13:02:49 <Qiming> #startmeeting senlin
13:02:50 <openstack> Meeting started Tue Aug 11 13:02:49 2015 UTC and is due to finish in 60 minutes.  The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:02:51 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:02:53 <openstack> The meeting name has been set to 'senlin'
13:02:57 <Qiming> hello
13:03:00 <yanyanhu> hi
13:03:04 <lkarm> hi
13:03:31 <Qiming> pls add agenda here if you have topics: https://wiki.openstack.org/wiki/Meetings/SenlinAgenda
13:03:58 <Qiming> just the 3 of us?
13:04:16 <yanyanhu> I think haiwei and linxinhui are also here?
13:05:12 <yanyanhu> they are still in senlin channel I guess
13:05:32 <Qiming> okay, let's go ahead
13:05:49 <Qiming> they can check the minutes if cannot join at the moment
13:06:11 <Qiming> #topic liberty-3 milestone targets
13:06:17 <Qiming> hi, heyongli
13:06:33 <Qiming> https://etherpad.openstack.org/p/senlin-liberty-workitems
13:07:01 <haiwei> sorry, a little late
13:07:07 <Qiming> we still have some backlog to handle, but most of them are related to unit tests
13:07:09 <yanyanhu> :)
13:07:20 <Qiming> don't worry, we just got started
13:07:44 <Qiming> as for the unit test coverage, the latest number is 78%
13:07:54 <Qiming> there are still some under review
13:08:33 <Qiming> another work is about triggers, I'll jump on to that in the following days
13:09:00 <yanyanhu> I think unit test of senlin service won't be a big problem
13:09:24 <Qiming> for liberty-3 goals, I'm listing them as separate topics for discussion today
13:09:29 <yanyanhu> I believe we can achieve our goal before deadline
13:09:45 <Qiming> that would be good
13:09:51 <haiwei> yap
13:10:11 <yanyanhu> what I'm a little worried about is the unit test of client
13:10:22 <Qiming> before moving on to other topics, haiwei, do you have some updates on exception handling?
13:10:33 <yanyanhu> it seems to be a big workitem
13:10:42 <haiwei> I have done that a little, recently
13:10:54 <Qiming> right, saw your patches
13:11:10 <Qiming> but it won't be an one night job as far as I can see
13:11:34 <Qiming> we will continue work on improving things along the way
13:11:36 <haiwei> what's difficult I think is how to catch the internalError in upper layer
13:11:58 <Qiming> yep, that would be a case by case analysis and handling
13:12:17 <haiwei> I also noticed one thing that senlin make all the ignore_missing as True
13:12:37 <Qiming> haiwei, just posted some comments on that
13:12:45 <haiwei> that means the driver will return None, if the resource is not found
13:12:58 <haiwei> ok, I will check that later
13:13:08 <Qiming> defaulting ignore_missing to False would be the right thing, except for those xyz_delete() methods
13:13:25 <Qiming> if you want to delete a resource, and the resource is not found, it should be okay
13:13:35 <Qiming> other than that, it should be treated separately
13:13:58 <haiwei> I will think about it
13:14:05 <Qiming> okay
13:14:23 <Qiming> we haven't started the client unit tests work
13:14:34 <Qiming> should start that asap
13:14:38 <yanyanhu> that's great. I saw the patch of start point
13:14:58 <Qiming> there are quite something outdated in the interaction with SDK
13:15:26 <yanyanhu> yes, especially we are still using resource call I think
13:15:29 <Qiming> as we add more resources to SDK, the client code will be change a lot
13:15:30 <yanyanhu> in client side
13:15:48 <yanyanhu> yes
13:16:06 <Qiming> right, that may render the test cases for them a waste of time --- they would have to be rewritten
13:16:20 <Qiming> will think about it and work out a framework
13:16:38 <yanyanhu> if you and haiwei dont have enough bandwidth, I will also work on it after finish functional test related jobs
13:16:50 <yanyanhu> agree
13:16:52 <Qiming> when we are filling the holes for test case coverages, please help review
13:16:59 <yanyanhu> sure
13:17:28 <Qiming> sometimes, I cannot find which one has been merged, sometimes, have to rebase a lot things if patches are not reviewed in time
13:17:43 <Qiming> let's move on
13:17:51 <Qiming> #topic functional testing updates
13:18:09 <Qiming> so we have functional testing gate job set up
13:18:26 <yanyanhu> yes, we are in good progress I think :)
13:18:33 <yanyanhu> https://review.openstack.org/#/c/209736/
13:18:37 <Qiming> I' wondering why it is not triggered every time when patches drop in?
