13:00:27 #startmeeting senlin 13:00:28 Meeting started Tue Jul 21 13:00:27 2015 UTC and is due to finish in 60 minutes. The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:29 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:29 hi 13:00:31 The meeting name has been set to 'senlin' 13:00:48 hello 13:00:49 let's get started? 13:00:54 hi 13:01:20 pls feel free to add agenda items: https://wiki.openstack.org/wiki/Meetings/SenlinAgenda 13:01:20 Qiming, your hello is a little earlier ;) 13:01:31 o/ 13:01:33 #topic add agenda items 13:01:38 hi lisa 13:02:33 #topic liberty-2 targets 13:02:46 #link https://etherpad.openstack.org/p/senlin-liberty-workitems 13:03:18 seems we have accomplished a lot, :) 13:03:28 yes, progress is good I think 13:03:34 if any update on binding.rst? 13:03:47 doc? 13:03:53 yes 13:04:34 latest unit test cover rate is 71% in my local env :) 13:04:36 here: https://review.openstack.org/#/c/201009/ 13:05:39 we still have some monsters to kill? 13:05:51 yes, about action and also some drivers 13:05:52 yes i am working on a few 13:05:57 I wanna try something about policy binding of node-placement policy 13:06:12 lixinhui_, that would be good 13:06:19 thanks, jruano 13:07:00 maybe we can target a placement policy implementation for liberty-3? 13:07:13 liberty-3 is 2015-09-01 I think 13:07:19 that will be cool 13:07:32 now we don't have any implementation about this policy 13:07:38 it doesn't have to be a full-featured version 13:07:46 yes 13:08:06 I think lixinhui_ has got some problems passing 'scheduler_hints' to nova 13:08:25 and sometimes this is intermingled with the server-group nova api 13:09:07 as far as I know, the nova server profile has something to be hardened 13:09:39 for example, the 'networks' property might be 'nics' 13:09:39 yes, seems openstacksdk will transfer all attrs into server: {} 13:09:50 we need to look into the nova-api to find out the details 13:10:04 but scheudler_hints should be os:scheudler_hints: {} 13:10:41 okay, it would be good if we can find the latest nova api doc 13:12:02 Qiming, maybe also the trust API doc 13:12:08 there are some other notions in sdk: server_interface, server_metadata and server_meta 13:12:57 yes, I noticed that the function calls related to server_metadata are still not there 13:12:58 yanyanhu, trust for api invocation? 13:13:35 yes 13:13:40 I have proposed a server_group resource to sdk, but their acceptance of new resource types is frozen for the moment 13:14:07 I don't think trust would be part of nova api 13:14:13 last time, you shared a link where the trust API description is a little different from the current one we are refering 13:14:18 yes, sure 13:14:19 it is completely an interaction with keystone 13:14:23 I mean keystone v3 API 13:14:59 ah, that wiki page, I checked the history, it was written in 2013, may be outdated long time ago 13:15:09 the only thing you can trust is the code itself 13:15:16 ok... 13:15:17 http://developer.openstack.org/api-ref.html 13:15:25 not sure whether this doc is updated 13:15:40 for some projects, I believe so 13:15:46 for others ... em 13:15:49 definitely no 13:16:04 but I found neutron related API doc is right 13:16:05 for example, you can check the ceilometer api there 13:16:09 ok 13:16:12 need test 13:16:30 right 13:16:38 report a bug 13:16:52 ? 13:17:57 http://paste.openstack.org/show/395684/ 13:18:05 poor network at home 13:19:08 lixinhui_, you will check the senlin-api output for this 13:19:37 okay 13:19:40 500 Internal Server Error means a real problem at api/engine side 13:20:06 look for this line "Failed checking service user checking." in senlin code 13:20:07 I found something similar like this 13:20:12 you will know what happened 13:20:37 please file a but for this 13:20:43 I think it could be a configuration problem 13:20:59 for example, if you are using 'demo' user in the 'demo' project 13:21:31 the user has to be configured as the 'admin' of the project, or else some keystone policies will block you from doing useful work 13:21:55 if that is the case, we need to update the setup document and the devstack plugin code 13:23:01 another improvement I see is the client code, the client can be improved to understand this kind of errors instead of reporting an 'ValueError' 13:24:06 so .. please file bugs when things don't work 13:24:52 anything else for the liberty-2 targets? 