13:01:08 <Qiming__> #startmeeting senlin
13:01:09 <openstack> Meeting started Tue Sep  8 13:01:08 2015 UTC and is due to finish in 60 minutes.  The chair is Qiming__. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:11 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:13 <openstack> The meeting name has been set to 'senlin'
13:01:21 <Qiming__> hello
13:07:11 <haiwei> sorry be late
13:07:20 <Qiming__> hi
13:07:32 <haiwei> hi
13:08:01 <Qiming__> yanyan has got some network problem
13:08:10 <Qiming__> let's wait a few minutes
13:08:16 <haiwei> ok
13:08:25 <Qiming__> anything new from your side?
13:08:56 <Qiming__> not sure cindia is online or not
13:09:06 <yanyanhu> hi, Qiming__ , sorry I'm late
13:09:11 <haiwei> nothing special
13:09:12 <yanyanhu> using web irc now
13:09:30 <haiwei> but maybe another person of my team will join senlin
13:09:41 <Qiming__> okay, let's do a quick update first
13:09:55 <Qiming__> haiwei, great!
13:10:16 <Qiming__> we have a looooooooot of things to do, really need more hands
13:10:21 <haiwei> and also I have invited one to senlin-dushboard
13:10:34 <yanyanhu> cool :)
13:10:51 <Qiming__> cool, please help link the UI guy to zhenguo
13:11:07 <haiwei> in fact, recently i have done few jobs
13:11:09 <Qiming__> looking at the work items etherpad now
13:11:21 <Qiming__> don't worry, you will catch up
13:11:24 <Qiming__> ;)
13:11:52 <haiwei> I am trying to understand as much as possible about senlin
13:11:52 <Qiming__> test cases for drivers ... em, we can remove them all today
13:12:03 <haiwei> godd
13:12:05 <haiwei> good
13:12:21 <Qiming__> haiwei, we are sorry for making things so complicated, :P
13:12:49 <haiwei> no, that's where senlin is interesting
13:12:50 <Qiming__> finally, we are about to clean the l-3 backlog
13:13:42 <Qiming__> functional tests ... yanyanhu has done good job there, hopefully scaling test cases will pass soon, after some bug fixes
13:13:51 <yanyanhu> yep
13:13:55 <haiwei> people are interested in when to make an namespace shift
13:14:08 <Qiming__> haiwei, we are in queue
13:14:10 <yanyanhu> hopefully this can be done tomorrow
13:14:49 <Qiming__> we won't be there when we are not proving the project's usefulness
13:15:42 <haiwei> yes, is there any risk that the namespace shift will be rejected?
13:15:54 <Qiming__> shifting namespace tomorrow may and may not be a good thing, if we cannot get a few POCs up and running
13:15:57 <haiwei> it seems kolla is still in stackforge
13:16:15 <Qiming__> haiwei, we keep work hard, things will happen
13:16:24 <haiwei> ok
13:16:48 <Qiming__> just had a conf call with SUR team this afternoon
13:17:02 <yanyanhu> shifting namespace is not the goal, making the project popular is :)
13:17:06 <Qiming__> they are stuck by some problems
13:17:19 <haiwei> who is SUR team
13:17:22 <Qiming__> yanyanhu, right
13:17:32 <Qiming__> it's a shared university research program
13:17:46 <haiwei> ok
13:18:00 <haiwei> what are the problems
13:18:26 <Qiming__> they are helping us developing clusters of containers
13:18:33 <Qiming__> in the framework of magnum
13:19:47 <Qiming__> they have just shared with us the code they have developed, unfortunately, I cannot find the links at the moment
13:19:56 <Qiming__> will share those with you later
13:20:07 <haiwei> ok
13:20:21 <Qiming__> they are not only working on operating container clusters, but also on auto-scale, load-balance them
13:20:23 <yanyanhu> nice, then we can make some reviews about the code
13:20:28 <Qiming__> yes
13:20:51 <haiwei> the problems are?
13:21:06 <Qiming__> I hope julio can play an important role there
13:21:48 <Qiming__> how to make auto-scaling/load-balancing work between the two layers of entities: VMs, containers
13:22:00 <Qiming__> that is an interesting topic to explore
13:22:41 <Qiming__> they already have a prototype using Docker, now they have started playing with kubernetes
13:23:04 <Qiming__> #topic placement policy
13:23:31 <Qiming__> Cindia has submitted a patch doing placement
13:23:43 <Qiming__> it is based on vSphere DRS
13:23:52 <Qiming__> but the idea is pretty generic
13:23:56 <Qiming__> we can generalize that
13:24:28 <Qiming__> basically, we have three layers to consider: hosts, availability-zones (aggregates) and regions
13:24:58 <Qiming__> for hosts, nova has different kinds of scheduler filters
13:25:22 <Qiming__> we evaluated the server-group API extension, which has many problems
13:25:32 <Qiming__> we may avoid using that
13:26:10 <Qiming__> the most significant problem is that you cannot add/remove members in a server-group
13:26:51 <Qiming__> as for availability zones and regions, a generic policy can be designed, Cindia and I will look into that
13:27:01 <yanyanhu> so you can just create/delete server-group as a batch
13:27:07 <haiwei> what are the scheduler filters in nova?
