08:02:40 #startmeeting Functest weekly meeting 6 Feb. 2018 08:02:40 Meeting started Tue Feb 6 08:02:40 2018 UTC. The chair is ollivier. Information about MeetBot at http://wiki.debian.org/MeetBot. 08:02:40 Useful Commands: #action #agreed #help #info #idea #link #topic. 08:02:40 The meeting name has been set to 'functest_weekly_meeting_6_feb__2018' 08:03:04 #topic Role call 08:03:09 #info Cédric Ollivier 08:03:12 #info Linda Wang 08:03:16 #info Juha Kosonen 08:03:37 #info xudan 08:03:55 #info Jose Lausuch 08:04:42 Before starting, the gate is currently broken (api_check). see https://gerrit.opnfv.org/gerrit/#/c/51695/ 08:05:00 i have tested locally, it works 08:05:07 #info Viktor Tikkanen 08:05:30 even though it also worked before updating on my local deployment 08:05:45 ok. I'm suprised that it's not verified +1. I was connected quite late yesterday and I missed the last discussions between Tim and Steven 08:06:15 it was quite late then 08:06:33 Also we are also facing with lots of issues when building our containers (network / Arm issues) 08:06:54 i have retriggered the docker build jobs just now 08:07:06 all containers were built successfully 08:07:20 ok. Let's check the next runs. 08:07:27 sure 08:07:28 hi all. Let's begin. 08:07:45 #topic Action point follow-up: 08:08:04 #info all pylint errors/warning were fixed in rally and refstack 08:08:12 our code has been rated at 8.91/10 which is great 08:09:29 We should update vnf, functest and openstack utils. https://build.opnfv.org/ci/job/functest-verify-unit-tests-and-docs-master/705/console 08:10:28 #info Delia Popescu 08:10:40 I think we should double check wich functions are useful in functest utils (see thread with SFC). And some of them are duplicated in vnf. 08:10:50 are the pylint of openstack_utils still needed to fix? 08:11:20 I don't think so. We should check wich functions are still used. 08:11:47 for functest_utils, I think several functions should be defined in run_tests instead (yaml parsing) 08:12:12 then we could remove functest_utils? 08:12:18 Not sure 08:12:59 you mean if the functions are only called by functest-xtesting, we could move it to run_tests? 08:13:00 I would like to check which functions are used internally or defined for other projects. 08:13:48 for sfc, we have to check the releng-xci and sfc repos? 08:13:49 I think functest_utils defines methods wich are relevant for ci. Then we could consider moving them in the right module. As we did for push_to_db 08:14:24 I think functest_utils makes sense for internal methods shared between testcases. 08:15:22 My proposal would have been to double check which methods should be moved outside, which are duplicated (see vnfs) and then to clean the remaining pylint errors. 08:15:25 for feature projects involved in F release, only bpgvpn and sfc are using such inernal functions from functest? 08:15:47 jose_lausuch: I think so. Jose could you confirm ? 08:16:57 yes 08:16:59 that´s right 08:17:16 I can do a grep in some repositories to see 08:17:22 About openstack_utils, we should check where it's still used (ODL) and then remove it. 08:17:56 def source_openrc is still used by run_tests 08:18:07 #action jose_lausuch to check where our openstack utils are still used 08:18:15 ok 08:18:19 I will do a summary 08:18:21 It should be moved to run_tests directly, shouldn't it? 08:18:31 and some other related functions 08:19:14 maybe 08:19:20 Thx. That's the best time to move functions as features are being updated. 08:19:46 Could I add an action point? 08:20:04 move source_openrc? 08:20:09 yes 08:20:11 sure, please 08:20:36 #action LindaWang move source_openrc to run_tests if possible (+ possible yaml parsing helpers) 08:21:50 I action myself to update ODL and epc regarding openstack utils. 08:22:15 Be free to update vnfs regarding openstack utils. 08:22:41 #action ollivier update ODL and epc towards snaps 08:23:00 #action world be free to update vnfs regarding openstack utils 08:23:28 could we disable orchestra_*ims? 08:23:38 Be free to fix any pylint erros elsewhere. juhak is fixing tempest testcases. 08:23:53 I think so. This is part of a next topic :) 08:24:36 Globally our code is good (~9/10). That's great. 08:25:04 #info docstring modules were added in snaps-related testcases 08:25:14 #topic Functest 08:26:12 #info please see SNAPS-263 which is blocking all installers 08:26:25 #info we renamed openstack.creds to a common credential file for OS and K8S 08:26:32 #link https://gerrit.opnfv.org/gerrit/51283 08:26:43 All xci and releng jobs have been updated. 