08:01:41 <jose_lausuch1> #startmeeting Functest weekly meeting May 30th 2017
08:01:41 <collabot`> Meeting started Tue May 30 08:01:41 2017 UTC.  The chair is jose_lausuch1. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:01:41 <collabot`> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:01:41 <collabot`> The meeting name has been set to 'functest_weekly_meeting_may_30th_2017'
08:01:57 <jose_lausuch1> #topic role call
08:02:02 <jose_lausuch1> #info Jose Lausuch
08:02:09 <SerenaFeng> #info SerenaFeng
08:02:29 <LindaWang> #info LindaWang
08:03:04 <morgan_orange> #info Morgan Richomme
08:04:03 <jose_lausuch1> #chair morgan_orange
08:04:03 <collabot`> Current chairs: jose_lausuch1 morgan_orange
08:04:07 <jose_lausuch1> #info agenda for today: https://wiki.opnfv.org/display/functest/Functest+Meeting#FunctestMeeting-30/05(8UTC)
08:04:12 <Shuya_ool> #info Shuya(OOL)
08:04:20 <boucherv> #info Valentin Boucher
08:04:29 <jose_lausuch> #chair jose_lausuch
08:04:31 <morgan_orange> jose_lausuch1: I added some points and just saw that I did not save
08:04:38 <ollivier> #info Cédric
08:04:44 <morgan_orange> so feel free to discard if you want
08:04:44 <jose_lausuch> morgan_orange: ok, what points?
08:04:51 <jose_lausuch> #topic Action Point follow-up
08:04:56 <jose_lausuch> #info AP: morgan_orange: contact Dave (Urschatz) to see if container suite could be done with functest/cengn mentor
08:05:01 <morgan_orange> Ci status and I put 2 links corresponding to the doc generated through pylint
08:05:13 <morgan_orange> #info done, mail sent wait for feedback
08:05:28 <jose_lausuch> ok, we can add them
08:05:47 <jose_lausuch> #info AP HelenYao prepare wiki page for Functest APi proposal (CLI / yardstick approach)
08:06:20 <jose_lausuch> she is not in today
08:06:35 <jose_lausuch> #action HelenYao prepare wiki page for Functest APi proposal (CLI / yardstick approach)
08:06:39 <jose_lausuch> #info SerenaFeng jose_lausuch review https://gerrit.opnfv.org/gerrit/#/c/35179/1
08:06:51 <jose_lausuch> morgan_orange: I think there are some comments you have to provide a new patch, right?
08:07:16 <morgan_orange> jose_lausuch: yes
08:07:24 <jose_lausuch> maybe we can merge it today
08:07:26 <morgan_orange> 2 comments, it is a high level view with empty sections
08:07:27 <jose_lausuch> when are you going to china?
08:07:33 <morgan_orange> but i can already take into account the first comments
08:07:43 <morgan_orange> and I need to change the dev guid on fucntest repo accordingly
08:07:51 <jose_lausuch> #action morgan_orange update https://gerrit.opnfv.org/gerrit/#/c/35179/1 and merge
08:08:00 <morgan_orange> I take off on the 6th of June
08:08:20 <morgan_orange> ok I will do it
08:08:34 <SerenaFeng> welcome to China :)
08:08:43 <jose_lausuch> ah ok next week
08:08:44 <jose_lausuch> sure
08:08:56 <morgan_orange> SerenaFeng: 谢谢
08:09:08 <SerenaFeng> 不客气
08:09:19 <jose_lausuch> #topic Progress on framework
08:09:47 <jose_lausuch> we have to learn chinese
08:10:27 <jose_lausuch> #info Framework guide by Cedric online: http://testresults.opnfv.org/functest/framework/
08:10:28 <morgan_orange> patch on VNF refactoring available for review https://gerrit.opnfv.org/gerrit/#/c/35283/
08:10:38 <morgan_orange> #info patch on VNF refactoring available for review https://gerrit.opnfv.org/gerrit/#/c/35283/
08:10:43 <jose_lausuch> #info API docs http://testresults.opnfv.org/functest/apidoc/apidoc/functest.core.html
08:11:00 <jose_lausuch> what was the problem with the -1 from jenkins yesterday?
