17:02:25 #startmeeting Rally 17:02:25 Meeting started Tue Nov 18 17:02:25 2014 UTC and is due to finish in 60 minutes. The chair is boris-42. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:02:26 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 17:02:29 The meeting name has been set to 'rally' 17:02:52 #topic OpenStack summit 17:03:05 Okay OpenStack summit was quite productive as usually 17:03:24 and I would like just to share with some interesting documents 17:03:43 #link https://etherpad.openstack.org/p/kilo-rally-osprofiler 17:03:57 result of Rally design session 17:04:34 #link http://www.youtube.com/watch?v=77B9Un4xtK0 (one talk) 17:05:04 #link http://www.youtube.com/watch?v=_uOnZKu7OSY (another talk) 17:05:56 #link http://www.youtube.com/watch?v=iNKKvLFhkpE (and one more time a bit about rally) 17:06:35 As well there were nonofficial discussions 17:06:42 about benchmarking Network with Rally 17:06:48 and Rally as a Service 17:06:55 we decide to do both=) 17:07:06 Any questions?) 17:07:21 (so I'll start mailing list threads about this) 17:07:27 yes 17:07:32 zhiyan: ?) 17:07:46 hi all:) 17:07:50 for osprofiler integration, i would like to take trove and zaqar 17:08:11 zhiyan: okay 17:08:13 actually trove ones are under review already 17:08:22 zhiyan: let's change topic 17:08:28 #topic OSprofiler integration 17:08:38 zhiyan: so could you share links on reviews? 17:09:12 another one is about adding a new mongodb based notifier 17:09:17 yes, 2 secs pls 17:09:32 https://review.openstack.org/#/c/116653/ 17:09:49 https://review.openstack.org/#/c/116654/ 17:09:55 #link https://review.openstack.org/#/c/116653/ 17:10:02 #link https://review.openstack.org/#/c/116654/ 17:10:09 https://review.openstack.org/#/c/116671/ 17:10:15 oops, thanks 17:10:19 ya 17:10:39 So I will try to find time to review those patches 17:10:57 zhiyan: about mongodb backend we will need to setup dsvm functional job 17:11:04 btw, i have talked them with trove team closely, them all agreed to land osprofiler stuff asap in k 17:11:05 zhiyan: for osprofiler before doing it.. 17:11:16 zhiyan: nice 17:11:23 zhiyan: btw what about using trove for backend?) 17:11:26 you are late.. lol 17:11:34 zhiyan: where where?) 17:12:06 i'm making it locally, just verifiying it before your ping :) 17:12:15 ok, i see your idea 17:12:44 i mean dsvm one 17:12:58 zhiyan: I am late with what?) 17:13:03 zhiyan: adding job? 17:13:05 but probably i need to know more details 17:13:37 joke, because i already started doing it 17:14:25 zhiyan: ah okay, so actually before we land your work we need to have functional testing 17:14:43 zhiyan: cause it's already merged to heat, cinder, glance and I don't won't to break whole world= 17:14:54 make sense 17:15:01 zhiyan: okay I'll look at your patches when you publish them=) 17:15:07 so let's move to next topic=) 17:15:08 thanks! 17:15:14 #topic Generic cleanup 17:15:38 Okay we finally finished refactoring of generic cleanup mechanism 17:15:53 so now it's resource oriented and it's really simple to extend by plugins 17:16:15 all plugins are here https://github.com/stackforge/rally/blob/master/rally/benchmark/context/cleanup/resources.py 17:16:29 so it's just about making subclasses and overriding probably some functions.. 17:16:58 This new cleanup engine is much safer and it has repeats and speed limiations and better logging of non deleted resources 17:17:06 so one more nice improvments 17:17:43 Any questions?) 17:18:06 okay seem no questions lol 17:18:17 #topic Next steps for generic cleanup 17:18:50 So there is only one left step, to make disaster cleanup, when another instance of Rally can clean up everthing in case of failure 17:19:00 msdubov is going to work on this 17:19:31 msdubov_: aorund? 