15:30:06 <DinaBelova> #startmeeting Performance Team
15:30:07 <openstack> Meeting started Tue Mar 28 15:30:06 2017 UTC and is due to finish in 60 minutes.  The chair is DinaBelova. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:30:08 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:30:11 <openstack> The meeting name has been set to 'performance_team'
15:30:20 <DinaBelova> hello folks
15:30:21 <DinaBelova> let
15:30:29 <vbala> hi
15:30:31 <DinaBelova> let's see who is around :)
15:30:51 <DinaBelova> vbala hey! welcome back!
15:30:56 <DinaBelova> vbala how' your relocation going on?
15:31:21 <vbala> thanks
15:31:45 <DinaBelova> tovin07 rook rcherrueau akrzos o/
15:31:54 <rook> hello
15:31:58 <tovin07> o/
15:32:00 <rook> o.
15:32:02 <rcherrueau> o/
15:32:05 <rook> o/
15:32:25 <DinaBelova> ok, it looks like we have kinda quorum here :)
15:32:50 <openstackgerrit> Merged openstack/performance-docs master: Kubelet density test on scaling by nodes  https://review.openstack.org/446974
15:32:51 <DinaBelova> let's start with action items
15:32:52 <DinaBelova> #topic Action Items
15:33:09 <DinaBelova> tovin07 one was on you sit and I see it in the open discussion agenda for this meeting
15:33:16 <DinaBelova> sir*
15:33:19 <tovin07> yes
15:33:34 <DinaBelova> #link https://review.openstack.org/#/c/450072/
15:33:40 <DinaBelova> rcherrueau ^^
15:33:52 <DinaBelova> rcherrueau that was requested by you if I'm not mistaken
15:33:53 <rcherrueau> nice
15:34:02 <tovin07> This is my first solution for this
15:34:13 <tovin07> thanks vbala for pointing out some issues
15:34:18 <DinaBelova> let's review and improve patch, I guess it should be merged in some future
15:34:19 <DinaBelova> indeed
15:34:47 <tovin07> vbala: do you have any other idea to improve this?
15:35:11 <vbala> tovin07: i need to look at it more closely
15:35:27 <tovin07> I’m thinking of adding result of WSGI to the trace info as well
15:36:19 <DinaBelova> one more action item is on msimonin
15:36:32 <DinaBelova> rcherrueau don't you know if msimonin pinged Rally PTL?
15:36:51 <DinaBelova> about the osprofiler + rally integration we have discussed some time ago?
15:37:01 <rcherrueau> Yes, he did
15:37:17 <rcherrueau> In his response, Andrey explained that integration of OSProfiler to Rally was planned from the beginning.
15:37:38 <rcherrueau> But, the philosophy to integrate it has to be different that the way it has been done in openstack client.
15:37:50 <DinaBelova> rcherrueau any details?
15:37:52 <rcherrueau> In other word, something like `rally task start ... --os-profile SECRET_KEY', which is the way we integrate it, is not good a integration.
15:38:17 <rcherrueau> The reason, and we (Inria) are agree with him, is that it is better to initialize profiling on each iteration of workload separately.
15:38:35 <rcherrueau> Andrey also said that the trace should be included in the rally report.
15:38:51 <rcherrueau> I don't know if we could help here. I don't know the rally roadmap.
15:38:52 <DinaBelova> yeah, in this case it'll require much more work in the rally itself
15:39:05 <rcherrueau> indeed
15:39:31 <DinaBelova> rcherrueau it depends i guess on how are you ready to help implementing this
15:39:53 <DinaBelova> if you have resources, rally roadmap is flexible enough
15:39:55 <rcherrueau> :D
15:39:58 <DinaBelova> :D
15:40:15 <DinaBelova> so please lemme know if you have a wish to take part in this
15:40:51 <DinaBelova> ok, it looks like we may proceed
15:40:52 <DinaBelova> #topic Current progress on the planned tests
15:40:53 <rcherrueau> I guess this will wait after the end of Boston Summit, but yeah, I will tell you
15:41:01 <DinaBelova> rcherrueau let's start with you sir
15:41:03 <DinaBelova> rcherrueau ack
15:41:15 <rcherrueau> OK
15:41:40 <rcherrueau> Last week, I was on another topic for the massively distributed WG -- understand oslo_messaging specification and its implementation: RMQ/ZMQ/QPid.
