07:57:42 <ollivier> #startmeeting Functest weekly meeting 02 Oct. 2018
07:57:42 <collabot> Meeting started Tue Oct  2 07:57:42 2018 UTC.  The chair is ollivier. Information about MeetBot at http://wiki.debian.org/MeetBot.
07:57:42 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic.
07:57:42 <collabot> The meeting name has been set to 'functest_weekly_meeting_02_oct__2018'
07:57:53 <ollivier> #info Cédric Ollivier
07:58:05 <juhak> #info Juha Kosonen
07:58:36 <viktor_t> #info Viktor Tikkanen
07:58:40 <delia_> #info Delia Popescu
07:58:49 <ollivier> #topic Action point follow-up
07:59:02 <ollivier> #info Rally OpenStack CI test cases have been merged
07:59:12 <ollivier> #link https://gerrit.opnfv.org/gerrit/#/c/61895/
07:59:28 <ollivier> juhak: Thank you for this new testcase.
08:00:12 <ollivier> It works well in gates.
08:00:18 <ollivier> #info juju has been updated to 2.3.9
08:00:36 <ollivier> it bypassed the subnet discovery issues as expected
08:00:55 <ollivier> #info default security groups are removed now
08:01:18 <ollivier> #link https://gerrit.opnfv.org/gerrit/#/c/62867/
08:01:44 <ollivier> #info tempest.conf has been fixed
08:01:53 <ollivier> And tempest_full is passing in Orange Openlab now
08:02:32 <ollivier> Apex will be good as well as soon as storage protocol is correctly set
08:02:35 <ollivier> #link https://artifacts.opnfv.org/logs/functest/lf-pod1/2018-10-01_06-30-03/tempest_full/tempest-report.html
08:02:53 <ollivier> viktor_t: thank you for the support on this.
08:04:40 <ollivier> #info ONS
08:04:49 <ollivier> #link https://events.linuxfoundation.org/wp-content/uploads/2017/12/LFN-Xcommunities-Testing-Morgan-Richomme-Cedric-Ollivier-Orange.pdf
08:04:57 <ollivier> #link https://events.linuxfoundation.org/wp-content/uploads/2017/12/LFN-Test-Frameworks-Certification-What-a-Telco-Really-Needs-Morgan-Richomme-Orange.pdf
08:05:31 <ollivier> I sent you few notes yesterday. Be free to ask for details if needed.
08:07:32 <ollivier> #topic Functest
08:07:53 <ollivier> #info all installer gatings are quite bad, but our testcases seem more than fine vs Orange Openlab (Queens)
08:08:04 <ollivier> (all testcases pass, tempest_full included)
08:08:17 <ollivier> byw we are closed to fully run Functest in parallel (only one rally_full iteration failed)
08:09:10 <juhak> is that about cinder volumes or something else?
08:09:54 <ollivier> yes. only one failure when running rally_full! I will work on it this week
08:10:36 <ollivier> I may check iptables locks in l3-agent as well. Here I can't verify two Functest releases in 1 run.
08:11:17 <juhak> ok, just let me know if I can check something
08:11:46 <ollivier> juhak: Thank you.
08:12:16 <ollivier> delia_: Fuel (arm64) deployments seem falsy. https://build.opnfv.org/ci/view/functest/job/functest-fuel-armband-baremetal-daily-master/lastFailedBuild/console
08:12:29 <delia_> yes, we're working on it
08:12:45 <ollivier> delia_: It's ok on Functest side?
08:13:28 <delia_> ollivier: it should be ok, I will though make sure
08:13:46 <ollivier> delia_: thank you.
08:14:06 <ollivier> delia_: byw Fuelm is not running the right container
08:14:07 <delia_> ollivier: it looks now that api_check tried internal endpoint
08:14:17 <delia_> ollivier: the same does cloudify_ims
08:14:29 <ollivier> delia_: cloudify_ims tries public endpoint
08:14:47 <ollivier> delia_: export OS_INTERFACE=internal
08:15:00 <delia_> ollivier: yes, that is why
08:15:24 <ollivier> delia_: you may remove the former env var (export OS_ENDPOINT_TYPE="internal")
08:16:11 <delia_> ollivier: I will speak with pma. I tried before..
08:16:21 <delia_> ollivier: https://jira.opnfv.org/browse/FUEL-376
08:16:21 <ollivier> delia_: In case of cloudify_ims, it may create the orchestrator vm via internal endpoints as configured. Then cloudify uses the public endpoints.
08:16:38 <delia_> ollivier: I updated it with the result , it is the endpoint
08:16:57 <delia_> ollivier; I don't have now an env for investigation, but I will
08:17:51 <ollivier> delia_: no pb. It seems ok for the time being. I don't know which env vars are read by snaps (OS_INTERFACE or OS_ENDPOINT_TYPE)
08:18:17 <delia_> ollivier: the orchestrator looks like created, but it fails to create security_groups becouse it ca't reach the url. Simple "cloudify" testcase, which only creates the orchestrator passes.
08:18:38 <delia_> ollivier: I need though to test on baremetal, I only had virtual env
08:19:35 <ollivier> https://git.opnfv.org/functest/tree/functest/opnfv_tests/vnf/ims/cloudify_ims.py#n117
08:20:06 <ollivier> Yes cloudify doesn't use the public endpoints.
08:20:51 <ollivier> cloudify only prepares the orchestrator but doesn't deploy any vnf.
08:21:52 <ollivier> delia_: Fuel should run gambia containers instead of latest.
