16:00:01 #startmeeting Performance Team 16:00:02 Meeting started Tue May 31 16:00:01 2016 UTC and is due to finish in 60 minutes. The chair is DinaBelova. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:03 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:05 The meeting name has been set to 'performance_team' 16:00:22 hey folks! 16:00:29 are there people around? :) 16:03:51 interesting, it looks like nobody has joined yet :) 16:04:06 here 16:04:11 AugieMena hey! 16:04:21 let's see if people will join :) 16:04:23 gokrokve o/ 16:04:29 o/ 16:04:33 hi 16:04:56 gokrokve, AugieMena - I know lezbar won't we able to join, as well as ilyashakhat and rohanion 16:05:04 not sure about others 16:05:55 let's see if we'll have quorum today :) 16:06:14 three of us are here 16:06:21 although, gokrokve, AugieMena - I may go through the action items and agenda just for you :) 16:06:27 yeah, sure 16:06:29 thanks 16:06:32 ok, sure 16:06:37 ok, so let's get started 16:06:48 DinaBelova: Error: Can't start another meeting, one is in progress. Use #endmeeting first. 16:06:52 oh, sorry :) 16:06:57 #topic Action Items 16:07:03 #link http://eavesdrop.openstack.org/meetings/performance_team/2016/performance_team.2016-05-24-16.00.html 16:07:12 first of all action items 16:07:22 rook are you around? 16:07:58 in the meanwhile - lezbar reached me before the meeting 16:08:21 so he had a quick conversation about MidoNet vs OVN comparison test plan sharing inside his company 16:08:39 right now this task is on hold, so he cannot share anything with community so far 16:09:05 he promised to check when the task will be reopened and raise this question asap 16:09:19 as for my action item - about Ironic scale and performance testing effort progress with Mirantis Ironic team 16:09:38 i had a chat with our Ironic team lead 16:10:00 they have performed primary scalability testing several weeks ago 16:10:11 against 1000 "virtual" HW nodes 16:10:26 and have plans to continue the research soon 16:10:43 so we agreed to publish this research and its results to the perofrmance-docs 16:11:01 no ETAs so far due to the lack of real HW to work as a test env 16:11:11 i hope to have some ETAs next week 16:11:41 #action DinaBelova check what ETAs can be used for Ironic tests/results publishing 16:12:11 as for the left action item - it's on rook and frankly speaking I don't know about its status 16:12:22 so let's leave it for the next time 16:12:27 #action rook ilyashakhat go through the offloading model suggestion 16:12:44 that's all regarding our action items from the last time 16:13:14 gokrokve, AugieMena - if no comments, let's proceed 16:13:35 just a quick update on the Identity V3 Rally support issue 16:13:50 AugieMena - yes, sir? 16:14:11 there was a comment just added to the bug this morning: https://bugs.launchpad.net/rally/+bug/1441379 16:14:11 Launchpad bug 1441379 in Rally "Need Rally deployment create support on domain_id or domain_name in order to support keystone v3 test (which needs domain scoped token). The current project scoped token only works in service API but not keystone v3 API (DTUG domain, user, group...)." [Medium,Fix committed] 16:14:29 #link https://bugs.launchpad.net/rally/+bug/1441379 16:14:43 oh, it looks like we have a volunteer :) 16:14:46 very cool :) 16:14:58 hey DinaBelova 16:14:59 I've asked Kavitha (on our team) to add info to the bug in response to Michael's question on the work she's done so far, etc 16:15:10 AugieMena ack, thanks 16:15:32 ok, cool 16:15:40 moving on 16:15:46 #topic Current progress on the planned tests 16:16:07 during last week Mirantis team was working on the final lab preparation/automatizartion 16:16:56 #info today we started running data plane tests using the test plan: http://docs.openstack.org/developer/performance-docs/test_plans/tenant_networking/plan.html 16:17:24 right after this baseline rally tests and basic load tests are planned 16:17:38 listomin will add test plan on review before end of week 16:17:45 * ad_rien_ just joined the meeting sorry for the delay 16:17:50 ad_rien_ o/ 16:18:24 so for now we're waiting for the data plane test results, once they will be ready, we'll switch to the next ones 16:19:08 I do not have any particular updates except the mentioned ^^ 16:19:26 ad_rien_, do you have any news about your ZeroMQ research? 16:19:30 no 16:19:35 :D 16:19:36 ack 16:19:37 we postponed this action untill end of june 16:19:47 We are working on Rally/Openstack on G5K 16:19:56 as discussed we want to identify all bottleneck 16:20:01 I was offline last week 16:20:14 #info ad_rien_ 's team has postponed additional ZeroMQ research till the end of June 16:20:21 but the week before we ended the meeting with the idea of trying to mutualize the effort 16:20:25 on that point. 