08:00:29 #startmeeting Functest weekly meeting 9 Jan. 2018 08:00:29 Meeting started Tue Jan 9 08:00:29 2018 UTC. The chair is ollivier. Information about MeetBot at http://wiki.debian.org/MeetBot. 08:00:29 Useful Commands: #action #agreed #help #info #idea #link #topic. 08:00:29 The meeting name has been set to 'functest_weekly_meeting_9_jan__2018' 08:01:01 #info Cédric 08:01:14 My best wishes for the year! 08:01:22 #info Juha Kosonen 08:01:38 #info Jose Lausuch 08:01:39 #info Linda Wang 08:01:56 #info Cristina Pauna 08:01:56 ollivier: Happy New Year! 08:02:10 LindaWang: thank you 08:02:11 happy new year to all 08:02:57 lots of actions done during the last 3 weeks. I will try to list them quickly. Please let me know if we must detail some of them. 08:03:03 #info  Action point follow-up 08:03:23 #topic Action point follow-up 08:03:34 #info the 3 new repositories were created (functest-requirements, functest-xtesting and functest-kubernetes) 08:03:43 #info the docker container garbage collector was published  (21/11 AP1) in both master and stable/euphrates 08:03:55 #info release notes were updated to simply list manifests (19/12 AP1) then no need to publish additional tags (19/12 AP2) 08:04:07 #info getting auth token is tested when checking deployment (19/12 AP6) 08:04:19 #info ceilometer tests failing in rally sanity were removed (19/12 AP7) 08:04:27 thank you all for that! 08:05:23 Regarding getting auth token, it may be incomplete to only check OS_AUTH_URL. Fuel is buggy (wrong public identity endpoint) and we couldn't detect it. 08:06:08 #topic Euphrates 08:06:29 #info orchestra_* testcases were now enabled for Daisy only 08:06:40 #info "inject all envs in openrc into CONST" was backported 08:06:48 #info minor errors were fixed in documentation 08:06:49 #info Delia Popescu 08:07:33 I think Euphrates is done. We will only fix new bugs. 08:07:39 Agree? 08:07:52 agree 08:08:23 yes 08:08:45 ollivier: about getting auth token issue, we could talk about it later. 08:09:23 i did not notice the bug from fuel 08:09:33 yes. From the time being, it's fine. Fuel bug could have been detected before running vnf testcases. 08:10:10 A wrong v2 url is registered in catalog. Only vnf gets the public endpoint before creating resources. 08:11:17 #topic functest 08:11:40 I think we don't list the next topic last meeting. 08:12:21 We need to check testcases if our testcases are able to manage unversioned endpoints (key change in Pike). 08:13:06 now v2 or v3 should not be listed in url. Then it could break several middleware (see juju) 08:13:32 #action all to check vnfs if they manage unversioned endpoints 08:14:34 maybe it only concerns juju but I want to share this possible side effects. 08:15:12 #info juju_epc has been successful vs Daisy but they still remain errors for the other installers 08:15:35 ok, i have not seen other tests with the same issue 08:17:09 Yes. I think *ims* testcases are fine. 08:18:10 #info all Functest testcases have been successful vs Daisy 08:18:48 Great even if it's the simplest case (http endpoint, management network reachability) 08:19:27 It seems that we are fine for the next release :) 08:20:30 MS4 you mean? 08:20:37 Jan. 15th? 08:21:01 My feeling is MS4 is ensured by unit tests. Then we must be successful 08:21:46 The issue is jsut about xtesting as we will change some python namespaces. Normally we shouldn't change the framework. 08:22:17 the milestone is about framework ready 08:22:20 They are trivial modification in 2 or 3 projects. That's not a big deal 08:22:23 minor changes can be done, specially bug fixes 08:22:30 exactly 08:22:38 I reported to David. That's MS4 is ok. 08:22:43 I saw that 08:23:24 Our rules are better simply because we check MS4 automatically (and before merging) 08:23:46 David also mentioned the feature patch gating job in the last release meeting, need to check that with Tim from Apex, 08:24:18 Tim would provide a snapshot to enable deployment openstack in a few minutes 08:24:40 I am not sure about this progress 08:25:24 Yes we discussed in case of Functest as well. In our case, it's more about checking MS3 automatically (running healthcheck) by functional tests 08:25:34 That's the same topic as XCI 08:26:04 (during release meeting right after the last Functest meeting) 08:27:11 My point is we require Functional gating and Tim said it's could be possible by snapshoting APEX and by hacking the current jenkins jobs. 08:28:04 which tests could be used for functional gating? 08:28:13 only healthcheck is enough? 08:28:29 Regarding Feature, I think the issue is more about gating several features every 3 days (SDNVPN) when APEX allows that. 08:29:03 I would say as much as possible. I would have added tempest as well. 08:29:06 the functional gating you mean is for functest patch gating? 08:29:37 the gating tests should be simple and quick 08:29:45 yes. To deploy a referent platform and to check healthcheck before merging the patch 08:29:53 healthcheck 08:29:57 I wouldn't add tempest there 08:30:41 yes, tempest tests would take a longer time 08:30:59 If the platform is considered as stable... Tempest could be executed. 08:32:14 OpenStack runs tempest before merging. 08:32:33 only on nosd-nofeature-ha scenario? 