14:04:07 #startmeeting neutron_qos 14:04:08 ajo: pull me in if needed…I am pulled into some other meeting 14:04:08 Meeting started Wed Jul 8 14:04:07 2015 UTC and is due to finish in 60 minutes. The chair is ajo. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:04:09 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:04:11 The meeting name has been set to 'neutron_qos' 14:04:16 armax, ack, thank you 14:04:35 #topic Announcements 14:04:52 We had the QoS mid cycle in Israel last week, 14:05:10 a little report here: http://www.ajo.es/post/123458887419/neutron-quality-of-service-coding-sprint 14:05:17 #link http://www.ajo.es/post/123458887419/neutron-quality-of-service-coding-sprint QoS mid cycle report 14:05:45 it was very nice to collaborate in sprint mode 14:05:53 lots of patches merged 14:06:50 not everything fully tested because we wanted to assemble a fast POC, and then iterate over TO-DO's and missing tests 14:07:08 regarding that 14:07:15 #topic Testing gaps 14:07:21 #link https://etherpad.openstack.org/p/neutron-qos-testing-gaps 14:07:29 we have put this together today, 14:07:49 defining a list of POC-essential patches, which we may allow to go in without full testing until the POC is complete 14:08:04 any other patches should start to add unit test/functional test/or any relevant tests 14:09:02 and, we should build a checklist of tests to start including as a follow up 14:09:18 ajo, and those who were happy to get their patches in without proper testing/docs, should go back and write those to those patches already merged ;) 14:09:19 as our goal is to get everything working, correctly tested, and merged back to master 14:09:50 yes, or ask for help where they don't have time, and we will find somebody to do it. But ideally the writer of the patch is the best to test it. 14:10:07 (not sure what a testing expert would say about that statement ^) 14:10:13 it will be faster at least 14:10:40 The full list of files that build the POC (+/-) is http://paste.fedoraproject.org/241292/36363675 14:10:44 yes, schedule is tight 14:10:55 #link http://paste.fedoraproject.org/241292/36363675 POC files which may need testing 14:11:12 if I dint' screw it up when doing the git diff :) 14:11:40 Ajo, neutron client changes started unit test cases. Done for policy clis, pending for rule cmds and working on the same 14:11:45 #link http://paste.fedoraproject.org/241295/14363637 the full POC patch (don't try to test as-is, it's only for reference) 14:11:54 ramanjaneya_++ 14:12:19 thanks, that's awesome, we will definitely need them on python-neutronclient, since we don't have a feature branch there. 14:12:38 ramanjaneya_, cool. I believe we won't be able to merge neutronclient changes though until we are merged back into master 14:12:55 #link http://paste.fedoraproject.org/241300/63640691 grep -C 5 "TODO([Qq]o[Ss])" over the previous link 14:12:59 Also please review the patches and let us know if any comments 14:13:20 ramanjaneya_ I will test every command manually soon 14:13:21 ajo on the Agent side patches you are missing https://review.openstack.org/#/c/195441/ 14:13:45 moshele, I didn't consider it essential for a "POC" :) 14:13:52 I don't have a SRIOV card for testing it :) 14:14:05 Ajo, ok. 14:14:06 moshele, but of course it should go in as soon as we verify/fix the full stack 14:14:19 ajo: ok 14:14:41 moshele, but of course, this should be part of the final artwork and get into master, no doubt 14:14:59 hi irenab :) 14:15:05 hey :-) 14:15:09 irenab: https://etherpad.openstack.org/p/neutron-qos-testing-gaps 14:15:26 ajo: I will start with https://review.openstack.org/#/c/195440/ + it's unit tests 14:15:26 you may want to look into this, we were discussing until the full diff, then look at the meeting logs later :) 14:15:51 moshele, ack, put the unit tests in a separate patch 14:16:02 moshele, as we may want 195440 merged ASAP to have the full POC in 14:16:37 ajo: ok 14:16:41 ajo: sorry if you already discussed it, I have a question 14:17:10 irenab, go ahead :) 14:17:12 we have time 14:17:14 is there any gap that need to be coded or just review + unit tests? 14:17:37 irenab, yes, there are gaps to be coded, 14:17:41 rule updates, for example 14:17:55 irenab, all todo (well, most) should go away 14:17:59 at worst scenario we could just disable "PUT" of rules in this version, for example 14:18:11 ok thanks. so the scope for POC is create/delete? 