13:00:05 <anac1> #startmeeting Test&Performance Ops meeting 13:00:06 <collabot> Meeting started Fri Oct 9 13:00:05 2015 UTC. The chair is anac1. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:06 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic. 13:00:06 <collabot> The meeting name has been set to 'test_performance_ops_meeting' 13:00:14 <anac1> #info Ana Cunha 13:00:29 <jose_lausuch> morgan_orange: ping? 13:00:32 <jose_lausuch> #info Jose Lausuch 13:00:41 <morgan_orange> #info Morgan richomme 13:00:58 <morgan_orange> jose_lausuch: pong 13:01:01 <jose_lausuch> I have a question not related to the agenda 13:01:02 <jose_lausuch> but 13:01:11 <anac1> yes 13:01:11 <jose_lausuch> if we integrate "push results to db" in each test 13:01:18 <jose_lausuch> if there is a non-federated lab running functest 13:01:21 <jose_lausuch> what will happen? 13:01:28 <jose_lausuch> are they allowed to push also to that db? 13:01:35 <jose_lausuch> or only for the pharos labs connected to CI 13:01:49 <morgan_orange> good question 13:01:57 <jose_lausuch> because they might not have access 13:02:07 <jose_lausuch> (it came to my mind 5 min ago ) 13:02:08 <jose_lausuch> :) 13:02:15 <morgan_orange> I was about to write a mil related to the security aspects we discussed on wednesday 13:02:26 <anac1> #topic test result dB 13:02:32 <morgan_orange> for me anyone whishing to share its results should be able to do it 13:02:38 <jose_lausuch> yes, we have also to make sure that if there are not credentials in the current lab, dont push 13:02:41 <jose_lausuch> or something like that 13:02:45 <morgan_orange> yes 13:02:52 <jose_lausuch> if credentials 13:02:52 <jose_lausuch> push 13:02:52 <jose_lausuch> otherwise 13:02:52 <jose_lausuch> no push 13:03:04 <anac1> #info who can push data to opnfv result-dB? 13:03:10 <morgan_orange> I prefer the basic auth approach rather than IP rules 13:03:38 <anac1> #info non-members of pharos ? 13:03:46 <morgan_orange> and currently if we imagiend we had security rules or you run the test from a non connected env 13:04:08 <morgan_orange> it will work anyway you will just get a tiemout on the push data request 13:04:33 <jose_lausuch> ok 13:04:46 <jose_lausuch> but avoiding that timeout would be nice :) 13:05:03 <anac1> conclusion? 13:05:18 <morgan_orange> coudl be a parameter 13:05:34 <morgan_orange> we could create an option in the config file 13:06:27 <morgan_orange> #info to be discussed with the testing community => would a basic authent be enough for pushing test resuilts in test result DB 13:06:35 <jose_lausuch> ok 13:06:35 <anac1> thanks 13:06:41 <jose_lausuch> that also sounds good 13:06:46 <morgan_orange> #info or shall we restrict the access to DB to Pharos lab 13:07:01 <jose_lausuch> its not critical, but its a design thing that we will face sooner or later 13:07:05 <jose_lausuch> at least, after R2 13:07:33 <morgan_orange> yes 13:07:41 <anac1> #topic test dashboard 13:07:56 <morgan_orange> #info meeting with LF last Wednesday 13:07:59 <anac1> i couldn't make to the meeting, any news to share? 13:08:25 <morgan_orange> #info I think that both parts now have the same understanding... 13:08:36 <morgan_orange> #info on one side a Mongo DB with a rest API to access 13:08:55 <morgan_orange> #info on the other side LF web team will develop PHP scripts to transform data into graphs 13:09:21 <anac1> #info what is next step? 13:09:21 <morgan_orange> #info LF will expect access to the DB and description of graph to be produced 13:09:59 <jose_lausuch> yes 13:10:06 <morgan_orange> #info LF web team needs to evaluate the resource needs (more code to do) 13:10:21 <morgan_orange> #info should be included in next Sprints 13:10:30 <morgan_orange> #info wait for feedback from LF web 13:10:41 <morgan_orange> meanwhile it is anyway possible for us to fill in the database 13:10:55 <morgan_orange> Rodrigue updated the APi yesterday 13:11:04 <morgan_orange> I must document it asap 13:11:14 <anac1> with yardstick request already? 13:11:20 <morgan_orange> #info standard steps for a test project 13:11:30 <morgan_orange> #info test project in CI 13:11:41 <morgan_orange> #info from test project use the API to push results 13:11:43 <jose_lausuch> but not the final db , right? 