08:33:27 We must be confident about SUT. Then I would say os-nosdn-nofeature-ha 08:33:47 yes, me too 08:34:07 If SUT is buggy, we can't run implement functional gating simply because we will lose time to debug and to override jenkins vote. 08:34:08 we can always extend it later, but let's start with the base scenario 08:34:27 ok 08:34:57 But I fully agree. At least MS3 must be covered. Let's see for tempest later (for the time being only Functest is ready for thsi gating) 08:37:26 Regarding the global planning, Fraser status is more that fine as os-nosdn-nofeature is already working in the simplest scenario. We must check the https certificates and the identify endpoint management. 08:38:06 that's very positive :) 08:38:28 Yes, great work! 08:38:38 byw, the issues during basic heat operations were fixed in XCI 08:39:05 this issue is only related to vnf tests? 08:39:51 the policy.json generated by XCI was wrong and Functest detected that. 08:40:56 can you share the link for the fix? 08:41:12 you mean when running tempest and rally/ 08:41:13 ? 08:41:48 before. Any basic heat operation fails (connection_check) 08:42:08 juhak: yes. I am searching the OpenSTack review. 08:42:41 i think policy.json defines the role for the user to take some actions 08:43:36 #link https://review.openstack.org/#/c/529147/ 08:44:20 ollivier: thanks! 08:44:36 I think everyone is added as CC to Morgan's email. If not I will complete the list. 08:44:38 but this patch has been abandoned? 08:44:46 Yes. 08:45:21 It's the first proposal submitted by Fatih. They simply switch the proposal but I think the topic is closed. 08:45:43 If I remember well it also remains one issue regarding the external network. 08:45:58 It has been fixed in XCI? 08:46:31 Yes. From the time being Functest is used to qualify XCI. But then if the platform is stable we could apply the opposite :) 08:46:49 That's the purpose of snapshots. 08:47:27 #topic functest-kubernetes 08:47:35 ollivier: could you forward Morgan's email to me? 08:47:45 ollivier: thanks 08:48:14 Sure. I think Morgan added everyone. But I will double check and complete it if required. 08:48:34 #info Functest Kubernetes Dockerfile was merged 08:49:24 #link https://gerrit.opnfv.org/gerrit/#/c/50063/ 08:50:20 i have tested it locally, it works well 08:50:24 Thank you, Konrad. 08:50:51 The test or the container builds ? 08:51:07 #info ollivier/functest-kubernetes is built automatically per change 08:51:20 i have pulled ollivier/functest-kubetetes 08:52:00 and i have created a testcases.yaml to run conformance tests inheriting from feature 08:52:06 it works 08:52:11 Great! 08:52:28 thats good news 08:52:40 But i would prefer to make it inherit from testcase.TestCase, 08:52:59 Thank you. It confirms the go version and we will have a coverage for K8s as well. 08:53:00 an email has been sent to you 08:53:26 ok. I will check it. Then we will have to implement the releng part 08:53:39 are we inheriting from bashfeature? 08:54:01 the testcases.yaml has not be published 08:54:19 first to build the containers (I think we can simply add functest-kubernetes in the classical releng build) 08:54:28 second to run it. 08:54:42 ollivier: ok 08:54:48 I can take care of adding that to the classical build 08:54:55 thank you 08:55:12 even though the testcases.yaml has not be published, i thought you wanted to make it inherit form feature? 08:55:17 #action Jose to take care of adding functest-kubernetes to the classical build 08:55:19 am i wrong? 08:55:32 Now this proposal could be fine as well. 08:55:36 jose_lausuch: thanks 08:55:42 feature or bashfeature? 08:55:50 bashFeature 08:55:53 It allows to manage deeper data sucvh as detail when publishing in DB. 08:56:38 ok 08:57:07 Jose could we first finish xtesting (next topic) before releng. There are many releng tasks but I think we should first focus on framework (MS4 related) Agree? 08:57:17 #topic xtesting 08:57:23 ok 08:57:30 #info lots of changes have been published to break dependencies to OpenStack or to OPNFV utils 08:57:32 addding the classical build is only 5 min I think 08:57:39 I need to finish my patch about push to db 08:57:44 sorry for being late with it 08:57:48 no pb. 08:58:36 It's the last one. I can help you this afternoon if needed. 08:58:51 ok, thank you 08:59:00 it could take until tomorrow as I have a workshop this afternoon 08:59:03 but will do my best 09:00:36 I think your proposal in comment is fine and you have done the biggest part. 09:00:50 I will merge all in 1 function 09:00:54 ok. 09:01:01 instead of 1 static and 1 class method 09:01:43 Yes. Your proposal is fine. I'm fine to wait tomorrow. Then I will publish my local changes in functest-xtesting (19/12 AP5) 09:01:46 let's discuss it later if needed, thanks :) 09:02:02 #topic AoB 09:03:03 I also noted actions on releng that I will share later by email. 09:03:46 We are still working on fixing the issues when compiling ruby modules on ARM (mismatch in dependency versions). 09:04:04 (vnf + testapi) 09:04:49 I will be partially off the next 3 days as I am giving an ODL training session. I will do my best to be connected on irc and emails. 09:05:04 got it. 09:05:37 Thank you very much. Great work. 09:05:38 ok, thanks 09:05:46 good meeting 09:05:58 #endmeeting