08:21:53 <delia_> ollivier: I will try to remove the  OS_INTERFACE=internal
08:22:15 <delia_> but I guess we shou;d also make sure we use public endpoints in all tests
08:22:39 <ollivier> delia_: Normally internal endpoints should be reachable from jumphost.
08:23:08 <delia_> ollivier: they aren't in fuel (for what I know)
08:24:13 <ollivier> delia_: in such cases, you have to switch to public but it may raise side effects when running admin identity tests in tempest_full
08:25:07 <ollivier> #info do we remove the patch logics ?
08:25:18 <ollivier> #link https://gerrit.opnfv.org/gerrit/#/c/61913/
08:25:30 <ollivier> If we do so, endusers may have to override the default config_functest.yaml
08:25:39 <ollivier> but it would ease maintaining Functest (most of that scenarios are out Functest gates)
08:25:55 <ollivier> my concern is that extra specs or image properties could differ between testcases (or even orchestrator and vnfs)?
08:27:01 <delia_> my ideea started with  Fuel wanting to go with scsi
08:27:13 <delia_> so either we have a patch or env var
08:27:29 <delia_> if we use env var we do need to make sure we can keep the ame facilities
08:27:32 <delia_> we have
08:27:39 <delia_> just in another way
08:28:14 <delia_> ollivier: how do you see it/
08:28:17 <delia_> ?
08:29:05 <ollivier> delia_: interesting :). The esiest way is simply to override the default config yaml file via a docker volume.
08:30:58 <delia_> ollivier: ok, I see no problem there
08:32:03 <ollivier> Why not suggesting new env vars. Then it will be applied to all testcases compared to the classic config which allows specifying per testcase.
08:32:48 <delia_> that was my sec idea..
08:32:51 <ollivier> We can first fix the issue by overriding the default config file and continue on thinking about removing patches or suggesting new env vars. No rush?
08:33:05 <delia_> ok
08:33:56 <ollivier> We could implement more complex logic. First read env vars, then check functest config else take class attributes as default...
08:34:24 <delia_> sounds good
08:35:00 <ollivier> I think it would only work for extra specs and image properties
08:35:29 <ollivier> (vio asks for other filenames and disk formats)
08:36:35 <ollivier> We could also remove the patch logics in such a case and force enduser to override the default config.
08:37:16 <delia_> this seems like an idea to digest, work on
08:37:37 <delia_> so maybe I can go to the initial temporary sol as a patch for scsi?
08:37:56 <delia_> with img_extra_properties that now reside on arm patch
08:38:03 <delia_> like split arm patch temporary
08:38:05 <delia_> ?
08:40:02 <ollivier> to override the default config yaml file via a docker volume?
08:40:29 <delia_> no
08:41:14 <ollivier> to create the new env vars ?
08:41:25 <delia_> no, my initial patch version
08:41:36 <delia_> I split arm patch in 2 peaces
08:41:55 <delia_> all img_extra properties were sent to the sec patch
08:42:02 <delia_> all facilities we have are kept
08:42:04 <ollivier> The first patchset is fuel dependent.
08:42:17 <delia_> and have VOLUME_DEVICE_TYPE sent by fuel
08:42:22 <delia_> yes
08:42:30 <delia_> or scsi dependent
08:42:38 <delia_> you may put it this way
08:43:26 <ollivier> I think the logic should be in releng if we consider the first patchset.
08:43:33 <delia_> basically we would have VOLUME_DEV_TYPE and arm patch split in scsi patch and arm patch
08:43:44 <delia_> in releng?
08:45:01 <delia_> I guess we have other topics too, we can take this one by mail, have more details there :)
08:45:32 <ollivier> Yes it's OPNFV specific. I would consider we override the default values instead in jjb/functest/functest-alpine.sh.
08:46:33 <ollivier> yes.
08:46:49 <ollivier> #info Intel NFV CI test-cases are under-progress
08:47:08 <ollivier> libvirt (qemu) credentials should be defined by env vars instead of testcases.yaml or class attributes
08:47:15 <ollivier> patches vs openstack/intel-nfv-ci-tests should be rebased
08:47:28 <ollivier> few dependencies (in OpenStack) are rated -1
08:47:55 <ollivier> #link https://gerrit.opnfv.org/gerrit/61151
08:48:35 <ollivier> #info LFN team asks for modifying our tox.ini + doc trees
08:48:50 <ollivier> #link https://gerrit.opnfv.org/gerrit/#/c/62523/
08:49:03 <ollivier> #link  https://gerrit.opnfv.org/gerrit/#/c/62347/
08:49:29 <ollivier> quite late regarding milestones and disruptive. Let's see how to manage that.
08:49:49 <ollivier> Nothing else on Functest part?
08:50:06 <ollivier> Anything else?
08:50:36 <viktor_t> No
08:50:38 <delia_> no
08:50:47 <juhak> nope
08:50:58 <ollivier> #topic Functest Kubernetes
08:51:07 <ollivier> #info 1 patch is still pending (it requires deep updates in Xtesting)
08:51:16 <ollivier> nothing new :)
08:51:22 <ollivier> #topic AoB:
08:51:45 <ollivier> We should get LFN resources to introduce fonctional gates.
08:52:02 <ollivier> I'm still waiting for vpn credentials
08:52:32 <ollivier> That should allow keeping Functest always stable by design
08:53:30 <ollivier> Anything else?
08:55:50 <ollivier> so let's close the meeting. Thank you. Have a great week!
08:55:53 <ollivier> #endmeeting