17:20:03 okay seems like he is not around=) 17:20:07 let's move to next topic 17:20:11 lol 17:20:17 zhiyan: ya=) 17:20:31 #topic New better Rally reports and future steps 17:20:37 amaretskiy: could you share with updates 17:20:51 https://review.openstack.org/#/c/131844/ 17:21:01 here are a lot of improvements 17:21:06 + bugfix 17:21:33 SLA, Failues, Scenario output 17:21:36 etc... 17:22:00 http://logs.openstack.org/44/131844/18/check/gate-rally-dsvm-rally/9872f9c/ 17:22:16 http://logs.openstack.org/44/131844/18/check/gate-rally-dsvm-rally/9872f9c/rally-plot/results.html.gz#/VMTasks.boot_runcommand_delete 17:22:28 http://logs.openstack.org/44/131844/18/check/gate-rally-dsvm-rally/9872f9c/rally-plot/results.html.gz#/Dummy.dummy_exception_probability#failures 17:22:45 amaretskiy: --verbose plese tell us what are you doing 17:22:52 lol 17:23:00 there are links above 17:23:05 take a look at tabs 17:23:22 Failures - we see there an exceptions 17:23:35 that happened during the scenario run 17:23:45 another Tab - Output 17:24:02 this is where we see scenario output (if any) 17:24:21 Also there is a SLA data at the top of Overview tab 17:25:00 there are also another improvements, like re-styling of page http://logs.openstack.org/44/131844/18/check/gate-rally-dsvm-rally/9872f9c/ 17:25:41 That is what is going to be merged soon (I hope :) ) 17:25:51 And there ar esome plans for next patch 17:25:53 amaretskiy: so what about future stuff that you are going to do?) 17:26:09 I'm going to introduce "overview page" 17:26:31 where we can observe all scenarion within the benchmark in condensed view 17:26:46 so there will be more easy to compare the results 17:27:23 amaretskiy: nice 17:27:47 and I'm going to aggregare Failures - do not repeat similar 17:28:01 just show differebt failures and their count 17:28:32 + fix browser back-forward buttons behavoir 17:28:46 now there is no effect - that is an issue 17:30:14 boris-42, that's all from my side :) 17:31:14 amaretskiy: great work on reports 17:31:21 amaretskiy: now our bosses are happy=) 17:31:30 :) 17:31:42 arcimboldo: hi hi 17:31:45 hi all 17:31:50 happy to join you 17:32:03 arcimboldo hi 17:32:17 arcimboldo: so we have here some kind of weekly meeting 17:32:42 Sorry seems like I've had some connection issues 17:32:43 arcimboldo: where we are trying to share updates and dicuss different stuff 17:33:07 #topic Dropping root requirements for rally setup.py 17:33:25 ok I think I can start 17:33:28 arcimboldo: so your idea is to use better locations for files * 17:33:38 yes, right now, there are two issues: 17:33:46 1) the install script fails if you are not root 17:33:58 2) setup.cfg tries to install the completion file in /etc/bash_completion.d 17:34:10 however, rally does not need to run as root, so I would suggest the following: 17:34:24 1) replace /etc/bash_completion with etc/bash_completion in setup.cfg. 17:34:44 this will install in either VIRTUALENV_PATH/etc/bash_completion, if you are running inside a virtualenv, or in /usr/local/etc/bash_completion otherwise 17:35:03 2) the install script will check if you are root or not, and if you are installing in a virtualenv 17:35:05 hm maybe ~/etc/bash_completion ? 17:35:14 otherwise it may produce some issues? 17:35:16 it's not standard 17:35:24 ah 17:35:32 aahhh I see 17:35:43 if you are running as root, _and_ not installing in a virtualenv, it makes sense to copy the file in /etc/bash_completion 17:35:50 this could be done by teh script 17:35:57 arcimboldo: yep 17:36:07 amaretskiy: one more question, what about rally.conf 17:36:09 if you are running as a user, a line can be added to the .bashrc 17:36:16 arcimboldo: ^ 17:36:56 well, rally.conf is not automatically copied in /etc/ as far as I understand, right? 17:37:13 it's copied by the script 17:37:14 rally.conf can be installed into user $HOME dir. I think 17:37:33 so again, if the script is run as normal user can be copied in ~/.rally/rally.conf 17:37:43 (which is where rally is looking for the configuration file) 17:38:11 arcimboldo: okay I think this totally makes sense 17:38:11 maybe it is reasonable to store rally.conf in virtualenv root dir? 17:38:29 imo nothing to discuss here. it is enough to file a bug, and assign it to amaretskiy or me 17:38:29 amaretskiy: good point=) 17:38:45 rediskin: arcimboldo is going to try to do it=) 17:38:54 arcimboldo: are you?) 17:38:56 amaretskiy, well, data files are usually installed in the virtualenv, but configuration files arent, as far as I know 17:39:02 boris-42, yes I can do it 17:39:20 arcimboldo: great so you can actually file a bug it's quite simple to do 17:39:44 arcimboldo: https://bugs.launchpad.net/rally 17:39:51 arcimboldo: and then try to fix it 17:39:56 ok 17:40:01 arcimboldo: as far as I remember I shared with you HowTo? 