanks, will try that for https://review.openstack.org/#/c/377074/35/neutron/objects/l3agent.py@42 15:49:21 dasanind, yes, thanks 15:49:53 #topic Open discussion 15:50:07 ok, do you have any more questions comments? 15:52:08 ok, if not, then 15:52:11 #endmeeting