13:00:30 <Qiming> #startmeeting senlin
13:00:30 <openstack> Meeting started Tue Jul 14 13:00:30 2015 UTC and is due to finish in 60 minutes.  The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:00:31 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:00:33 <openstack> The meeting name has been set to 'senlin'
13:00:39 <Qiming> hello, guys
13:00:47 <lixinhui> hello
13:00:57 <yanyanhu> hi
13:00:59 <lkarm> hey
13:01:09 <Qiming> hi, xinhui and lisa
13:01:50 <Qiming> hi, jruano
13:01:53 <jruano> hello
13:02:19 <Qiming> everyone please add agenda items to the wiki, if you can access it, :)
13:02:41 <Qiming> I was able to add some using a machine located in the states
13:03:25 <Qiming> first thing I'd like to update is ... IRC log
13:03:37 <Qiming> with the merge of this patch: https://review.openstack.org/#/c/196993/
13:03:43 <Qiming> now we have IRC logs, ;)
13:03:49 <Qiming> http://eavesdrop.openstack.org/irclogs/%23senlin/%23senlin.2015-07-14.log.html
13:04:23 <yanyanhu> cool :)
13:04:26 <Qiming> it's pretty nice
13:04:51 <jruano> very good
13:05:04 <lkarm> awesome it looks good
13:05:14 <Qiming> #topic liberty-2 milestone targets
13:05:27 <Qiming> #link https://etherpad.openstack.org/p/senlin-liberty-workitems
13:05:58 <Qiming> so far we have made goo progress on the work items
13:06:20 <yanyanhu> yes
13:06:29 <Qiming> yanyanhu has proposed some test cases for the policies
13:06:37 <Qiming> most of them are already merged
13:06:43 <jruano> i have a few more test cases that i will add
13:06:47 <jruano> working on those now
13:06:58 <Qiming> that would be great, jruano
13:07:21 <yanyanhu> I will try to work on test case for some drivers after finishing policy realted ones
13:07:29 <Qiming> driver test cases would be much easier considering the code is not very complex
13:08:29 <yanyanhu> so hopefully, we can finish most of the unit tests by the end of liberty2 cycle
13:08:37 <Qiming> when haiwei is back, hopefully we can get the first phase exception handling landed as well
13:09:10 <Qiming> yep, a lot of work, but I'm pretty sure we can make it, :)
13:09:28 <yanyanhu> yes :)
13:09:35 <Qiming> starting tomorrow, I am gonna focus on the scaling policy rework
13:09:46 <lixinhui> cool
13:10:00 <Qiming> hopefully, we can get something for review by next time we meet
13:10:30 <Qiming> when we were discussing the action triggering mechanism with the heat team
13:10:48 <Qiming> there were suggestions to turn to zaqar based notification
13:11:00 <Qiming> I don't think we can make it by liberty-2
13:11:10 <Qiming> we can postpone that to liberty-3
13:11:31 <Qiming> for liberty-2, we will make webhook based triggering work
13:12:15 <jruano> yes, zaqar support would seem like a big effort
13:12:40 <Qiming> my first experience with it was .... not a pleasant one, to be honest
13:13:16 <Qiming> I tried their test case by following the instructions in README, it didn't work, there were some multi-processing problem I think
13:13:52 <Qiming> the advantage of having zaqar support is that it allows the trigger be sent from anywhere ...
13:14:51 <Qiming> #topic documentation - getting started doc
13:15:28 <Qiming> now we have the getting started doc covering most frequent operations
13:16:07 <Qiming> we don't have docs for events, actions and webhooks, I'm leaving that for future work because we may revise the webhook flow somehow
13:16:34 <Qiming> for events and actions, we only support listing and show operations, pretty straightforward
13:17:04 <lixinhui> do we have guidence on policy side?
13:17:13 <Qiming> may jump back on to this when I'm stuck by the scaling policy work
13:17:18 <Qiming> yes, lixinhui
13:17:59 <lixinhui> good to know, qiming, I will try more on policies in next days
13:18:04 <Qiming> lixinhui, doc on policy types: https://review.openstack.org/#/c/200596/
13:18:18 <Qiming> doc on policy objects: https://review.openstack.org/#/c/200801/
13:18:31 <lixinhui> okay
13:18:32 <Qiming> doc on policy bindings: https://review.openstack.org/#/c/201009/
13:19:16 <lixinhui> great!