08:11:05 <morgan_orange> #info API docs http://testresults.opnfv.org/functest/apidoc/
08:11:14 <ollivier> jose_lausuch: we do modify job to publish them as artifact (functest verify)
08:11:39 <morgan_orange> on my patch it was a doc issue (bad format in docsting => tox told me it locally but I did not see it :))
08:11:49 <ollivier> first Jenkins was down.
08:12:02 <jose_lausuch> aha
08:12:12 <jose_lausuch> couldn't follow much yesterday, sorry
08:12:17 <morgan_orange> too many open files see mail from Cedric yesterday...
08:12:24 <morgan_orange> a nice java exception...
08:13:30 <morgan_orange> the last jenkins update seems to habe brought some instabilities
08:13:45 <morgan_orange> but at the moment it sounds better..
08:13:56 <morgan_orange> if you could review the vnf patch, it will be cool
08:14:10 <morgan_orange> I would like to test the VNF implementation (did not do it so far..)
08:14:20 <jose_lausuch> ok
08:14:46 <jose_lausuch> #action review VNF class from morgan_orange
08:14:50 <jose_lausuch> we have to discuss https://gerrit.opnfv.org/gerrit/#/c/35237/
08:15:24 <jose_lausuch> so, my concern was
08:15:48 <jose_lausuch> the initial idea that was discussed in the release plan is to use snaps libraries with self-cleaning
08:16:02 <jose_lausuch> and we get rid of openstack-snapshot and openstack-clean
08:16:20 <jose_lausuch> because with our current snapshot/clean it is not possible to run parallel tests for example
08:16:31 <jose_lausuch> it is possible, but there can be issues cleaning things that shouldn't
08:16:43 <jose_lausuch> so, we agreed on using snaps library
08:16:45 <jose_lausuch> ok so far
08:17:11 <jose_lausuch> the refactor that cedric proposes is very good if we want to continue the snapshot approach
08:17:15 <jose_lausuch> but it's not the case
08:17:17 <jose_lausuch> what do we do?
08:17:20 <morgan_orange> please note that at the moment on CI several scenarios are failed due to snaps smoke errors
08:17:27 <jose_lausuch> we merge and we remove it later when all the tests are snap-arized?
08:18:04 <ollivier> My patch is not related with any snaps integration...
08:18:10 <jose_lausuch> I know I know
08:18:22 <jose_lausuch> but we will have to remove it later on anyway
08:18:24 <jose_lausuch> that's my concern
08:18:51 <jose_lausuch> we won't probably need that OSGC class later, because we won't run the current snapshot/clean
08:19:17 <jose_lausuch> the patch is fine for today
08:19:26 <jose_lausuch> but it
08:19:42 <jose_lausuch> but it's extra work if we consider where we want to be in 1 month or so
08:19:45 <LindaWang> Are there any progress about reusing snaps for other testcases besides vping?
08:19:57 <jose_lausuch> I don't think so...
08:20:15 <SerenaFeng> with regards to how to use snaps, I think we need to study it first
08:20:44 <ollivier> +1
08:20:53 <SerenaFeng> yesterday I try to fix snaps smoke issue, it is difficult to do it because I have no idea whow to do
08:21:10 <LindaWang> There is an issue about unit test, when I tried to reuse get_endpoint in snaps for odl.
08:21:42 <jose_lausuch> shall we have a session with Steven during the summit?
08:22:02 <SerenaFeng> can we arrange a moment in Beijing Summit let Steven or someone else to give us a training about snaps
08:22:03 <ollivier> snaps must be declared in requirements.txt. Badly the patch was abandonned.
08:22:25 <ollivier> Steven and I have fixed its setup.py to do so
08:23:07 <jose_lausuch> which patch ?
08:23:32 <jose_lausuch> #action jose_lausuch ask Steven to propose a session with Functest team to teach how to use snaps
08:23:39 <LindaWang> https://gerrit.opnfv.org/gerrit/#/c/34909/
08:23:40 <LindaWang> jose_lausuch:
08:24:40 <jose_lausuch> ah ok
08:24:41 <jose_lausuch> yes
08:24:50 <jose_lausuch> too many different things in the same patch
08:25:09 <morgan_orange> we may have been a bit optimistic on the snaps integration. There are hidden impacts and even if we are supposed to contribute, it is not trivial...