08:27:00 #info a new wiki page about running latest containers has been initiated 08:27:10 #link https://wiki.opnfv.org/pages/viewpage.action?pageId=13211751 08:27:29 #info all results printed in configguide were updated (both stable/euphrates and master) 08:27:53 #info only refstack_defcore(under investigation), cloudify_ims and vyos_vrouter are failed (as for all installers) vs Orange POD 08:28:38 have you tested tempest_full_parallel also? 08:28:42 Yes 08:28:57 how many sub tests passed? 08:29:03 It works because criteria is 80%. 08:29:56 but rally_full is easy to achieve 100%? 08:30:15 should be 08:30:22 1506 08:30:50 1506 over 1600+? 08:31:27 Success 1133. Skipped 353 Failures 20 08:31:43 ok, got it. 08:33:19 About rally, it seems fine to increase criteria to 100. We already disabled the wrong testcase (race conditions) and Juha is working on the Cinder testcases. 08:34:15 add new test cases about Cinder, juhak ? 08:34:35 We could apply the same for tempest (to reach 100% by disabling any possible wrong tests). But we could check that later. 08:36:03 but we do not the results of temepst_full on other deplotments. 08:36:05 LindaWang: checking failed cinder case, maybe the scenario needs to be modified 08:36:47 juhak: oh, i remember, live migration on Apex. 08:37:44 LindaWang: that a different case 08:37:44 not really. It's about neutron issues in this case. I can share with you my logs. We also have to work on live migration for "smoke" rally. 08:38:22 live migration works well here. 08:38:40 (I should check if it's enabled :)) 08:38:56 ollivier: good point 08:39:36 #link https://gerrit.opnfv.org/gerrit/51679 08:40:58 We should still find the best solution about "Fix the config of fixed_network_name in tempest". 08:42:13 #topic vnf 08:42:27 #info juju_epc is now working on Orange POD (we have to modify the dns server from the time being) 08:42:46 #link https://gerrit.opnfv.org/gerrit/51419 08:43:01 We have to wait the next CI runs but the results should be ok. 08:43:13 #info all emails have been sent to the first VNF authors 08:43:25 #info juju_epc: they can start working from this week 08:43:38 #info: vyos_vrouter: they will start working on their JIRA tickets after 26/02 08:44:23 I will try to fix the issue about authentification before. Be free to have a look. 08:44:45 #info orchestra*: no answer (or I missed them?) 08:45:04 i did not see the answer, either 08:45:06 I consider that we could disable the 2 testcases as they waste CI runs for any good result 08:45:20 my patch is already prepared and I just need to submit 08:45:28 agree 08:46:06 even though it is only on one installer 08:46:14 Cedric Ollivier proposed functest: Disable orchestra_* testcases https://gerrit.opnfv.org/gerrit/51749 08:47:02 I hope they will update the code. We could also move the related python modules in a new orchestra repository. 08:48:09 I will forward any news. 08:48:31 #topic functest-kubernetes 08:48:56 # info the compass job has been published https://gerrit.opnfv.org/gerrit/#/c/51093/ 08:49:00 Is it running? 08:49:11 compass has some issue deploying k8s 08:49:37 the admin.conf is missing, so the step of fetching admin.conf failed 08:50:07 i also sent email to Joid team to ask info about k8s 08:50:50 At least we are ready. 08:51:03 #info a second testcase is under review https://gerrit.opnfv.org/gerrit/51635 08:52:22 #topic AoB 08:52:36 #info http://functest.readthedocs.io/ 08:53:12 It's better than the classical shared via testresults. 08:54:36 #info Apex announced (last TSC meeting) that they will propose a master version. 08:54:39 looks great 08:55:42 They mainly follow the XCI model. They could raise side effects here as functest is used in their verify jobs. 08:56:48 when will the master version be proposed? 08:56:54 It could indirectly help us if several installers are ready from Day-0. 08:57:16 It would break the circular dependencies between Functest and installers. 08:57:21 but functest master is for Pike 08:58:25 Yes. There is no rush for that topic. We could consider Functest following master as well but nothing is clear (and we depend on snaps) 08:59:35 I just want to inform you. I will discuss on that this afternoon. 09:00:08 We should continue working on testcases as we did + prepare xtesting. The key question is about release management if several installers follow master. 09:00:46 but that requires functional gating then we could indirectly benefit from that. 09:00:56 We are on time. 09:01:21 Thank you all. Great week! 09:01:26 #endmeeting