15:42:07 <rcherrueau> I try to understand the control flow (no how it behave in at scale).
15:42:07 <rcherrueau> 
15:42:28 <rcherrueau> So, not performance WG related.
15:42:41 <rcherrueau> However, if I am right, the Performance WG conducted some bench on RMQ/ZMQ implementations?
15:42:41 <rcherrueau> 
15:42:52 <DinaBelova> rcherrueau it's ok :)
15:42:54 <DinaBelova> rcherrueau true
15:42:57 <DinaBelova> lemme fin dthe links
15:43:38 <DinaBelova> rcherrueau https://docs.openstack.org/developer/performance-docs/test_plans/mq/plan.html here is the oslo simulator testing for both rabbitmq and zeromq
15:43:48 <rcherrueau> DinaBelova: thanks. I will take a look
15:44:05 <DinaBelova> additionally there was a test with RabbitMQ with and without HA queues being enabled https://docs.openstack.org/developer/performance-docs/test_plans/mq_ha/plan.html
15:44:10 <DinaBelova> rcherrueau ^^
15:44:19 <rcherrueau> DinaBelova: thanks
15:44:31 <DinaBelova> ok, from mirantis side
15:44:52 <DinaBelova> several test results were pushed to the review
15:44:58 <DinaBelova> #link https://review.openstack.org/446974
15:45:04 <DinaBelova> #link https://review.openstack.org/449087
15:45:21 <DinaBelova> the first relates to pure k8s density testing
15:45:33 <DinaBelova> the second one to containerized control plane reliability
15:46:17 <DinaBelova> several more things to be added this week (including what we have done for containerized openstack upgrades testing)
15:46:51 <DinaBelova> rook anything to share from the redhat?
15:47:22 <rook> DinaBelova: Posted a blogpost on a workload integrated into Browbaet
15:47:26 <rook> Browbeat
15:47:33 <rook> but it is really a integration of Rally + Pbench
15:47:34 <DinaBelova> link please :)
15:47:37 * rook gets link
15:47:51 <rook> #link https://browbeatproject.org/2017/03/23/network-performance-rally-pbench-uperf-integration
15:47:58 <DinaBelova> thanks! will take a look
15:49:14 <DinaBelova> folks, anything else to share in "current progress section"?
15:49:41 <DinaBelova> it looks like nope, let's proceed
15:49:42 <rcherrueau> Just one question
15:49:45 <DinaBelova> ack
15:49:51 <DinaBelova> rcherrueau please go ahead
15:49:56 <rcherrueau> yesterday I tried to setup a kolla-ansible deployment with ceilomenter and panko as event dispatcher to then make OSProfiler trace. But it doesn't work, and no idea why!
15:50:14 <DinaBelova> rcherrueau heh :) what symptoms do you have?
15:50:18 <rcherrueau> The main difference between kolla-ansible and devstack deployment is that ceilometer and panko refer to their database through configuration parameters `database.metering_connection' and  `database.event_connection' rather than `database.connection'. I don't know the difference so maybe my problem comes from that.
15:50:18 <rcherrueau> 
15:50:52 <DinaBelova> rcherrueau I don't think this could influence, but frankly speaking I'm not sure
15:50:54 <rcherrueau> Also, I am not a ceilometer expert, but what really surprise me is that, there are no logs in ceilometer-*.log when I do, for instance, a `openstack hypervisor list --os-profile SECRET_KEY'
15:51:56 <DinaBelova> rcherrueau well, ceilometer does not take part in this request actually, so I do not expect them to be, although some logs have to be there I guess - if there are VMs, if polling is hapenning
15:52:15 <rcherrueau> hum ...
15:52:33 <tovin07> rcherrueau:  did you check rabbitmq?