17:40:09 yes 17:40:22 reading it it's in my todo list :) 17:40:24 arcimboldo: first time it's a bit painfull=) 17:40:49 https://github.com/openstack/oslo.config/blob/master/oslo/config/cfg.py#L476 17:41:13 olso.config looks for configs not only at /etc 17:42:03 oanufriev: yep 17:42:13 oanufriev, I've already installed rally as standard user, and put my configuration in ~/.rally/rally.conf, so it works already 17:42:18 the only issue is with the installation 17:43:35 arcimboldo: so okay I think we discussed this well= 17:43:56 arcimboldo: btw did you run sucessfully against Folsom?) 17:44:08 boris-42, yes! 17:44:13 arcimboldo: WTF 17:44:14 =) 17:44:28 I've put a try/execpt around call that was raising an exception 17:44:37 arcimboldo: http://m.memegen.com/6vu5mf.jpg =) 17:44:58 arcimboldo: I think we should do that in upstream 17:45:04 inside _remove_associated_networks, around the to utils.check_service_status 17:45:10 arcimboldo: yep 17:45:14 but I only tested creation of virtual machines 17:45:15 arcimboldo: another good patch to do 17:45:24 arcimboldo: at least something works! 17:45:26 I don't know what else could fail 17:45:41 arcimboldo: but at least something works that is great=) 17:46:08 arcimboldo: so it makes sense to put try/except there 17:46:18 arcimboldo: one more patch for u=) 17:46:20 but the try/except should probably stay in benchmarks.utils, inside function check_service_status 17:46:40 if anything throws an exception, it should returns false. 17:47:16 arcimboldo: not sure about that, cause it can hide evil bugs 17:47:50 arcimboldo: maybe just put in user context on removing networks.. 17:48:15 what about logging the exception raised? 17:48:29 arcimboldo: ya something like that 17:48:30 raising an exception there is quite nasty, because you cannot cleanup 17:48:47 I had to delete the tenant by hand 17:49:00 arcimboldo: hm 17:49:08 arcimboldo: that shouldn't happend 17:49:14 so either you run it *before* creating anything, or you try to continue and then decide if perform the tests or just exit 17:49:20 (after proper cleanup) 17:50:08 arcimboldo: ah yep 17:50:10 arcimboldo: https://github.com/stackforge/rally/blob/master/rally/benchmark/context/users.py#L199 17:50:21 arcimboldo: msdubov_ ^ you forgot to put there try/excpet 17:50:42 arcimboldo: so the idea is to log failures * 17:50:55 arcimboldo: or at least warning there and put there try/except 17:51:11 It seems to me that this failure is not critical, since you can proceed even if you don't get the list of the services. 17:51:16 so you should continue 17:51:33 at least, in this very specific case: Folsom with nova-network 17:51:54 arcimboldo: so this code _remove_associated stuff is realted only to nova-network 17:51:58 boris-42 As far as I remember it wasn't there before this patch? I tried not to change anything related to this 17:52:01 arcimboldo: and shouldn't be run agiant nova + neutron 17:52:22 msdubov_: hm you were last person who touched that piece of code 17:52:25 msdubov_: so=) 17:53:00 https://github.com/stackforge/rally/blob/9089ea4e56a8426538303a7c4abaa4884576a90d/rally/benchmark/context/users.py that's the previos version 17:53:08 https://github.com/stackforge/rally/blob/9089ea4e56a8426538303a7c4abaa4884576a90d/rally/benchmark/context/users.py#L173 17:53:09 msdubov_: doesn't matter =) 17:53:30 There is try..except in the look 17:53:32 *loop 17:53:42 msdubov_: it doesn't delete anything 17:53:47 msdubov_: if it raises exception 17:54:27 boris-42 hm it just logs a warning, doesn't it? 17:54:29 arcimboldo: so as you found that bug, could you add some try/excpet good patch for beggign 17:54:32 https://github.com/stackforge/rally/blob/master/rally/benchmark/context/users.py#L130 17:54:54 boris-42 Exception is in the nova client initialization? 17:54:55 msdubov_: https://github.com/stackforge/rally/blob/master/rally/benchmark/context/users.py#L120 and what if this raises exception 17:55:04 msdubov_: ^ that line 17:55:09 boris-42 Okay 17:56:03 Okay so meeting is quite close to end 17:56:07 #topic Free discussion 17:56:19 any questions? ideas? proposals?) 17:56:41 Well since I missed some part of the meeting, I don't know whether you discussed rally info 17:56:58 There is going to be an update: https://review.openstack.org/#/c/134666/ 17:57:12 It introduces some explanatory texts about main entities in Rally 17:57:22 and improves the overall user experience with this command 17:57:26 Just to share 17:57:55 As well as I'm hoping to finish soon covering Rally with docstrings, so that "rally info find ..." will return info for any query 17:58:06 msdubov_: nice 17:58:25 okay we have to end meeting 17:58:29 #endmeeting