07:30:03 <anac1> #startmeeting Yardstick work meeting
07:30:03 <collabot> Meeting started Thu Aug 27 07:30:03 2015 UTC.  The chair is anac1. Information about MeetBot at http://wiki.debian.org/MeetBot.
07:30:03 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic.
07:30:03 <collabot> The meeting name has been set to 'yardstick_work_meeting'
07:30:11 <anac1> #info Ana Cunha
07:30:22 <raindirve> #info Sean Wenström
07:30:25 <anac1> hi, please info your names
07:30:33 <QiLiang> QiLiang
07:31:01 <anac1> #topic Interaction between dispatcher & output manager
07:31:09 <Patrick_> #info Patrick
07:31:35 <anac1> let's discuss diaptacher and output manager to synchronise and agree on the solution
07:31:49 <QiLiang> ok
07:31:57 <QiLiang> I think we can use OutputMgr as the base class, and FileDispatcher, HTTPDispatcher be the subclass of OutputMgr.
07:32:34 <anac1> raindirve?
07:32:56 <QiLiang> Each subclass implement it's own sotrage part, just rewrite _output_serializer_streamer
07:33:05 <raindirve> interesting, I definitely agree that the dispatcher would make a great backend for the OutputMgr
07:33:17 <raindirve> subclass sounds like a good way to go
07:33:41 <anac1> #info suggestion to use OutputMgr as base class
07:33:57 <QiLiang> use subclass can dismiss confisuion
07:34:33 <anac1> #info and FileDispatcher, HTTPDispatcher as subclass
07:35:06 <anac1> raindirve, do i understand you agree?
07:35:21 <raindirve> yes
07:35:28 <anac1> jnon, any opinion?
07:35:52 <anac1> qiliang, ok for you, right?
07:35:59 <QiLiang> yes
07:36:01 <Patrick_> So which dispatcher should be configured in command?
07:36:13 <jnon_> i missed the beginning
07:36:46 <QiLiang> for the configuration part i think we can use a config file like rally did
07:38:26 <Patrick_> ok
07:38:45 <anac1> jnon, propose to use OutputManager as base class and dispacher as subclass
07:39:16 <jnon_> ok that sounds good to me
07:39:42 <Patrick_> So we need a base config file for common configuration in Yardstick Testing?
07:39:50 <anac1> #agreed use OutputManager as base class and dispacher as subclass
07:39:51 <QiLiang> and i will rename the name, not use dispatcher , just xxOutput
07:39:59 <anac1> ok
07:40:03 <jnon_> yes
07:40:42 <anac1> #action QiLiang will rename dispatcher to fit in the agreed solution
07:41:00 <anac1> anything else on outputmanager & dispatcher?
07:41:25 <Jingwen> #info Jingwen
07:41:26 <jnon_> no
07:41:29 <anac1> let's move on
07:41:37 <Jingwen> sorry, i am late.
07:41:43 <anac1> no problem
07:41:53 <anac1> #topic Test Case for HA
07:42:11 <anac1> #info meeting with HA yesterday
07:43:02 <anac1> #info requirement from HA to Yardstick: test case to "kill" one instance of a service running in the controller node
07:43:13 <anac1> #info slides available, presented yesterday
07:43:44 <anac1> #info not decided what service we should start with
07:43:52 <anac1> #info it could be nova
07:44:18 <anac1> #info yardstick can use china mobile opnfv lab
07:44:19 <Patrick_> I have a question, do we know the behavior of OPNFV 3+2 HA system when we kill the OpenStack service on Control Node?
07:44:52 <anac1> that is what they (HA) want to understand
07:45:33 <Patrick_> for example, if we killed nova in one Control node, does another nova will give the service in another Control Node?
07:45:58 <jnon_> It should probably failover to another or same node
07:46:04 <anac1> if we kill one, the other have to come up and tke over
07:46:22 <anac1> HA wants to measure how long time does it take for that to happen
07:46:26 <Patrick_> ok.
07:47:15 <anac1> #info a test case will be drafted together with HA
07:47:29 <QiLiang> i think this ha test case is very different for the test cases we implemented
07:47:39 <anac1> yes
07:47:40 <jnon_> yes its different
07:47:59 <anac1> #info what questions we have to HA?
07:48:36 <QiLiang> if we kill nova instacne this looks like we test if nova api is available
07:48:46 <anac1> #info do we understand what HA wants to do?
07:49:37 <jnon_> Currently I dont think it fit so well with what yardstick is doing now, but we can have look at it
07:50:03 <jnon_> yes check nova-api and crm status maybe\
07:50:11 <anac1> #action all - analyse the requirments from HA until next monday and bring opinion
07:50:50 <anac1> is that ok that we discuss again next meeting?
07:50:59 <jnon_> yes
07:51:02 <QiLiang> ok
07:51:04 <Patrick_> ok
07:51:53 <anac1> great !
07:51:54 <jnon_> Another thing, we should probably from now on write unit tests for all new features we push.  !?
07:52:41 <QiLiang> aggree
07:52:51 <anac1> ok for everyone?
07:52:52 <Patrick_> totally agree
07:52:55 <raindirve> yes
07:53:03 <Jingwen> aggree
07:53:26 <anac1> #agreed write unit tests for each new feature added
07:53:51 <anac1> #topic miscellaneous
07:54:24 <anac1> #info youtube video with yardstick installation in opnfv channel
07:54:56 <anac1> #info design meeting with vsperf next monday
07:55:02 <Patrick_> Ana, about Yardstick-112， I found that the tools Iperf/Iperf3 can measure the UDP jitter( packet delay variation) between client and server.
07:55:29 <Patrick_> As Yardstick has alrealy used Iperf3 to measure the bandwidth, I think we can expend this scenario to support more network metrics measurement.
07:55:37 <anac1> great idea
07:56:06 <anac1> #agreed use iperf3 for Yardstick-112
07:56:08 <Jingwen> Ana, i'm sorry , i used packet delay variation as end-to-end latency, but they are not the same
07:56:22 <Patrick_> and about processing speed, I 'm still investigate in this area. Hoping next week I will finish this investigation and start coding.
07:56:30 <anac1> Jingwen, ok - thanks for clarification
07:57:15 <anac1> Patrick, you have the task, let us know when you have a proposal, that sound a good way forward
07:57:30 <Patrick_> ok
07:58:29 <anac1> #info a discussion on a dashboard to report test results has started
07:59:05 <anac1> #info functest is driving
07:59:26 <QiLiang> when ?
07:59:42 <Patrick_> the topic will continue discussing tonight in Testing Meeting
08:00:00 <QiLiang> i see
08:00:02 <anac1> discussion on mailing list, continue in test&performance today
08:01:07 <anac1> ok, that's all for today
08:01:11 <anac1> #endmeeting