16:20:29 we can discuss it later 16:20:37 ad_rien_ - very nice, thank you 16:20:57 in fact we may cover this topic now 16:21:25 ad_rien_ - we're going to publish our variant of baseline rally scenarios before the end of the week 16:21:34 ok nice. 16:21:38 I mean test plan with these scenarios 16:21:49 once I'll have the review number I'll ping you here, sir 16:22:13 ok my idea was to share the effort on the engine that performs all the tests 16:22:18 in fact we wanted to start this activity 1 week earlier, but had several issues with lab preparation / automatization 16:22:24 ok 16:22:37 If I'm correct you are doing that on your own testbed 16:22:55 our idea is to complete these experiments by conducting the same ones on Grid'5000 16:23:09 we're using rally for testing, but Jenkins for automatic jobs running 16:23:20 i mean rally for these types of tests 16:23:26 we have an experiments engine : http://execo.gforge.inria.fr/doc/latest-stable/ 16:23:41 #link Inria's experiments engine : http://execo.gforge.inria.fr/doc/latest-stable/ 16:24:04 yes but I guessed you need to customize everything (i.e. deployment of specific version of OpenStack, Rally, … and then launch rally) 16:24:09 ad_rien_ - I need to go through the docs 16:24:17 ok 16:24:18 ad_rien_ - sure, we use jenkins for this purpose 16:24:22 ok 16:24:27 so I need to compare the effort 16:24:41 ok, lemme create an action item for myself 16:24:50 ok it would be great if we can share our scripts and see the best of the two worlds 16:25:03 #action DinaBelova go through the http://execo.gforge.inria.fr/doc/latest-stable/ and compare it with the approach used inside Mirantis 16:25:10 ad_rien_ agreed 16:25:30 gokrokve - do you think we can make the jobs / scripts we are using publicly available? 16:25:36 :-) ? 16:25:50 sure 16:25:53 as a part of some new repo (to be created) 16:25:55 cool 16:26:01 I think this is a great approach 16:26:09 we need to make them available 16:26:45 #action DinaBelova check current status of internal Mirantis JJ and python scripts - if they can be made public right now. Evaluate what needs to be done otherwise 16:27:03 ad_rien_ - I'll try to complete this research before next meeting 16:27:35 on our side, we have an engine to deploy on demand an openstack infrastructure and then have an engine to setup the rally server and start the benchs 16:27:48 sadly I'm a bit busy with my current Mirantis OpenStack release managere role, so I cannot guarantee this.. I'll try 16:28:00 #link https://www.grid5000.fr/mediawiki/index.php/OpenStack_deployment_on_Grid5000 16:28:06 ad_rien_ can this engine to be modified to run not only rally? 16:28:10 yes 16:28:25 very cool - for instance, right now rally cannot process data plane tests 16:28:32 therefore I asked :) 16:28:55 thanks, I'll go through the links you published before next meeting 16:28:55 the goal of execo is to run a complete experimental protocol with different parameters 16:29:20 It is quite hard for Inria people to attend meetings each week, 16:29:33 ad_rien_ - where is source code of this engine stored? 16:29:41 I think that you can put such an action for middle of June 16:29:48 ack, cool 16:29:54 anyay, I'll try to do this asap 16:30:03 and let's see when we'll be able to intersect 16:30:24 https://github.com/asimonet/rally-g5k/blob/master/rally-g5k.py 16:30:31 #link https://github.com/asimonet/rally-g5k/blob/master/rally-g5k.py 16:30:33 ack, thanks 16:30:56 it looks like that's all for this particular topic 16:31:19 * hockeynut apologizes, just got done with another meeting 16:31:24 hockeynut :D 16:31:35 ad_rien_, hockeynut - is the time of the meeting not ok for you? 16:31:41 we may try to rearrange it 16:31:55 DinaBelova normally this is fine with me. Just not today 16:32:01 hockeynut ack 16:32:03 ad_rien_ ? 16:32:09 please keep in mind that the engine has been done for Grid'5000 (i.e. we are using our own baremetal deployment tool, called kadeploy) 16:32:25 time is ok, frequency rather hard right now 16:32:31 ad_rien_ ok, thanks 16:32:36 our FTE will join us by end of June 16:32:43 I'll keep this in mind 16:32:47 right now we have only postdoc and Phd guys 16:33:16 and Mathieu (msimonin) is for the moment working only 40% on the project 16:33:27 ok, I got it 16:33:31 good to know 16:34:26 I think that's all regarding the test plans / test runs for now. I think we may skip next topic (OSprofiler updates) as no major things happened since previous week 16:34:39 and rohanion is on vacation (he works on the new drivers structure) 16:35:00 so most probably I'll jump to Open Discussion if no objections 16:35:05 DinaBelova: just a last #link for Execo (can be easier than RTFM) : https://hal.inria.fr/hal-00861886/en 16:35:24 #info Execo (can be easier than RTFM) : https://hal.inria.fr/hal-00861886/en 16:35:28 thanks you sir 16:35:38 #topic Open Discussion 16:35:47 so jumping to the Open Discussion part 16:36:00 as said, no major updates from my side so far 16:36:14 same on the Inria side 16:36:17 probably someone would like to raise his/her own topic 16:36:25 hockeynut, gokrokve, AugieMena ? 16:36:40 I dont have any 16:36:44 nothing new 16:36:46 * hockeynut is working on some rally+barbican integration to get up to speed with rally plugins 16:37:14 hockeynut - may you please share link to the patches (if any of them ready)? 16:37:22 As I know Alex is working on 1000 nodes results document 16:37:30 It should be published in our docs 16:37:31 DinaBelova I certainly will 16:37:46 gokrokve ack. Hope to see this plan published soon :) 16:37:49 hockeynut ack 16:38:19 hockeynut - now? :) 16:38:35 DinaBelova hoping to have them up later today...nothing yet 16:38:51 hockeynut ok, I just wanted to add #link if you have the links now :) 16:38:58 if no - it's also ok 16:39:10 it looks like we covered all important topics for today 16:39:21 thanks for attending and see you later 16:39:24 bye! 16:39:30 #endmeeting