13:01:54 <Qiming> #startmeeting senlin
13:01:55 <openstack> Meeting started Tue Feb 23 13:01:54 2016 UTC and is due to finish in 60 minutes.  The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:56 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:58 <openstack> The meeting name has been set to 'senlin'
13:02:38 <yanyanhu> a little bit earlier
13:02:49 <elynn_> o/
13:02:53 <yanyanhu> hi
13:03:03 <Qiming> little bit late, you mean?
13:03:04 <lixinhui> 9:04 on my laptop :)
13:03:28 <yanyanhu> looks my local time need refresh :)
13:03:38 <Qiming> was trapped by the policy doc thing I'm working on
13:04:34 <Qiming> pls check agenda and see if you have things to add
13:05:02 <Qiming> first things first, let's run the work item etherpad
13:05:12 <Qiming> #link https://etherpad.openstack.org/p/senlin-mitaka-workitems
13:05:40 <Qiming> api side, I'm not aware of any high priority bugs/fixes
13:06:07 <Qiming> I'd suggest we move advanced 'filters' back to TODO list
13:06:32 <yanyanhu> agree
13:06:47 <Qiming> the reason is ...
13:06:51 <yanyanhu> it's a work with lower priority
13:06:59 <Qiming> it is very complicated ... see: http://specs.openstack.org/openstack/api-wg/guidelines/pagination_filter_sort.html#filtering
13:07:20 <Qiming> consider: GET /app/items?size=gt:8
13:07:29 <Qiming> GET /app/items?foo=in:buzz,bar
13:08:13 <Qiming> heat resource type
13:08:19 <elynn_> yes
13:08:33 <elynn_> only one resource left.
13:08:36 <Qiming> OS::Senlin::Node
13:09:16 <elynn_> I update it according to reviews
13:09:29 <Qiming> there is not serious problems as I see it in the code
13:09:34 <Qiming> just one thing
13:09:47 <Qiming> do we want to expose a node to be created with cluster specified?
13:10:02 <Qiming> the final answer to that would be YES
13:10:16 <Qiming> do we want to support that right now?
13:10:23 <Qiming> I'm not sure, to be honest
13:10:50 <Qiming> haven't got bandwidth to deal with the NODE_CREATE problem
13:11:10 <elynn_> What's the problem with node_create?
13:11:14 <Qiming> i.e. item listed on line 45
13:11:31 <Qiming> when you are creating a new node INTO a cluster
13:11:45 <Qiming> you will respect all policies attached to that cluster
13:11:47 <yanyanhu> elynn_, policy check I think
13:12:00 <Qiming> say, region placement, zone placement, load-balancing, ...
13:12:44 <Qiming> the same applies to NODE_DELETE
13:12:45 <elynn_> Do I need to hide the property in heat resource?
13:13:02 <Qiming> I'd suggest so
13:13:46 <Qiming> speaking of Heat resource types, I don't see a strong use case to create an out-of-band node into a cluster, via a template
13:14:07 <elynn_> So for now we don't support attach a node to cluster?
13:14:28 <Qiming> that is the question
13:14:38 <elynn_> Qiming: Yes, it's not that necessary to create a node from template.
13:14:52 <elynn_> Qiming: I mean in heat.
13:14:54 <Qiming> we do support cluster-add-nodes
13:15:12 <Qiming> we are checking that action in our policies
13:15:41 <Qiming> but we never checked NODE_CREATE, NODE_DELETE
13:16:14 <elynn_> ok... That would make codes of node resource more complicate.
13:16:28 <Qiming> so .. I'd suggest we hide 'cluster' property of nodes
13:16:53 <Qiming> at least for now
13:17:13 <elynn_> Qiming: ok, I will hide it in next patchset.
13:17:15 <Qiming> it can be left as an attribute
13:17:31 <Qiming> also, node details should be left as an attribute
13:18:43 <elynn_> ok
13:18:46 <Qiming> is there a problem if we have 'cluster' as an attribute of Node and in future we add it as a property?
13:20:08 <Qiming> a practical goal for mitaka is to have the basic cluster/node/policy/profile concepts demonstrated in a template
13:20:10 <elynn_> Qiming: I don't think so, since we can say that cluster attr is ID of cluster while cluster property could be name.
13:20:11 <Qiming> that is enough
13:20:34 <Qiming> fair point
13:20:43 <Qiming> moving on, testing
13:20:55 <Qiming> have we started stress testing?
