08:04:16 <ollivier> #startmeeting Functest weekly meeting 19 June. 2018
08:04:16 <collabot_> Meeting started Tue Jun 19 08:04:16 2018 UTC.  The chair is ollivier. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:04:16 <collabot_> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:04:16 <collabot_> The meeting name has been set to 'functest_weekly_meeting_19_june__2018'
08:04:25 <ollivier> #info Cédric Ollivier
08:07:10 <ollivier> juhak: depo: CristinaPauna: Are you free? Else I will simply close it.
08:07:42 <depo> #info Delia Popescu
08:07:43 <depo> yes
08:07:51 <depo> Hi!
08:08:10 <ollivier> hello
08:09:13 <ollivier> let's wait few minutes if Juha is connecting.
08:09:40 <depo> ollivier: sure
08:10:27 <ollivier> depo: else we can focus on the point you're interested in. Everything is already written in the wiki page. https://wiki.opnfv.org/display/functest/Functest+5.+Meeting
08:11:30 <depo> OK, I was to come with an answer regarding barbican support on fuel
08:11:41 <depo> It is not supported now, but it will be
08:12:47 <ollivier> #info Fuel will support Barbican
08:12:53 <ollivier> great news!
08:13:44 <ollivier> so let's start. I will simply list the actions done and we can discuss on the new scenarios which raise side effects on cinder_test
08:13:57 <depo> ok
08:14:21 <ollivier> #topic Action point follow-up:
08:14:33 <ollivier> #link https://jira.opnfv.org/browse/FUNCTEST-976
08:14:45 <ollivier> #info a new opnfv/functest-tempest was published
08:15:00 <ollivier> #info lots of uncovered/untested code have been removed (parser, cli, api, etc.)
08:15:08 <ollivier> #info config flavor metadata can be configured for snaps testcases
08:15:21 <ollivier> #info a new dedicated CI has been setup to allow testing OpenStack master
08:15:49 <ollivier> depo: let me know if you need details about this topics
08:17:02 <depo> yes, last point
08:17:23 <depo> where? how? :)
08:18:10 <ollivier> Apex is deploying master and I don't want to use N-1 (Queens) clients vs N
08:18:40 <ollivier> it would be a mess to build to do so without branches due to the design of Releng (templates + rolling model)
08:19:18 <ollivier> then I reuse my travis-ci jobs to build ollivier/functest-xxx with OpenStack master's constraints
08:19:21 <depo> ah, ok. got it
08:19:35 <ollivier> we don't know yet which testcases could run in master (tempest, rally, etc.)
08:19:57 <ollivier> And as it was discussed here, we can't support 3 branches without fonctional gates
08:20:09 <depo> how would we know?
08:20:27 <ollivier> chicken and egg situation :). We need a full run vs OpenStack master
08:20:51 <depo> thought so... :)))
08:20:57 <ollivier> We already know that vmtp and refstack_client can't work vs Queens.
08:21:00 <depo> ok
08:21:29 <ollivier> Normally we should select all tempest-based +rally testcases. But vnfs could be a real added value.
08:21:39 <ollivier> even for OpenStack gates.
08:22:39 <ollivier> by default, we should select testcases which are run in OpenStack gates else it's more or less impossible to maintain.
08:23:45 <ollivier> Without a full run vs OpenStack master, I would say that smoke could be reused (except snaps smoke). Snaps smoke remains an added value for Gambia.
08:24:26 <depo> ok, so we test and see
08:24:57 <ollivier> yes. That's why a dedicated toolchain is better and I would be a mess to hack the jjbs for that (2 builds per changes)
08:26:07 <ollivier> I'm switching to a new proposal.
08:26:21 <ollivier> #info new scenarios are published in Functest core
08:26:55 <ollivier> it avoids duplicating code as for vnfs and eases writing testcases for endusers
08:27:20 <ollivier> #link https://gerrit.opnfv.org/gerrit/#/c/58501/
08:27:57 <ollivier> Lots of testcases (vnfs) require one vm rechable via ssh
08:28:07 <ollivier> (+vping and conder)
08:28:09 <depo> yes, I saw it, it looks good
08:29:00 <ollivier> And the good point is to configure all the testcases in the same way (flavor specs, ...) or even to remove useless configs.