13:18:40 <yanyanhu> https://review.openstack.org/#/c/208355/
13:18:52 <yanyanhu> since it is experimental now
13:19:13 <yanyanhu> so if we want gate to help run functional test, we need manually trigger it by adding comments like 'check experimental'
13:19:26 <Qiming> I understand it has been set non-voting
13:19:31 <Qiming> okay, I see
13:19:41 <yanyanhu> this is a recommended way to add functional test when we can't ensure it can work correctly
13:19:46 <Qiming> thanks for clarification
13:19:57 <yanyanhu> no problem :)
13:20:00 <Qiming> is it working as expected so far?
13:20:07 <Qiming> for some simple tests?
13:20:08 <yanyanhu> after it can work correctly, we will move it to check pipeline
13:20:31 <yanyanhu> seems not...
13:20:45 <Qiming> pls help diagnose then
13:21:04 <Qiming> it is really important for a stable service
13:21:06 <yanyanhu> I think there could be some problems when gate setup the basic test environment which includes senlin service
13:21:10 <yanyanhu> sure
13:21:12 <Qiming> we are not creating just a toy
13:21:12 <yanyanhu> will do that
13:21:20 <yanyanhu> right
13:21:24 <Qiming> cool
13:21:51 <Qiming> let's move on
13:21:57 <Qiming> #topic clusters of containers
13:22:33 <Qiming> this is something we plan to show case during the coming summit
13:22:50 <Qiming> Julio has been helping doing some investigations
13:23:12 <Qiming> and some students from Tongji univ have been working on that since early this month
13:23:26 <Qiming> the progress is amazing, I can say
13:23:57 <Qiming> the students were able to create a cluster of Docker containers in three days
13:24:07 <lkarm> can we see a demo? or are there instructions somewhere to test it ourselves
13:24:23 <Qiming> they were not that familar with any OpenStack code before
13:24:41 <haiwei> they use magnum to do that?
13:24:43 <Qiming> they gave us a demo via webconference
13:25:09 <Qiming> not yet, they created their own python program to call Senlin APIs
13:25:33 <lixinhui_> amaizing
13:25:33 <Qiming> create profile, prepare cloud-config scripts, create VM clusters, start dockers inside VMS
13:26:02 <haiwei> great !!!
13:26:03 <yanyanhu> yep, I think those guys are excellent
13:26:04 <Qiming> I was shocked ...
13:26:40 <Qiming> so, after some discussions with the team yesterday, we have assigned some more tasks to them
13:26:41 <lixinhui_> It is a interesting cooperation between senlin and container
13:26:55 <Qiming> play with load-balancing policy and auto-scaling policy
13:27:19 <Qiming> stand up some meaningful apps from inside containers
13:27:30 <Qiming> the chosen app would be wordpress
13:28:28 <Qiming> if things are all going well, we may coach them on doing auto-scaling based on container metrics next week
13:29:04 <Qiming> lkarm, I'll ask the team to record something that can be shared
13:29:38 <lkarm> that would be great! would love to see it
13:29:50 <Qiming> definitely
13:30:46 <Qiming> if there are thoughts on improving/polishing the demo case, please share with the team
13:31:35 <Qiming> I've already asked the team to try use Senlin to create Docker bay in Magnum
13:31:47 <Qiming> would be curious to see if that is possible
13:32:21 <Qiming> any suggestions on this work?
13:33:06 <Qiming> okay, let's move on
13:33:09 <yanyanhu> do we need to revise the code of magnum?
13:33:10 <Qiming> #topic open discussions
13:33:29 <Qiming> yanyanhu, yes, that is inevitable
13:33:44 <Qiming> team will create a branch for experiments
13:34:01 <yanyanhu> ok, so if it is something like plugin or driver, that would be nice
13:34:31 <Qiming> I'm planning to collect their work here: https://github.com/tengqm/conclusur
13:34:58 <Qiming> that would depend on Magnum's design
13:35:15 <Qiming> conclusur = CONtainer CLUster SUR
13:35:39 <haiwei> not to senlin directory directly?
13:35:56 <Qiming> no, it would be a work for magnum
13:36:25 <yanyanhu> nice, hope they can update their work soon, then other guys can help to review and give some comments
13:36:28 <Qiming> there would be some sample profiles that can be checked into senlin project
13:36:39 <Qiming> yanyanhu, that's the plan
13:37:10 <Qiming> oh, forgot to add an item about the HA (high-availability) work
13:37:12 <lixinhui_> I am also interested at auto-scaling + container demo
13:37:25 <lixinhui_> let me know the demo when it is ready
13:37:35 <Qiming> lixinhui_, you are a hero, you can do everything, :)
13:37:39 <Qiming> sure
13:37:42 <lixinhui_> :)
13:37:51 <haiwei> a little confused, their jobs will be invited to senlin finally, right?
