[From IRC] andreykurilin : :)
[From IRC] hai_shi : pls help review this patch https://review.openstack.org/#/c/431624/ ;)
[From IRC] tovin07 : andreykurilin: hi
[From IRC] tovin07 : recently, rcherrueau from inria asked you about Rally + OSprofiler
[From IRC] tovin07 : Do you remember that? :D
[From IRC] andreykurilin : hai_shi: yeah... I remember about it. It is in my opened tabs. Actually, I dislike how is validation in embedded in CLI and not sure that we should save in database errors like "unable to open file". Today I want to experiment with validation and then will return to you patch
[From IRC] andreykurilin : tovin07: yes, I remember about it
[From IRC] andreykurilin : :)
[From IRC] tovin07 : so, do you guys have any plan to make it work?
[From IRC] andreykurilin : tovin07, my reply to inria folks maybe was not accurate, so they thought that it will take too much human-resources which they do not have :(
[From IRC] andreykurilin : actually, it should not be a hard task
[From IRC] andreykurilin : it is true
[From IRC] andreykurilin : but as I said, it should not be hard to implement it
[From IRC] tovin07 : for example, we init nova client with something like this: nova.Client(something but **kwargs)
[From IRC] tovin07 : yup
[From IRC] tovin07 : so, I think it’s good to do that soon ==> we can measure overhead of using OSprofiler all-the-time in production
[From IRC] tovin07 : I agree that it is not a hard to implement :D
[From IRC] andreykurilin : tovin07: [implementation details] as far as I understand how osprofiler works, while osprofiler initialization it uses global variable, so initialization of one openstack client means initialzation fro all clients. That is why there is no need to transmit profiler_enable to nova client itself.
[From IRC] andreykurilin : If you are interested in implementation, I'm ok to extend rally with osprofiler feauture
[From IRC] andreykurilin : we can start from the spec
[From IRC] tovin07 : I will follow it if I can
[From IRC] tovin07 : thank you andreykurilin
[From IRC] tovin07 : that’s all from me
[From IRC] andreykurilin : Ok, I'll start a spec about it. Can you share your email(you can do it in direct messages to avoid logging)? I'll add you to thread with inrea folks when the spec will ready.
[From IRC] tovin07 : okay, I’ll send a dirrect message
[From IRC] andreykurilin : Anything else to discuss?
[From IRC] hai_shi : a small question
[From IRC] hai_shi : if cinder_v1 have not update_volumes() function, it looks we can not add this update_volues() in BlockStorage
[From IRC] chenhb : we call osprofile.trace while calling client?
[From IRC] hai_shi : yep
[From IRC] andreykurilin : If update_volume_type is a specific for Cinder V2, only Cinder V2 should include that method and scenario should have proper validator that checks that that cinder v2 is enabled.
[From IRC] andreykurilin : chenhb: no we will just initialize osprofiler before each iteration and save trace id into iteration result then it will be displayed in reports as text area.
[From IRC] andreykurilin : it is a first step for integration
[From IRC] andreykurilin : then, it will be possible to embed osprofiler report into rally's report
[From IRC] chenhb : got it
[From IRC] hai_shi : fine
[From IRC] tovin07 : andreykurilin: (y)
[From IRC] chenhb : so we can get a unify report
[From IRC] andreykurilin : *we will able to get
[From IRC] andreykurilin : now we cannot do it :)
[From IRC] hai_shi : no question
[From IRC] chenhb : nothing from me
[From IRC] andreykurilin : ok, let's finish
[From IRC] andreykurilin : #endmeeting