08:29:33 <ollivier> From the time being it's rated -1 because I can't test it right now (my testing POD is broken)
08:30:02 <ollivier> but it allows sharing the whole idea
08:30:24 <depo> I will see if I can figure a way to test it
08:30:37 <depo> and let you know
08:30:39 <ollivier> tenantnetwork can be merged already (I will add unit test as soon as duplicated codes are removed)
08:30:43 <ollivier> yes thank you
08:30:54 <ollivier> I can build a dedicated container if it helps
08:31:05 <depo> yes, it would
08:31:17 <ollivier> thank you
08:31:30 <depo> my pleasure
08:31:52 <ollivier> #topic tempest
08:32:10 <ollivier> I think we do run tempest test in parallel as for refstack
08:32:50 <ollivier> and it seems that tempest_full has to be updated (it has to conform with Neutron's tempest full)
08:33:18 <ollivier> That's minor compared to remove the duplicated code.
08:33:22 <depo> we have a JIRA for this?
08:34:02 <ollivier> No I can open one. The JIRA tickets takes much more time that the gerrit patchset in the first case ;)
08:34:38 <depo> :))
08:34:48 <depo> make sense
08:34:52 <ollivier> I know that David loves JIRA but in opensource gerrit should be more important and JIRA mainly use by enduser
08:36:07 <depo> Companies in generall love JIRAs
08:36:28 <ollivier> Yes. Do you have another topic ? We would like to push your previous change on multijob.
08:36:47 <ollivier> I'm also considering implementing the task manager in Python if releng can't be updated.
08:36:51 <depo> regarding the topic I mentioned last week
08:36:59 <ollivier> logs?
08:37:11 <depo> logs, I abord, it would complicate things too much
08:37:25 <ollivier> yes. I saw it.
08:38:03 <depo> I mean I aboard the idea of splitting logs
08:38:30 <depo> multijobs would still be an improvement, just not sure how to implement it
08:39:11 <depo> maybe if we remove fetures as single job template would help ( having less job templates )
08:39:21 <ollivier> Yes. I will send a new email to Travor, Fatih and Aric.
08:39:36 <ollivier> The issue is simply that Functest can't declare any multijob by design
08:40:21 <ollivier> regarding logs, we could propose a new input to mute all logs out of Functest as before.
08:41:25 <ollivier> but we could mute them by default but it would be great to activate in jjbs to ease debugging
08:41:39 <depo> it would be an idea.. having all API calls and responses can be useful in debugging but it is also noisy
08:41:49 <depo> yes
08:42:16 <depo> have like DEBUGING=false ( default ) variable
08:42:38 <ollivier> Yes as the previous CI_DEBUG. We would have to enable it in Features in all cases.
08:43:07 <ollivier> Or we do declare all the namespaces know by Functest.
08:43:23 <ollivier> (sdnvpn, sfc, baro_tests ...
08:44:55 <ollivier> Let's go for that :)
08:45:17 <depo> I am confused with that
08:45:23 <depo> please detail
08:45:25 <depo> :)
08:45:53 <ollivier> previously we filtered all python namespaces starting by functest. (__name__)
08:46:51 <ollivier> We could propose to allow all logs from functest, and OPNFV features by default then to list all the python namespaces.
08:47:27 <ollivier> https://git.opnfv.org/functest/tree/functest/ci/logging.ini#n22
08:47:48 <ollivier> or 2 allows everything as the current behavior
08:47:52 <ollivier> to allow
08:48:09 <ollivier> We could write logging.ini on the fly to get the right behavior
08:49:00 <ollivier> I would say that we should print everything in gates and filter by default for endusers.
08:49:14 <ollivier> installers could be free to mute them if they want.
08:50:37 <ollivier> still unclear?
08:50:47 <depo> no, just processing
08:51:00 <ollivier> I will send you 2 logging.ini as examples
08:51:23 <depo> ok
08:51:34 <depo> I will take care of this
08:51:40 <ollivier> thank you.
08:52:08 <ollivier> anything else?
08:52:30 <depo> just processing on logging everything in gates, I found it noisy because it had everything there, but is ok
08:52:33 <depo> not, not from me
08:53:35 <ollivier> We could mute them by default and allow to print everything on purpose. We have to discuss with Feature teams as well.
08:54:05 <ollivier> Then we can close the meeting. Thank you Delia.
08:54:06 <depo> sure
08:54:11 <ollivier> #endmeeting