13:38:14 <Qiming> haiwei, it will be a show case of Senlin
13:38:22 <Qiming> but the code should go to Magnum
13:38:36 <Qiming> if is an application from Senlin's perspective
13:39:11 <Qiming> there would be something we can harvest in the end
13:39:24 <haiwei> got it, it's just like a use case how magnum uses senlin?
13:39:31 <Qiming> right
13:39:51 <Qiming> that is already a very strong argument how useful senlin could be
13:40:15 <Qiming> okay, back to the HA discussion
13:40:31 <Qiming> lixinhui and I visited China Mobile this afternoon
13:40:48 <Qiming> for a discussion on the coming NFV summit presentation proposal
13:41:32 <Qiming> we will try to demonstrate to those telecom vendors how Senlin could help achieve service resiliency
13:42:00 <Qiming> the conclusion is that we will show three things in the presentation
13:42:45 <Qiming> 1. placement policy 2. HA for stateful apps (VNFs) 3. HA for stateless apps (VNFs)
13:43:32 <Qiming> the good news is that lixinhui has already prototyped an integration of Senlin with VMware DRS service
13:44:13 <Qiming> though it is only about placement policy (something we can still improve), it is a very good starting point already
13:44:52 <Qiming> the OpenNFV summit is scheduled at early November
13:45:11 <Qiming> hopefully, we can have all three scenarios supported by then
13:45:48 <Qiming> this would be another strong evidence how Senlin could enable different usage scenarios
13:45:52 <lixinhui_> Yes, will make efforts there
13:46:51 <lixinhui_> Not just NFV, High availability is super important
13:47:07 <lixinhui_> even to VMware openstack product VIO
13:47:23 <Qiming> exactly
13:47:41 <lixinhui_> but I think senlin provide a super good/flexible structure/framwork
13:47:50 <lixinhui_> to resolve end2end HA
13:48:41 <Qiming> glad to hear that
13:48:53 <yanyanhu> yes, agree. actually besides HA policy, placement policy is a place to provide HA ability as well
13:49:19 <yanyanhu> although maybe in different layer
13:49:43 <lixinhui_> Yes, Yanyan. So we can cooperate together to show this capability.
13:49:59 <yanyanhu> definitely :)
13:50:31 <Qiming> anything else?
13:50:52 <yanyanhu> I want to talk a little bit about driver plugin workitem
13:51:48 <yanyanhu> since we have started this work and have finished nova_v2 related support, maybe we can start use this way to refer to nova driver :)
13:51:52 <yanyanhu> https://review.openstack.org/#/c/211058/
13:52:00 <Qiming> btw, in case you are interested in how VMware DRS is integrated with Senlin, please check it here: https://vmware.webex.com/vmware/lsr.php?RCID=76dafb3818d14ed69a99f49fdcf1a4ef
13:52:37 <yanyanhu> sure, we are absolutely interested in it :)
13:52:44 <Qiming> yanyanhu, it all looks good so far
13:53:04 <Qiming> basically this enables Senlin to talk to different clouds
13:53:33 <yanyanhu> yes, that's our purpose to do this job
13:53:49 <Qiming> I'd suggest we make the 'SenlinDriver' a global var lazily initialized
13:53:50 <yanyanhu> we hope to provide more flexibility using this way to manage drivers
13:54:17 <yanyanhu> ok
13:54:28 <yanyanhu> agree with this
13:54:30 <Qiming> but considering that lazy loading may bring some new problems
13:54:35 <Qiming> we can postpone that later
13:54:45 <yanyanhu> and we need to refresh this variable each time reload engine
13:54:49 <yanyanhu> ok
13:55:33 <yanyanhu> I think maybe I can give an introduction with more details about this job next week
13:55:53 <yanyanhu> to let everybody know more about it
13:56:11 <Qiming> if we have a global var, we will have cloud.compute(...) instead of driver_base.SenlinDriver().compute(params)
13:56:22 <yanyanhu> yes
13:56:41 <Qiming> sure, please add an item on the agenda
13:57:00 <yanyanhu> ok
13:57:13 <Qiming> 3 minutes left
13:57:49 <Qiming> are we done?
13:57:57 <yanyanhu> no more topics from me
13:58:03 <Qiming> 2 mins
13:58:22 <Qiming> seems we are done
13:58:27 <haiwei> ok
13:58:29 <Qiming> thanks for joining, guys
13:58:37 <Qiming> talk to you next week
13:58:40 <Qiming> #endmeeting