14:18:12 ajo, it won't fly for merge 14:18:14 I believe 14:18:37 ah, you're talking about poc. 14:18:38 ihrachyshka, worst case scenario, I don't know, there was pull not to include updates during reviews 14:18:46 (spec reviews) 14:18:56 ihrachyshka: for poc, 100% no updates 14:19:16 but ok, disallowing PUT of rules won't be necessary I think :) 14:19:30 most of the layers and the design is ready to handle it 14:19:46 and yes, all the TO-DOs 14:19:53 check for the 3rd fpaste link in etherpad 14:19:57 irenab ^ 14:20:25 ihrachyshka, can I ask a question about the etherpad details or could you go over "We also need a high level, per module description of the architecture, with logical split of the changes into separately reviewable pieces." ? 14:20:26 :) 14:21:24 #topic Test/TODO checklist 14:21:32 I should have changed topic before ;) 14:22:01 so far, priorities: 1) minimal POC, 2) add testing 3) Handle the TO-DOs 14:22:19 4) Get everything ready with testing included in patchsets to be merged right away 14:22:40 ihrachyshka ping 14:22:41 ajo, ehm, so now that I think how acceptance into master could look like, I realize that it will be hard for anyone to comprehend all changes in one go. So we'll need to come up with description of the system, per module, and ask people to look at those separately 14:22:58 because just providing 2000+ lines of code won't fly 14:23:20 the plan after POC to move to master? 14:23:30 but that's after we get coverage for existing code and TODOs handled 14:23:30 ihrachyshka, ok, that makes sense, we can assemble an small devref with description of the design 14:23:43 "small" 14:23:47 irenab, no way. we need to cover tests, write docs, handle TODOs 14:23:52 that's a huge piece of work 14:23:57 yes 14:24:07 and we have ~month if we set start of L3 as goal 14:24:22 ajo, devref is needed anyway. we have a stub in the tree. 14:24:22 And we should target L-2 (optimistically) -end of july- or start of L-3, 14:24:32 the later the worst, because L-3 end is feature freeze 14:24:48 yes, if we get later, we risk to slip to M 14:26:00 I guess major concern would be for agent/rpc changes, the rest is quite decoupled from the core neutron 14:26:10 we need to do a final push, and properly communicate that 14:26:31 irenab, correct, but agent/rpc changes are very cool IMHO ;) 14:26:44 good design ;) irenab++ moshele++ 14:26:46 maybe need core reviewers of relevant areas already looking 14:26:59 irenab, yes, and we should make sure all is as clear as possible for those who will decide on getting the code in master (not us) 14:27:09 ago: + gampel 14:27:21 +gampel +gsagie ;) 14:27:37 We don't want to start bikesheeding for merging back, 14:27:47 but we may expect requests to change things 14:28:03 for example, we should sync with the new callback interface armax is working on to extend resources 14:28:11 instead of using AFTER_READ dict extension 14:28:33 ajo: this this already for review? 14:28:45 I mean armax change 14:28:53 ajo, I believe this one can be actually postponed to when it's in master 14:28:55 irenab, no, it's just a plan, I think he started it and will send for review, we talked about it yesterday 14:29:11 cool, thanks 14:29:20 ihrachyshka, it could be, as long as the long-term plan is clear.. I'm ok with that 14:29:32 so, 14:29:36 with this plan in mind 14:29:56 we want to talk to mestery to make sure everything is aligned 14:30:09 any comments about the plan? 14:30:55 moshele, irenab , sc68cal , vhoward , will you have some time to write tests? :) 14:31:10 otherwise, ping for help fast, so we can look for resources, we don't want to risk delaying to M 14:31:29 ajo: I will need help with tests 14:31:34 also, to make things funnier, I'll be PTO since Jul 20th to Aug 3 14:31:49 ajo: I am afraid I have limited cycles, but can try to help. 14:32:16 Ok, I will try to pull in help from Red Hat 14:33:23 moshele, irenab , please identify the modules you know about which need testing, and ping us with what you won't be able to test, so we can find stakeholders :) 14:33:42 you know about -> you participated on development 14:33:56 ajo: can we include the rule_types patch in the POC scope? Will be nice to get some feedback 14:34:39 irenab, It needs to be finished, isn't it? :) right now it has a few "pass" which will do nothing 14:34:55 if I'm not wrong 14:35:07 the API is there, so it need to be connected with ihrachyshka code 14:35:23 via service_plugin. it's in review. 