13:19:18 <Qiming> please feel free to -1 when seeing something wrong, unclear ...
13:19:45 <lixinhui> :-)
13:20:06 <Qiming> #topic update on node placement experiments
13:20:13 <Qiming> lixinhui, this is yours
13:20:16 <yanyanhu> \]\\
13:20:22 <yanyanhu> sorry
13:20:28 <lixinhui> okay
13:20:57 <lixinhui> I am running senlin on vsphere nova driver
13:21:27 <lixinhui> and target to build some demo on smart auto-scaling on vsphere vm placement
13:22:33 <lixinhui> first step of this is to enable hint in senlin
13:22:39 <lixinhui> scheduler_hints
13:22:47 <Qiming> so the emphasis is on scaling or placement? we are currently treating them as two different policy types
13:23:01 <Qiming> that is about placement, right?
13:23:07 <lixinhui> yes
13:23:10 <yanyanhu> em, we have two different polices for them respectively
13:23:33 <lixinhui> but I think the two types can work together?
13:23:38 <lixinhui> right?
13:23:51 <Qiming> cool, we only have a sketch for the placement policy at the moment
13:24:12 <lixinhui> :-)
13:24:15 <Qiming> yes, it was designed to be work separately and cowork together
13:24:47 <Qiming> even for a non-scaling cluster, you may still want to control node placement
13:25:00 <yanyanhu> so nova vsphere driver can understand the scheduler_hint from Nova and use it to guide the placement of virtual machine?
13:25:26 <lixinhui> yes, qiming, we meed discusss more about the design
13:25:42 <lixinhui> right, yanyan
13:26:27 <Qiming> okay, it would be great if you can check in your current work, then we can use it as the starting point
13:26:33 <lixinhui> for demo purpose, we can use server_group to simulate but in future vsphere team should expose more functions from driver to make this work beter
13:26:38 <yanyanhu> right
13:27:07 <Qiming> is vsphere using the server_group api from nova?
13:27:08 <lixinhui> Will do it, qiming
13:27:49 <lixinhui> it can work, qiming, I tried it by nova cli
13:27:56 <Qiming> great
13:28:14 <yanyanhu> this will be very good use case for the preliminary version of placement policy
13:28:35 <lixinhui> Yes, Yanyan:_)
13:28:37 <Qiming> by the way, Carolyn from IBM has proposed something about cross-az, cross-region placement support in Heat
13:28:46 <lixinhui> oh
13:28:53 <Qiming> we may need to support those use cases as well
13:29:04 <yanyanhu> yes
13:29:27 <lixinhui> how for heat to support the two placement?
13:29:34 <Qiming> spec here: https://review.openstack.org/#/c/179250/
13:29:39 <lixinhui> what is design difference
13:29:43 <Qiming> code here: https://review.openstack.org/#/c/116139/
13:30:27 <Qiming> it is an important feature for managing a group of resources
13:30:49 <Qiming> but implementing it in Heat is a little bit cumbersome
13:31:09 <Qiming> for example, the resource being scaled could be anything
13:31:42 <Qiming> we can limit the PROFILE_TYPE to 'os.nova.server' in senlin, and see if it works as expected
13:32:04 <jruano> yes... this request feels like something that should exist in senlin. separate from heat
13:32:41 <Qiming> right. if we can support it well, when we add senlin resource type back to Heat, users still can enjoy this feature
13:32:51 <jruano> exactly
13:33:11 <lixinhui> yes
13:33:15 <Qiming> the problem as I see it, is the policy itself
13:33:39 <Qiming> you can base your placement decision on a thousand different factors ...
13:34:19 <Qiming> loading balancing, energy efficiency, fault isolation, resource cost .... to name a few
13:34:45 <yanyanhu> so maybe like our lb_policy design, we need to narrow down our scope in the first step
13:35:18 <Qiming> yes, for "built-in" policy types, we can just support one or two scenarios
13:35:27 <Qiming> users know best what they want
13:35:37 <Qiming> they can provide their own policy plugins
13:35:39 <jruano> need some sort of importance ranking, or prioritization attribute
13:35:47 <jruano> user defined
13:36:19 <Qiming> yes, relative priority among policies of the same type, policies of different types
13:36:57 <Qiming> that would be a huge design space, we cannot dictate the usage scenario
13:37:12 <yanyanhu> agree
13:37:35 <Qiming> so, back to your experiment, lixinhui
13:37:46 <lixinhui> :-)
13:37:48 <Qiming> it sounds pretty important to Senlin
13:38:09 <lixinhui> yes, firstly, we have to enable scheluler_hints
13:38:09 <Qiming> please let us know if you need more hands and/or eyes on that
13:38:17 <lixinhui> great!