13:27:30 <Qiming__> haiwei, http://git.openstack.org/cgit/openstack/nova/tree/nova/scheduler/filters
13:28:01 <haiwei> thanks
13:28:27 <Qiming__> if you are making good use of them, you can already do a good job managing the placement of your cluster nodes
13:28:58 <Qiming__> yanyanhu, yes, there is no update operation to a server-group
13:29:16 <yanyanhu> understand
13:29:35 <Qiming__> I'm thinking maybe we need to get the PRS team involved in the design
13:29:38 <haiwei> this is the host layer?
13:29:59 <Qiming__> haiwei, mostly host layer and also some support to availability zones
13:30:04 <Qiming__> aggregates
13:31:09 <Qiming__> maintaining the distribution of nodes is a challenge
13:31:40 <Qiming__> any questions on placement policy?
13:31:56 <haiwei> the first step of placement policy is in host layer?
13:32:06 <haiwei> and also azs
13:32:29 <Qiming__> yep, haiwei, we are targeting a POC on that
13:33:43 <Qiming__> some months ago, I saw a demo from PRS (platform computing) team doing various kinds of placement control
13:33:48 <haiwei> I am not understand senlin generally well, does senlin can do autoscaling of vms in one host, don't consider the az
13:34:16 <haiwei> now
13:34:23 <Qiming__> generally speaking, we leave that scheduling decision to nova
13:34:33 <Qiming__> when we are talking about VMs
13:34:53 <Qiming__> however, we can provide some hints to Nova to gain a better control
13:35:07 <Qiming__> that is what the 'filters'/scheduler is about
13:35:07 <haiwei> so using nova scheduler the answer is yes
13:36:16 <Qiming__> #topic summit planning
13:36:16 <haiwei> ok
13:36:48 <Qiming__> for the coming summit, we need to schedule some discussions
13:37:13 <Qiming__> it seems that the meeting rooms are pretty limited
13:37:38 <haiwei> yes, tokyo is small
13:37:41 <Qiming__> hopefully we can still find some places to have a discussion, for planning
13:37:50 <yanyanhu> that will be the best
13:38:08 <haiwei> you'd better reserve your hotel earlier
13:38:22 <Qiming__> need to check who will/won't come
13:38:34 <yanyanhu> haiwei: we have booked the room :)
13:38:47 <haiwei> great
13:38:53 <yanyanhu> although not the nearest hotel...
13:38:55 <Qiming__> sadly, our talk proposal wasn't accepted this time
13:39:17 <yanyanhu> it's about 2.5 miles away from the conference hotel
13:39:19 <haiwei> yes
13:39:21 <Qiming__> totally confused about the voting and selection process
13:39:35 <Qiming__> it is not transparent
13:39:37 <haiwei> a little far
13:40:05 <Qiming__> last time in Vancouver, I know that an India company got 4 sessions accepted
13:40:06 <haiwei> i know some secrets about the voting
13:40:08 <Qiming__> sigh ...
13:40:42 <yanyanhu> anyway, we can plan a meetup I think
13:40:46 <Qiming__> there should be no secrets at all
13:40:51 <haiwei> they make some groups, for compute or network and so on,
13:41:00 <yanyanhu> to discuss the road map for the coming cycle
13:41:19 <Qiming__> surely we will schedule a meetup
13:41:37 <Qiming__> prioritize things we want to complete during M cycle
13:41:55 <yanyanhu> yes
13:42:07 <Qiming__> will check with julio and lisa and others
13:42:34 <haiwei_> I dropped, maybe someone found I am talking secret
13:43:00 <Qiming__> :(
13:43:17 <Qiming__> #topic open discussions
13:43:32 <haiwei_> besides you two and me, are there other guys will join the senlin discussion
13:43:58 <yanyanhu> hope xinhui can also join us
13:44:10 <Qiming__> I think some guys from huawei, intel and vmware will join
13:45:13 <Qiming__> anyway, with the last pieces of test cases merged
13:45:35 <Qiming__> we will focus on getting everything together and do some show case
13:46:02 <yanyanhu> nice
13:46:39 <Qiming__> anything else?
13:46:49 <yanyanhu> nope from me
13:46:53 <haiwei_> 3:0 just now
13:47:03 <yanyanhu> soccer?
13:47:27 <haiwei_> you know it
13:47:41 <yanyanhu> :)
13:47:47 <Qiming__> okay, let's release the channel
13:47:54 <Qiming__> do your small talks somewhere, ;)
13:47:58 <Qiming__> thanks for joining
13:48:04 <Qiming__> #endmeeting