13:11:47 <jose_lausuch> the one that you setup 13:12:07 <morgan_orange> no it is a temp db but as the APi has been updated, I think I could ask aric to integrate the API 13:12:11 <morgan_orange> then we could use the target one 13:12:19 <jose_lausuch> where is the temp db located right now? 13:12:23 <morgan_orange> but we need to implement security mecahnism 13:12:36 <morgan_orange> the temp DB is in a VM in an Orange public Cloud 13:12:43 <jose_lausuch> ok 13:12:53 <jose_lausuch> thanks 13:13:05 <morgan_orange> we will be able to save the data and reinject them if needed 13:13:13 <morgan_orange> as the APi was not completed 13:13:26 <morgan_orange> the results were only associated with a pod, not with the installer and the version 13:13:30 <morgan_orange> which shall be possible now 13:13:34 <morgan_orange> I have to update the API... 13:13:39 <morgan_orange> and send the mail to aric 13:13:47 <morgan_orange> and document the API 13:13:52 <morgan_orange> I got a question on that 13:14:10 <morgan_orange> for the API, would a description on the wiki enough or shall we provide a sagger like API description 13:14:20 <morgan_orange> ? 13:14:30 <morgan_orange> anac1: yes you shall be able to push yardstick results 13:14:35 <morgan_orange> at least the simple ones 13:14:39 <anac1> thanks ! 13:14:51 <morgan_orange> I think you ask for a put method in orfer to modify test results 13:14:59 <morgan_orange> as far as I know it is not implemented yet 13:15:11 <morgan_orange> but you can declare testcases and push results, get results,... 13:15:31 <morgan_orange> I planned a demo during Maryam meeting next week 13:15:35 <jose_lausuch> morgan_orange: I think a description on the wiki is ok 13:15:37 <morgan_orange> we will do the exercice practically 13:15:51 <morgan_orange> jose_lausuch: if so that is something I can do quickly 13:15:53 <anac1> i think an API is preferable 13:16:07 <anac1> sorry... 13:16:35 <jose_lausuch> but for a first step 13:16:36 <morgan_orange> I will try to do that this aftenoon 13:16:44 <jose_lausuch> so that there is something 13:16:45 <morgan_orange> so people will be able to use 13:16:53 <morgan_orange> theoretically the existing page already gives the details 13:16:59 <anac1> yes 13:17:08 <morgan_orange> but I gave too many details on the json files and we do not really care 13:17:16 <morgan_orange> I will try to do something simpler 13:17:21 <jose_lausuch> ok 13:17:38 * fdegir thinks if db sits in lf infra, it might not be possible to open firewall to the whole world 13:17:52 <morgan_orange> we use a rest API 13:17:53 * fdegir thinks php sucks 13:17:58 <morgan_orange> we do not connect the DB directly 13:18:06 <morgan_orange> opening port 80... 13:18:40 <anac1> ok, anything else? 13:18:49 <morgan_orange> fdegir: I planned to discuss with aric on the security mehcnaism they want to put in place 13:18:58 <morgan_orange> #topic cross matrix 13:19:12 <morgan_orange> despite the 25 C milestone, we got new requests.. 13:19:19 <morgan_orange> OpenContrail, VPN 13:19:29 <morgan_orange> I think we can take opencontrail in Functest 13:19:43 <anac1> vpn is included in yardstick 13:19:48 <morgan_orange> BGP VPN? 13:20:08 <fdegir> ok 13:20:11 <anac1> yes 13:20:21 <anac1> project SDNVPN 13:20:48 <morgan_orange> ok as far as I understood they planned to support it through ODL, Neutron/Bagpipe, and Nuage? 13:20:58 <morgan_orange> my colleagues are working on the second scenario 13:21:12 <anac1> i informed, i think, last week ? sorry if not 13:21:18 <morgan_orange> but it makes sense to do it in Yardstick 13:21:32 <morgan_orange> ok 13:21:40 <jose_lausuch> your collegues can become yardstick contributors :) 13:21:46 <jose_lausuch> *colleagues 13:21:56 <morgan_orange> they are commiting code in Openstack on the backpipe plugin 13:22:02 <jose_lausuch> ok 13:22:12 <morgan_orange> but would surely become yrdstick contributors for teh test scenario 13:22:30 <anac1> makes sense to me 13:22:32 <morgan_orange> so Ana I will update https://wiki.opnfv.org/feature_test_project_matrix 13:22:48 <anac1> thanks 13:22:50 <morgan_orange> #info SDNVPN => yardstick 13:22:57 <morgan_orange> #info OpenContrail => Functest 13:23:10 <anac1> i got a request from ovs 13:23:31 <anac1> #info OVS => yardstick 13:24:11 <morgan_orange> ok it shall be now somehow relatively stable 13:24:17 <anac1> i hope 13:24:36 <morgan_orange> main issues will be for all the projects to see if dependencies may be taken into account by the installer 13:24:44 <morgan_orange> OVS 2.4, patch XYZ,.. 13:25:09 <anac1> agree, that's the feature project responsibility, right? 13:25:11 <morgan_orange> ok for this topic 13:25:19 <morgan_orange> #topic Summit 13:25:39 <morgan_orange> I wrote 3 lines last week end about the presentation on test dashboard 13:26:01 <morgan_orange> I created a css for reveal, I do not knwo if you are interested or if you prefer pptx 13:26:19 <morgan_orange> I submitted the code https://gerrit.opnfv.org/gerrit/#/c/2327/ 13:26:30 <morgan_orange> the first slides are there :) 13:26:45 <morgan_orange> the deadline is end of October 13:26:47 <jose_lausuch> thats nice 13:26:49 <anac1> ok 13:27:11 <morgan_orange> so anac1 I suggest we could have a first version of Test strategy for end of next week 13:27:31 <morgan_orange> anac1: you will be in Düsseldorf? 