13:21:00 <yanyanhu> no yet
13:21:05 <yanyanhu> just have some basic idea
13:21:29 <yanyanhu> may need further discussion with xinhui and brian
13:21:39 <Qiming> need to fork this work thread soon so that branw can work on it
13:21:46 <yanyanhu> yes
13:22:00 <yanyanhu> will try to setup a discussion with them this week
13:22:14 <Qiming> more functional tests for cluster/node coming in? or we can delete line 16?
13:22:24 <Qiming> cool, yanyanhu
13:22:28 <yanyanhu> still need to add a test case for cluster update
13:22:36 <Qiming> okay
13:22:38 <yanyanhu> other parts have been done I think
13:22:59 <Qiming> I'm removing the senlinclient unit test line
13:23:09 <yanyanhu> ok
13:23:24 <Qiming> health management/policy ?
13:23:38 <lixinhui> I have submit the patch
13:23:39 <lixinhui> https://review.openstack.org/#/c/282299/
13:24:01 <lixinhui> collect comments
13:24:15 <lixinhui> I will change accordingly
13:24:29 <Qiming> ah, I reviewed, but haven't written comments
13:24:37 <lixinhui> :(
13:24:45 <Qiming> the general feeling is that the patch is still a half-baked solution
13:24:54 <lixinhui> en
13:24:59 <Qiming> it is dangerous if we are moving along this way
13:25:27 <Qiming> it may induce a lot of overheads if we are checking every single cluster and try to recover them
13:26:01 <Qiming> I'd suggest we only care about clusters that are somehow 'registered' for this kind of checking/recovery
13:26:06 <lixinhui> a candidate way to check only the clusters attached with health policy
13:26:27 <Qiming> the best way I can think of is to see if they have a health policy attached AND enabled
13:26:42 <lixinhui> yes
13:27:05 <lixinhui> but not sure what priority with health policy
13:27:13 <Qiming> so a health policy, when being attached to a cluster, will "register" that cluster for health management
13:28:06 <lixinhui> okay
13:28:07 <Qiming> a health policy is triggered by timer events, not actions
13:28:21 <Qiming> so 'priority' may be not relevant
13:28:55 <Qiming> we can have a simple, naive version of health policy to drive this
13:29:23 <lixinhui> okay
13:29:33 <Qiming> cluster/node check/recover apis all done?
13:29:36 <lixinhui> could you give some example on timer events?
13:29:50 <lixinhui> yes, they all be done
13:30:09 <lixinhui> once the client patch got into, I will add doc accordingly
13:30:23 <Qiming> can be triggered by crond, for example, or timer event from ceilometer,
13:30:36 <lixinhui> and saw Yantanhu has helped to write the functional tests :)
13:30:45 <Qiming> can be invoked from within senlin engine, from a perioc_task
13:30:51 <yanyanhu> lixinhui, yes :)
13:30:56 <yanyanhu> those API are available now
13:31:13 <lixinhui> okay
13:31:19 <lixinhui> I will try
13:31:20 <Qiming> how about the SDK side, I think we have them all merged?
13:31:27 <lixinhui> Yes
13:31:28 <Qiming> great, thanks!!
13:31:49 <Qiming> moving on
13:31:52 <Qiming> documentation
13:31:53 <lixinhui> One thing I also tried in the past week
13:32:03 <Qiming> okay?
13:32:06 <lixinhui> is about neutron member health status
13:32:12 <Qiming> LB based?
13:32:18 <lixinhui> Yes
13:32:34 <lixinhui> I still do not give up to get status trigged recover
13:32:47 <Qiming> please don't
13:32:49 <Qiming> :)
13:32:53 <lixinhui> oh
13:33:19 <Qiming> I mean, please don't give up
13:33:20 <lixinhui> weekly meeting is very important :)
13:33:47 <lixinhui> but found a bug that member status can be updated
13:33:52 <lixinhui> by healthmonitor in V@
13:34:00 <lixinhui> inv2
13:34:05 <yanyanhu> lixinhui, hope can have some time to work on healthmonitor support in lb policy in coming week
13:34:09 <Qiming> can or cannot?
13:34:13 <lixinhui> s/can/cann'
13:34:18 <lixinhui> s/can/cann't
13:34:25 <Qiming> could be some config problem?