08:25:25 <morgan_orange> and we may put lots of pressure on Steven :)
08:25:45 <ollivier> yes. But it must be declared in requirements.txt. I can apply it and I rebase his pending change.
08:26:16 <jose_lausuch> ollivier: https://gerrit.opnfv.org/gerrit/#/c/35411/
08:26:22 <morgan_orange> for vnf refactoring, I clearly priviledge the framework first and plan to make adaptation for snaps later..
08:26:52 <ollivier> Yes, we can spit it into two changes to help Linda merging her change
08:26:54 <jose_lausuch> the conversion to snaps should be for milestone "testcases ready"
08:26:59 <jose_lausuch> not framework ready
08:27:02 <jose_lausuch> so there is still time
08:27:44 <morgan_orange> it is my concern, snaps is testcase and framework, it has impact on framework
08:28:13 <morgan_orange> for instance regarding the work done on pep8 and pylint, I do not think it is OK on snaps side
08:28:18 <morgan_orange> our utils are not ok as well..
08:29:00 <jose_lausuch> yes, that's right
08:29:05 <LindaWang> I still need to update openstack_utils to enable https insecure.
08:29:20 <ollivier> About https://gerrit.opnfv.org/gerrit/#/c/35411/2
08:29:21 <jose_lausuch> we can force that in the next release (pep8+pylint) for snaps repo
08:29:40 <ollivier> Do not merge it, it modifies the framework too and I disagree the change.
08:30:01 <jose_lausuch> ollivier: ok, no problem. place your comments there
08:30:13 <ollivier> already done in abandonned change
08:31:06 <jose_lausuch> so, what do we do? what is our strategy?
08:31:14 <jose_lausuch> we keep our openstack_utils + snapshot+clean?
08:31:29 <jose_lausuch> it's very weak...
08:31:37 <jose_lausuch> our utils are a bit messy
08:31:43 <ollivier> LindaWang: I think we shouldn't link ODL with snaps. Get_endpoint shoudn't be part of snaps
08:32:02 <SerenaFeng> agree, we keep it for now, and let somebody study snaps first, make a proposal
08:32:09 <LindaWang> ollivier:  Why do you think so?
08:32:15 <SerenaFeng> after we review the proposal we can start to work on it
08:32:18 <ollivier> LindaWang: why should we use a middleware instead of keystone client for that.
08:32:30 <morgan_orange> yes it is messy but there is still lots of unknown areas on the generalization of snaps
08:32:42 <ollivier> LindaWang: to simply rely on keystone client.
08:32:44 <morgan_orange> it is not a simple replacement of the utils
08:33:18 <ollivier> Why not using CONST as you proposed. But I think it's shouldn't be linked with SNAPS for such a detail.
08:33:29 <LindaWang> OK, thank you for advice. I will make adjustment.
08:33:36 <morgan_orange> let's organize a discussion with Steven during the summit and share our experience.
08:33:37 <ollivier> perfect. Thank you
08:33:37 <SerenaFeng> it is hard to decide substituting the current framework without investigation
08:33:59 <ollivier> LindaWang:  then you can merge it even if snaps is not listed in requirements.txt :)
08:34:21 <jose_lausuch> we are not substituting our framework, just the openstack calls
08:34:30 <ollivier> I noted there are pending issues regarding logging in snaps too.
08:34:33 <LindaWang> ollivier:  I will try.
08:34:51 <ollivier> LindaWang: Thank you.
08:35:15 <morgan_orange> for me it is still a goal to use snaps to replace our weak parts but we should have a better understanding on the consequences on the generalization
08:35:36 <SerenaFeng> agree
08:35:42 <ollivier> #agree
08:35:46 <jose_lausuch> ok
08:35:50 <LindaWang> +1
08:35:58 <boucherv> +1
08:36:30 <jose_lausuch> shall we try to make better our utils then?
08:36:42 <jose_lausuch> shall we try to make better our utils then?
08:37:13 <LindaWang> I was confused with the file snaps_utils.py in functest. Now I have known more about it.