15:52:46 <rcherrueau> tovin07: nop
15:52:47 <tovin07> Is there any thing in profiler.info queue
15:53:15 <rcherrueau> OK thanks for the information, I will take a look into rabbitmq profiler queue
15:53:26 <tovin07> because OSprofiler will put raw trace info in rabbitmq first
15:54:00 <tovin07> then ceilometer pull it out and do something crazy and then —> panko —> mysql
15:54:11 <tovin07> that’s the default scenario
15:54:29 <rcherrueau> OK Great, maybe my connection_string is miss set
15:54:42 <tovin07> yup
15:54:44 <tovin07> maybe :D
15:54:53 <rcherrueau> tovin07: thanks for the clarification on OSProfiler workflow
15:55:18 <DinaBelova> #topic Open Discussion
15:55:25 <DinaBelova> tovin07 I see one more item from you
15:55:29 <tovin07> yes
15:55:32 <DinaBelova> regarding the messaging osprofiler driver
15:55:53 <tovin07> Why do we need admin context in messaging driver (with oslo.messaging) only?
15:56:05 <DinaBelova> frankly speaking I am not sure about admin context here
15:56:21 <tovin07> I asked you this a long time ago
15:56:42 <tovin07> If we see this https://review.openstack.org/#/c/316799
15:57:02 <tovin07> I don’t know why we need admin context :D (too)
15:57:15 <DinaBelova> tovin07 it looks like I was confused with something else
15:57:29 <DinaBelova> technically speaking admin context is not necessary I believe
15:57:53 <tovin07> Because all the security thing in OSprofile is based on HMAC mechanism
15:57:59 <DinaBelova> tovin07 true
15:58:20 <DinaBelova> and context is used only for transport generation in oslo messaging in messaging driver
15:58:40 <tovin07> yup
15:58:50 <tovin07> let me investigate it
15:58:55 <DinaBelova> tovin07 ack, thank you
15:59:03 <tovin07> maybe we can remove admin context in this case
15:59:05 <DinaBelova> originally this was coming from Boris code
15:59:13 <tovin07> and use normal context
15:59:17 <DinaBelova> ++
15:59:20 <vbala> I think we can remove it
15:59:27 <DinaBelova> vbala agree
15:59:44 <tovin07> vbala: agree too
15:59:59 <tovin07> ok, thanks :D
16:00:01 <DinaBelova> cool, thank you tovin07 for raising this question
16:00:19 <DinaBelova> anything else to cover today?
16:00:23 <tovin07> yes
16:00:28 <DinaBelova> please go ahead
16:00:39 <tovin07> Your presentation in Boston
16:00:52 <DinaBelova> tovin07 yeah
16:00:54 <DinaBelova> good you asked
16:01:09 <DinaBelova> #info sadly I won't be able to attend the Boston summit
16:01:24 <DinaBelova> I have an accidental type of event
16:01:28 <tovin07> oh, sad news
16:01:32 <DinaBelova> that I cannot move
16:01:39 <DinaBelova> so vbala is going to present without me
16:01:40 <tovin07> vbala, can you attend
16:01:52 <DinaBelova> vbala I'll help you with presentation creation
16:01:58 <vbala> tovin07: yes, i'm planning to
16:02:06 <vbala> thanks Dina
16:02:11 <tovin07> good to knows
16:02:23 <rcherrueau> And what about you tovin07?
16:02:24 <tovin07> this time, I also apply for TSP
16:02:31 <tovin07> however, I failed :))
16:02:51 <tovin07> I’m looking for other chance
16:02:52 <DinaBelova> tovin07 heh, those guys are cruel
16:03:04 <tovin07> Hope that I can go in other way
16:03:21 <DinaBelova> tovin07 please keep us updated
16:03:26 <tovin07> yes
16:03:36 <tovin07> that’s all for me today
16:03:41 <tovin07> rcherrueau:
16:04:05 <tovin07> Do you have any comment on my patch in SQL/function result in trace
16:04:16 <tovin07> Please let me know :D
16:04:32 <rcherrueau> tovin07: yes, I will do through gerrit :)
16:04:56 <DinaBelova> thanks rcherrueau :) waiting for comments
16:05:14 <DinaBelova> ok, it looks like we may wrap up today meeting
16:05:19 <rcherrueau> tovin07: btw, thanks for implementing this :thumbsup:
16:05:35 <DinaBelova> thanks tovin07 rcherrueau vbala rook
16:05:36 <tovin07> rcherrueau: my pleasure!
16:05:43 <vbala> thanks
16:05:47 <DinaBelova> #endmeeting