13:34:50 <yanyanhu> status of lb pool_member?
13:34:54 <lixinhui> great yanyanhu! I know your work efforts already very heavy
13:35:11 <lixinhui> I tried but think not config issue
13:35:19 <yanyanhu> lixinhui, no problem, just being trapped in some other stuff :)
13:35:26 <Qiming> okay, let's wait for feedback from neutron team
13:35:52 <lixinhui> okay, will not give up
13:36:18 <Qiming> cool
13:36:50 <Qiming> documentation side, I've started drafting something, hopefully they will serve a summary or design doc
13:36:54 <Qiming> http://docs-draft.openstack.org/58/283558/2/check/gate-senlin-docs/b75599a//doc/build/html/developer/index.html
13:37:11 <Qiming> you will see the Built-in Policy Types section
13:37:36 <Qiming> just commited a patch about deletion policy
13:37:49 <yanyanhu> great job!
13:37:52 <Qiming> link is here: http://docs-draft.openstack.org/58/283558/2/check/gate-senlin-docs/b75599a//doc/build/html/developer/policies/deletion_v1.html
13:38:40 <Qiming> writing docs really helps clear my mind
13:39:01 <Qiming> will continue working on that
13:39:04 <lixinhui> :)
13:39:11 <lixinhui> zan
13:39:24 <Qiming> I'm finding quite some bugs when writing the doc and reread the code
13:40:03 <lixinhui> will do that too
13:40:27 <Qiming> don't think we have any progress in other areas listed in the etherpad
13:40:54 <yanyanhu> yes, I'm afraid that's true...
13:41:06 <yanyanhu> oh, about the resource property renaming in sdk, have all related patches been merged yet?
13:41:12 <Qiming> we did something that are not listed on the page, :)
13:41:17 <yanyanhu> do we need to prepare rework our driver?
13:41:25 <Qiming> yanyanhu, not yet
13:41:29 <yanyanhu> Qiming, yep :)
13:41:38 <yanyanhu> ok
13:41:46 <Qiming> still some patches coming in
13:41:48 <yanyanhu> will wait for new release of sdk
13:42:51 <Qiming> Richard is working on message service
13:43:05 <yanyanhu> zaqar?
13:43:12 <lixinhui> when will it be possibly?
13:43:27 <Qiming> seems so
13:43:53 <Qiming> oh, I forgot that '-2' is sticky
13:44:09 <yanyanhu> which -2?
13:44:14 <Qiming> I was surprised to see Yanyan's patch still has a -2
13:44:35 <yanyanhu> ah
13:44:36 <yanyanhu> that one
13:44:46 <yanyanhu> I guess terry hasn't got time to watch it :)
13:45:06 <Qiming> okay, that is one of the SDK issues we need to deal with
13:45:10 <yanyanhu> we really need that workaround for pool_member resource creating
13:45:15 <yanyanhu> yes
13:45:25 <lixinhui> yes
13:45:33 <Qiming> are you aware of others?
13:46:10 <Qiming> I think I have heard some complaints about ksa1, things not found, etc etc
13:46:10 <yanyanhu> no, as of now, only pool_member has this problem I think
13:46:28 <yanyanhu> ksa1?
13:46:36 <Qiming> keystoneauth1
13:46:43 <yanyanhu> oh
13:46:59 <yanyanhu> really need spend some time to understand it
13:47:06 <Qiming> okay
13:47:14 <yanyanhu> as you know, I'm trying to install Senlin in SV
13:47:18 <yanyanhu> which is a Juno env
13:47:20 <Qiming> cool
13:47:29 <lixinhui> what is SV
13:47:31 <Qiming> that would be a very risky test, :)
13:47:32 <lixinhui> ?
13:47:38 <yanyanhu> do meet some issues when access keystone service through sdk
13:47:43 <yanyanhu> yes
13:47:43 <Qiming> lixinhui, a power cloud
13:47:53 <lixinhui> okay, Qiming
13:47:59 <Qiming> sdk 0.7.4 ?
13:48:00 <yanyanhu> especially about trust
13:48:01 <yanyanhu> yes
13:48:04 <yanyanhu> 0.7.4
13:48:10 <yanyanhu> authentication is ok
13:48:39 <Qiming> good luck, boy
13:48:41 <yanyanhu> but exception happened during listing trust
13:48:45 <yanyanhu> thanks :)
13:48:48 <yanyanhu> hope so
13:49:08 <Qiming> lixinhui, any outstanding issues building the autoscaling poc?