08:38:31 <jose_lausuch> if we go this way
08:38:32 <morgan_orange> the key question: where shall we put our efforts: update our utils or work on snaps integration. I have the intuition that it should be on snaps....I planned to do it on vnf refactoring once the patch related to the framework was accepted.. :)
08:38:37 <jose_lausuch> we still need to modify our test cases
08:38:42 <jose_lausuch> to clean what they produce
08:39:00 <jose_lausuch> according to cedric approach https://gerrit.opnfv.org/gerrit/#/c/35237/
08:39:08 <jose_lausuch> instead of functest taking a snapshot and cleaning later
08:39:25 <ollivier> There are 800 pylint violations between openstack_utils and its related unit tests
08:39:35 <jose_lausuch> I think we agreed on using snaps already but I see it's not the case :)
08:39:45 <jose_lausuch> 800? cool...
08:40:15 <morgan_orange> how do you see the link with the clean() and snaps? clean is part of the frameowrk and generic
08:40:42 <morgan_orange> snaps is for openstack related tests, it may include some cleaning, it does not prevent to have a clean method in testcase
08:40:50 <jose_lausuch> clean() can be part of the TestCase class, that's not for discussion, I think it makes a lot of sense
08:41:02 <jose_lausuch> exactly
08:41:26 <jose_lausuch> but what I wanted to get rid of for this release is our cleaning mechanism
08:41:31 <jose_lausuch> with snapshot+clean
08:42:03 <jose_lausuch> each test case should clean its own stuff, not the framework
08:42:16 <morgan_orange> that was the first proposal from Cedric ....
08:42:21 <ollivier> +10
08:42:35 <SerenaFeng> +1
08:42:39 <jose_lausuch> so
08:42:41 <jose_lausuch> 2 options
08:42:47 <jose_lausuch> either we do it with snaps, or with our own utils
08:42:56 <jose_lausuch> but we need to change that behavior for this release for sure
08:43:27 <LindaWang> Then  our testcases can be run parallelly?
08:43:45 <jose_lausuch> yes, that is one of the immediate good consequences
08:44:08 <jose_lausuch> we can probably gain in efficiency
08:44:26 <jose_lausuch> and run light test in parallel (for example).. but that's for the next release. First things first
08:45:06 <jose_lausuch> but if we do this in Euphrates (test clean itself), we won't need OSGC class, right?
08:45:37 <ollivier> as It's written and tested, we should keep it... smooth transistion.
08:46:13 <jose_lausuch> ollivier: and do you agree that we can remove it later if not needed any more? (in case snapshot is not longer used)
08:47:14 <LindaWang> Actually I often use the cli "functest openstack snapshot-create" locally.
08:47:22 <ollivier> both are possible. We can keep it anyway to offer possibilities.
08:47:37 <morgan_orange> we do not force to use it :)
08:48:03 <jose_lausuch> the reason behind is that snapshot is not a good approach for test cases
08:48:05 <jose_lausuch> but ok
08:48:10 <jose_lausuch> it seems there is agreement
08:48:23 <jose_lausuch> please review https://gerrit.opnfv.org/gerrit/#/c/35237/
08:48:35 <jose_lausuch> and we can continue with next changes
08:48:51 <ollivier> could we discuss about https://gerrit.opnfv.org/gerrit/#/c/35183/ ?
08:49:04 <jose_lausuch> yes
08:49:39 <jose_lausuch> why don't you like opnfv as a module? it's a central opnfv library
08:49:44 <jose_lausuch> opnfv_modules is too long
08:49:48 <jose_lausuch> and I like the way to import it
08:50:01 <jose_lausuch> from opnfv.X import Y
08:50:18 <ollivier> I think it doesn't change that
08:51:33 <jose_lausuch> mm
08:51:35 <jose_lausuch> what is this? https://git.opnfv.org/releng/tree/setup.py
08:51:43 <jose_lausuch> why do we need that setup there?
08:51:44 <ollivier> I tested in our tox env.. but I may be wrong..
08:52:14 <ollivier> Names are duplicated... I simply fix the bug
08:52:29 <jose_lausuch> the one that should be changed is that one, not the setup in /modules/
08:52:53 <jose_lausuch> actually I don't know where that comes from https://git.opnfv.org/releng/tree/setup.py
08:53:13 <ollivier> I disagree. We copied/pasted it wtihout modifying the name
08:53:26 <ollivier> we completed the previous change
08:53:31 <morgan_orange> arghhh apparently I am the guilty guy
08:53:50 <jose_lausuch> ?