13:38:46 <Qiming> okay, let's move on
13:38:51 <lixinhui> I definitely will need help from all of you
13:39:00 <Qiming> #topic update on interlock with TOSCA team
13:39:21 <Qiming> jruano helped bridge us to the TOSCA team last week
13:39:43 <Qiming> we had some very interesting discussions around the policy and policy type support
13:39:46 <jruano> yes so a big use case tosca is looking at right now is node updates
13:40:10 <Qiming> both from a standard standingpoint and from senlin's implementation perspective
13:40:26 <Qiming> it was very timely a discussion
13:41:10 <jruano> very much... my colleague matt had some discussions with customers last week, and they are all very interested in senlin
13:41:35 <Qiming> that sounds nice
13:41:42 <jruano> policies are very important to them in terms of cloud management
13:42:49 <jruano> so a first step may be proving out a node update modeled in tosca, with senlin as the engine
13:43:56 <jruano> i will keep this moving forward and report updates back to the team
13:44:10 <Qiming> thanks, jruano
13:44:20 <Qiming> we will keep an eye on the tosca standard, be conformant to it when possible
13:44:59 <Qiming> jruano, you will set up a regular call for us with the TOSCA team, right?
13:45:10 <jruano> i will do that
13:45:18 <Qiming> cool
13:45:20 <jruano> a checkpoint call
13:45:50 <Qiming> okay, it would be great if we can read some WIP documents they have
13:46:47 <jruano> sure i can send out a link to the team
13:47:07 <Qiming> great
13:47:16 <Qiming> #topic open discussions
13:47:35 <lixinhui> Jruano, could you help to include me inside the maillist? lxinhui@vmware.com
13:48:16 <jruano> will do lixinhui
13:48:32 <lixinhui> Thanks.
13:49:06 <yanyanhu> hi, I think maybe we need to further cleanse the setup progress of Senlin service to make its installation easy
13:49:28 <Qiming> guys, if you haven't "watch" senlin and python-senlinclient projects in your gerrit setting, please do so
13:49:35 <yanyanhu> especially by using devstack
13:49:54 <Qiming> TBH, I haven't tried that yet
13:50:01 <lixinhui> how?
13:50:07 <lixinhui> by devstack
13:50:09 <Qiming> the devstack plugin was contributed by elynn
13:50:38 <yanyanhu> yes, I think most of us meet some problems when first time setup senlin  ;)
13:50:53 <Qiming> http://git.openstack.org/cgit/stackforge/senlin/tree/devstack/README.rst
13:51:11 <yanyanhu> maybe this can be a work item targets on libert-3
13:51:14 <lixinhui> yes, I know, I works well
13:51:28 <lixinhui> It works well
13:51:37 <Qiming> lixinhui, you have tried it?
13:51:40 <Qiming> \o/
13:51:41 <lixinhui> yes
13:51:57 <yanyanhu> I also haven't tested it yet...
13:52:10 <yanyanhu> cool
13:52:11 <jruano> oh it works great for me
13:52:37 <yanyanhu> this is great :)
13:52:45 <Qiming> cool, I had a link to the README.rst file in the getting started doc: http://git.openstack.org/cgit/stackforge/senlin/tree/doc/source/getting_started/install.rst#n25
13:52:48 <yanyanhu> I will also try it later
13:53:36 <Qiming> any other outstanding issues?
13:54:08 <Qiming> bugs?
13:55:11 <yanyanhu> not yet. But I guess we might meet some after we start testing cluster scaling/updating with scaling/lb policies attched
13:55:32 <Qiming> okay, future bugs for future meetings
13:55:41 <yanyanhu> sure :)
13:55:47 <Qiming> I think we can release this channel now
13:55:53 <Qiming> thanks for participating
13:55:59 <Qiming> #endmeeting