13:49:32 <lixinhui> so far good
13:49:52 <lixinhui> I will ask for help once meet issues
13:49:57 <Qiming> you can point me to the ceilometer patch about lbaas v2
13:50:01 <lixinhui> :)
13:50:10 <Qiming> I can help poke the ceilometer team on that
13:50:24 <lixinhui> https://blueprints.launchpad.net/ceilometer/+spec/lbaas-v2-enablement
13:50:44 <lixinhui> Qiming, the code works in my tests
13:50:55 <Qiming> you mean that patch?
13:50:57 <yanyanhu> cool
13:51:01 <lixinhui> yes
13:51:06 <yanyanhu> ji zilian works on it :)
13:51:16 <lixinhui> yes, yanyanhu :)
13:51:19 <Qiming> cool, the author has been working with us for quite a long time
13:51:21 <yanyanhu> really a good news for us
13:51:28 <lixinhui> great!!!
13:51:40 <lixinhui> RP is so good
13:51:46 <elynn_> oh, his BP
13:51:47 <yanyanhu> haha
13:52:05 <Qiming> feel free to speak up when there are questions
13:52:12 <yanyanhu> yep, he has worked on monitoring in openstack for a long time
13:52:22 <lixinhui> okay!
13:52:36 <Qiming> #topic open discussion
13:52:57 <yanyanhu> I think you guys may need to prepare for the trip to Austin :)
13:52:59 <Qiming> ^ will be the first/last topic in the meeting minutes
13:53:19 <yanyanhu> especially the visa
13:53:22 <Qiming> trapped by several things these days
13:53:37 <elynn_> Are you all going to Austin?
13:53:42 <Qiming> flight connection is a headache, according to Deng Hui
13:53:56 <yanyanhu> :)
13:54:01 <Qiming> elynn_, we follow you
13:54:03 <lixinhui> I did not reply to him
13:54:05 <yanyanhu> elynn_, myself is not sure
13:54:38 <lixinhui> because I am still thinking the trip plan
13:54:41 <Qiming> I'm reading ttx's proposal on separating design summits out of the main conf
13:54:44 <Qiming> sigh ...
13:54:48 <lixinhui> go to PA or not
13:55:16 <Qiming> if you are very active, it means you will need to justify four trips a year now
13:55:17 <yanyanhu> any decision about separating design summit?
13:55:54 <Qiming> don't know the outcome yet
13:56:33 <Qiming> I planned to repond to that proposal, but ...
13:56:38 <Qiming> bandwidth
13:56:54 <yanyanhu> right...
13:57:24 <Qiming> TBH, I don't get the idea of separating devs into a 'clean' room for pure tech discussions
13:57:59 <Qiming> somehow we are stepping away from users
13:58:01 <Qiming> sigh
13:58:26 <Qiming> anything else?
13:58:29 <Qiming> 2 mins left
13:58:35 <yanyanhu> nope from me
13:58:46 <elynn_> nope from me
13:58:47 <lixinhui> Israel openstack day
13:58:51 <Qiming> haiwei is in hospital I think
13:58:57 <yanyanhu> just hope can have more time to work on those items in etherpad
13:59:05 <lixinhui> do you guuys feel interested?
13:59:06 <yanyanhu> oh?
13:59:10 <Qiming> lixinhui, going Israel?
13:59:11 <yanyanhu> is he ok?
13:59:22 <Qiming> hope so, yanyanhu
13:59:23 <lixinhui> poor guy
13:59:32 <elynn_> What happend?
13:59:33 <lixinhui> yes will send link to all of you
13:59:33 <yanyanhu> lixinhui, when is it?
13:59:41 <yanyanhu> lixinhui, thanks
13:59:47 <lixinhui> to see if anything want to propose
13:59:49 <Qiming> lixinhui, al ways interested in any tech discussions
13:59:50 <lixinhui> NP
13:59:55 <Qiming> the only problem is travel budget
13:59:59 <lixinhui> :)
14:00:02 <yanyanhu> yes...
14:00:03 <lixinhui> come to vmware
14:00:05 <yanyanhu> big headache
14:00:08 <Qiming> enough
14:00:09 <Qiming> thanks
14:00:13 <yanyanhu> haha
14:00:13 <Qiming> #endmeeting