13:27:36 <anac1> no 13:27:42 <morgan_orange> ok neither do I 13:27:50 <morgan_orange> #info dealine End of october 13:27:52 <anac1> friday ok for a meeting ? 13:28:05 <anac1> i will book 13:28:11 <morgan_orange> we may use this slot and do a P2P meeting afterwards 13:28:24 <anac1> good ide, will check with trevor 13:28:28 <anac1> idea 13:28:29 <morgan_orange> we can reused the slide we made to explain the test stratgey and update them 13:28:49 <anac1> yes, we have the base already 13:29:08 <morgan_orange> jose_lausuch: we may wait LF feedback but as you say it is a logical progression.. 13:29:18 <morgan_orange> anac1: if we could have some test results from yardstik 13:29:28 <morgan_orange> it is to illustrate the heterogeneity of the results 13:30:18 <anac1> ok 13:30:18 <jose_lausuch> yep 13:30:22 <jose_lausuch> agree 13:30:23 <morgan_orange> I think we had a third slot on Pharos 13:30:37 <morgan_orange> #topic misc 13:30:57 <morgan_orange> I had a question, Yardstick, Qtip and Vsperf are all using pktgen, right? 13:31:09 <anac1> not sure about qtip 13:31:20 <anac1> yardstick and vsperf yes 13:31:25 <morgan_orange> yardstick and Qtip integrated the ability to use heat to perform orchestrated tests 13:31:39 <morgan_orange> anyone from qtip on the chan? 13:31:51 <morgan_orange> maybe too early fo California 13:32:42 <morgan_orange> I had a discussion with another Telco (not part of OPNFV) they would be interested to provide a testcase of a orchestrated VNF running pktgen. 13:32:55 <morgan_orange> My first reaction was that it was already covered by yardstick 13:33:15 <morgan_orange> and in the qtip minutes yesterday I had the feeling that a similar scenario was considered 13:33:31 <anac1> could be, depends on the scenario 13:33:39 <morgan_orange> yes.. 13:34:03 <morgan_orange> Ok I will try to get more information 13:34:06 <jose_lausuch> I have a comment about functest 13:34:20 <jose_lausuch> we talked a bit about that yesterday 13:34:44 <jose_lausuch> shall we consider having an offline functest and yardstick? 13:34:51 <jose_lausuch> for now functest needs internet 13:34:57 <jose_lausuch> otherwise, it fails 13:35:03 <morgan_orange> I saw your commit on urllib2 13:35:27 <morgan_orange> and your colleague said that Chris wanted to have offline version 13:35:32 <anac1> we start with internet, but i think long term we should also have offline 13:35:55 <jose_lausuch> morgan_orange: we have to evaluate how much effort that is needed for us 13:36:07 <morgan_orange> I would say that it is out of scope for R2 13:36:16 <jose_lausuch> its not so complicated though... 13:36:20 <morgan_orange> we would see for intermediate version 13:36:32 <jose_lausuch> because Chris wanted to run the demo offline 13:36:33 <morgan_orange> no it is just bigger because you must get all the packages you need 13:36:37 <jose_lausuch> and even fuel is giving problems :) 13:36:46 <morgan_orange> I am not sure to follow the intereste of an offline demo 13:37:05 <jose_lausuch> offline I mean without internet access 13:37:06 <jose_lausuch> but anyway 13:37:16 <jose_lausuch> we will discuss it some other day 13:37:30 <morgan_orange> or we may say that we will consider it as an improvement 13:37:36 <morgan_orange> if we can do it in R2 why not 13:37:44 <morgan_orange> but we will have lots of testcases 13:37:52 <anac1> good to have but not mandatory? 13:37:57 <morgan_orange> yes 13:38:00 <anac1> agree 13:38:00 <jose_lausuch> for us is now mandatory.. 13:38:15 <morgan_orange> ah so it was not a question :) 13:38:16 <jose_lausuch> we can include the cirros image inside the docker image 13:38:42 <jose_lausuch> anyway, will see 13:39:04 <morgan_orange> ok 13:40:07 <anac1> nothing else from me 13:40:08 <morgan_orange> #info question on the possibility to run tests without Internet connection (retrieve of disk image, access to git repo,...) 13:40:16 <morgan_orange> #info could be needed for demo? 13:40:34 <morgan_orange> #info consensus for a good to have but not mandatory... 13:40:39 <jose_lausuch> yes, for further demos maybe 13:40:48 <morgan_orange> #info even if it seems mandatory 13:40:48 <jose_lausuch> not for this one :D 13:41:03 <morgan_orange> try to sell OPNFV in the middle of Australia... 13:41:23 <morgan_orange> ok any other topic we want to share? 13:41:42 <anac1> no 13:42:18 <anac1> #endmeeting