14:35:42 so if we merge the API, it can be connected together 14:35:57 just not to make more dependencies between patches 14:36:09 yes, we need to merge more patches to clear the dependency mess :=) 14:36:34 * ajo had a nice day today assembling the POC and going back to rebase patches because of horrendous conflicts while cherry-piling 14:37:11 irenab, when we have the integration ping me, we will need to write the tests anyway soon :) 14:37:19 link to the patch https://review.openstack.org/#/c/199066/ 14:37:50 ajo: sure 14:38:13 very low attendance today, 14:38:25 ajo, the problem with your poc is it seems the deps are broken. (or at least the order): I see one failure that suggests you use some modules in a patch that are not yet avail at that point in your mountain 14:38:42 ihrachyshka, it probably is 14:38:45 I suspect we should have better merged disconnected pieces and then integrate 14:38:53 I need to refine it during today/tomorrow to verify it's really a poc 14:38:57 ihrachyshka: +1 14:38:58 right now, it's just "a patch" 14:38:58 integrating in git is... fun 14:39:19 it's good to find out what we have done, and what's missing testing 14:39:20 but I will discover soon 14:39:24 ajo, yeah, but you updated existing, mergeable versions of patches 14:39:32 anyway 14:39:34 I just finished cherry-piling and rebasing stuff before themeeting 14:39:37 I trust you solve it :) 14:39:45 * ajo crosses fingers 14:40:24 ok, any more comments about this topic? 14:40:27 we need to run! ;) 14:40:58 ok 14:41:05 another topic which it's been around 14:41:10 #topic Summit presentation(s) 14:41:26 vikram approached me about working on a QoS presentation for next summit 14:41:39 I believe moshele , irenab , you may want to be in 14:41:44 yep 14:42:00 I will collaborate with content, but probably I won't attend this summit for family reasons 14:42:18 I need to check, but probably will 14:42:23 * ajo will envy all the Tokio visitors :) 14:42:55 It could be interesting even to work in two different sessions 14:43:08 one about general usage, and explaining what have we done, and how it works 14:43:21 and another about lower level, and... explaining vendors how to integrate their plugins with QoS 14:43:40 we may want to loop Eran about the 2nd one :) 14:43:58 and Gal 14:44:10 and Gal, true 14:44:21 we may divide people between both presentations, not to clutter the stage :) 14:44:54 cool, loop me in for content part for sure. And would love to participate in presentation if will attend 14:45:05 moshele, irenab , probably you must be a must for the user facing one (more visible) 14:45:27 API? 14:45:40 yes I think we should do two probably? 14:45:55 1) API explanation, CLI usage, configuration , etc... 14:46:12 2) low level integration for other vendors 14:46:23 may be 45m is enough for both? 14:46:25 I'm not sure 14:46:33 what do you mean by low level? 14:46:48 write your own qos plugin? 14:46:50 2) explain the internal design of the service, 14:47:05 explain how to write a backend driver for qos 14:47:29 just an idea, 14:47:38 I will put everybody interested in a mail thread 14:47:49 if we have Midokura support by then, it will be cool to show….I need to check 14:47:52 so we can decide how to do it 14:47:57 irenab++ 14:48:01 that'd be super awesome :D 14:48:24 we want to show also the agent side ? 14:48:36 for other ml2 agents ? 14:48:53 moshele, probably, as an example to others who use ovs, for example... or sr-iov 14:48:54 I think yes, end to end. Can be both sriov and ovs 14:49:00 yep 14:49:12 cool 14:49:57 There is talks submission deadline July 15 14:50:06 yes, time is tight :) 14:50:16 as always :-) 14:50:28 I'm too toasted now to even write an email, but I can do it first thing tomorrow morning 14:50:31 or ping vikram to do it 14:50:46 toasted = 35-37ºC 14:50:57 generally I don't need A/C in my office, but this summer is horrible! 14:51:23 ok 14:51:31 anything else, or shall we close the meeting? 14:51:55 I will also communicate our plan in openstack-dev with CC to mestery to see how it flies 14:52:09 #action ajo write email about presentations to all interested parties 14:52:27 or ihrachyshka , do you want to do that 2nd part? :) 14:52:32 I was about to write an action to myself.. 14:52:45 ajo, which part, sorry. presentation? 14:52:58 ihrachyshka, no, I mean, communicating or feature/qos plan to openstack-dev 14:53:10 to get it merged back to master by L2/L3.start 14:53:18 L-2, L-3 :D 14:53:19 hm, not sure I will have it today. let's sync tomorrow. 14:53:23 L2/L3 seems like OSI layers 14:53:31 ihrachyshka, tomorrow is ok, 14:53:36 I can come up with draft, then we'll discuss 14:53:39 it's a bit late already for CEST 14:53:40 ack 14:54:14 #action ihrachyshka communicating feature/qos plan to openstack-dev towards merging back to master 14:54:33 yes, my load 14:54:36 *lord 14:54:39 %) 14:54:40 thanks ihrachyshka :D 14:55:49 omg: https://weather.yahoo.com/spain/madrid/madrid-766273/# 39ºC 14:55:50 ok 14:55:55 we can close the meeting :) 14:56:02 o/ 14:56:07 ajo: time to go to the pool :-) 14:56:15 thanks moshele , irenab , ihrachyshka , ramanjaneya_ 14:56:21 cu. and everyone, keep bombing us with patches 14:56:23 yey! 14:56:28 go go go! :) 14:56:29 or to israel 14:56:43 bye 14:56:45 yeah, this weather is insane 14:56:45 moshele, yeah, Israel seems like a cold heaven 14:56:49 Bye 14:56:51 bye 14:56:55 yeah, the weather was very nice ;) 14:57:03 #endmeeting