08:54:12 <jose_lausuch> why?
08:54:54 <morgan_orange> I think I created the setup.py in the modules by copying the one that already existed in releng (at root)
08:55:19 <jose_lausuch> no
08:55:24 <jose_lausuch> the original one is in modules
08:55:37 <jose_lausuch> that is the one used to install the python util libraries, like the installer stuff
08:55:47 <morgan_orange> argh so I do not know what the one at root means..
08:56:05 <jose_lausuch> if you install the setup in root, it doesn't work
08:56:08 <jose_lausuch> you can't import opnfv
08:56:13 <jose_lausuch> becuase it doesn't install anything
08:56:19 <jose_lausuch> the one that works is the setup in modules
08:56:40 <jose_lausuch> and I worn't like to change the name of that one
08:56:59 <jose_lausuch> it's used by functest, sfc, sdnvpn, barometer, ...
08:57:10 <SerenaFeng> I might know the reason
08:57:38 <jose_lausuch> why?
08:57:42 <SerenaFeng> it is almost useless, put it there because it is required by tox
08:57:56 <jose_lausuch> then let's modify that name to something else
08:57:59 <jose_lausuch> releng or whatever
08:58:01 <SerenaFeng> I think we can change the name there
08:58:07 <SerenaFeng> yes, agree
08:58:09 <jose_lausuch> ollivier: can you do that in your patch?
08:58:16 <jose_lausuch> modify the root setup, not the one in modules
08:58:32 <jose_lausuch> SerenaFeng: can we name it "releng" ?
08:58:39 <SerenaFeng> change it to 'releng'
08:58:54 <SerenaFeng> sure, we can
08:59:09 <SerenaFeng> it doesn't do actually installing work
09:00:09 <jose_lausuch> ok
09:00:10 <jose_lausuch> thanks
09:00:14 <jose_lausuch> anything else?
09:00:19 <morgan_orange> ok to change the description on the root file. Cedric just tested whatever the name you indicate whatever_you_want (assumling it is not duplicated) you cna still import your module as you used to do import opnfv
09:00:45 <SerenaFeng> about the patch: https://gerrit.opnfv.org/gerrit/#/c/35181/
09:00:49 <ollivier> I simply change the egg file and avoids resing the previous one.
09:01:00 <SerenaFeng> can you all review it?
09:01:12 <jose_lausuch> SerenaFeng: sure
09:01:12 <morgan_orange> so basically here the patch is OK (just fix the duplication name)  but it should be releng and releng_module to be more consistant
09:01:24 <ollivier> ok. I update my change.
09:01:25 <jose_lausuch> morgan_orange: no
09:01:29 <jose_lausuch> please keep opnfv in /modules/ dir
09:01:43 <jose_lausuch> don't change that one please
09:01:48 <jose_lausuch> just the root setup
09:01:51 <morgan_orange> a priori it has no impact
09:01:53 <SerenaFeng> I think the root setup.py is more releng related
09:02:24 <morgan_orange> even with opnfv_modules, you do not need to change the import in any class using this module
09:02:39 <SerenaFeng> I checked the tox.ini, it checks the jjb format
09:02:45 <jose_lausuch> ya, but what changing something that it doesn't need to?
09:03:59 <ollivier> ofc we need to... Why do we name all gnome packages (gedit, gclac) gnome?
09:04:00 <SerenaFeng> and also executing unit tests in modules
09:05:01 <morgan_orange> ok it is already 11:04
09:05:39 <jose_lausuch> yes
09:05:40 <morgan_orange> we did not speak about CI, it seems that on master most of the errors are linked to snaps smoke...it impacts at least compass and daisy
09:05:44 <jose_lausuch> I guess tea time for you :)
09:05:46 <morgan_orange> but let's continue offline
09:05:54 <jose_lausuch> but we had good discussions today
09:05:59 <morgan_orange> I missed my tea break, i have to survive until 12
09:06:13 <jose_lausuch> you can still do it
09:06:15 <jose_lausuch> ok
09:06:17 <jose_lausuch> thanks for today
09:06:19 <jose_lausuch> #endmeeting