13:25:00 hi, Qiming, last time, you mentioned that maybe we can try to provide a first release 13:25:10 by the end of liberty-2 13:25:33 with just 1 week left? 13:25:47 em, a little difficult I think 13:26:17 but we need a plan I think ;) 13:26:37 it would be a good practice though I don't think the release itself will change the world 13:26:55 sure 13:27:12 #action Qiming to investigate release procedure 13:27:46 anyway, we will definitely do a release by liberty-3 13:27:54 yes 13:28:02 nice 13:28:49 let's think about it -- what do we want to go into the release 13:29:05 we can discuss it in our next meeting 13:29:16 em, especially what features will be included in the first release 13:29:17 ok 13:30:01 another thing I do not know if we should do in senlin is locale 13:30:22 lixinhui_, you mean internationalization? 13:30:26 yes 13:30:34 a light one ... 13:30:55 em ... I can read a little French, a little German ... 13:31:17 each time I wanna run senlin on my mac, I have to resolve local UTF-8 13:31:25 ... 13:31:35 haiwei_ can do some Japanese I think 13:32:02 UTF-8 is the standard encoding, especially for Python 3 13:32:14 I am using a English environment 13:32:26 :-) 13:32:37 we do have infrastructure to extract strings for translation 13:33:00 the community has a shared project doing that: i18n 13:33:49 so far we have been transmitting and storing data in UTF-8 I think 13:33:57 okay, let's move on 13:34:22 #topic update on complex (scaling) policy 13:34:32 this has been my focus recently 13:35:03 I haven't checked in any code yet, because nothing is in good shape yet 13:35:32 in this work, I'm trying to model a scaling policy in 3 sections: 13:35:44 signal, schedule, handlers 13:36:10 a signal can be a ceilometer alarm, a monasca alarm, a zaqar message, for example 13:36:45 but the policy will contain only a single signal entry, that is a restriction to simplify implementation 13:37:31 when modeling the signal section, I spent some time studying the ceilometer alarm API doc, which is in nowhere 13:37:55 there are threshold alarm, combination alarm and ... gnocchi alarms 13:38:31 properties vary between different alarm types 13:39:03 and .... I saw this command: 13:39:28 ceilometer alarm-gnocchi-aggregation-by-metrics-threshold-create 13:39:50 kind of driven crazy by this ... 13:39:54 a long command 13:40:31 and things like this: ceilometer alarm-gnocchi-aggregation-by-resources-threshold-create 13:40:38 sigh ... 13:41:13 any way, I will try my best to make the 'signal' specification applicable to most alarm types 13:41:26 the second 'section' is 'schedule' 13:42:00 this is designed for scheduled action triggering, it can be used with and without a signal section 13:42:13 knock knock 13:42:34 seems you once said ceilometer 13:42:40 if there is a signal and a schedule section, it means some action will be triggered during the 'schedule' window 13:42:42 will change the alar 13:43:01 yes, they have a plan to split alarm out of ceilometer 13:43:26 to where? 13:43:32 they are keeping ceilometer alarms for now until they make sure they don't break anything 13:43:42 if only 'schedule' section is provided, then the policy will take effect only during the target window 13:43:46 heat team almost had a heart attack 13:43:57 :-) 13:44:00 the switch is supposed to be transparent 13:44:03 yes, that is .... 13:44:30 I'm speechless over that patch 13:44:31 haiwei_, a new project named Aodh 13:44:47 difficult to read 13:45:13 so, we need to keep an eye on that, and hopefully understand their new api design soon 13:45:49 the last section in the complex scaling policy is about 'handlers', yes, plural form 13:46:11 it is a list of actions to be triggered when some 'alarm' or 'message' is received 13:46:43 the good news is that both Ceilometer and Monasca support a list of webhook urls 13:47:21 I have almost finished the parsing of the 'handler' section 13:47:40 maybe I should check in a WIP for you guys to comment 13:47:52 but don't +2 on it yet, cores 13:48:20 :) 13:48:35 will help to review the code one you submit it 13:48:46 thanks, that would be great 13:48:54 please as picky as you can be 13:50:22 any questions/comments so far? 13:50:30 i like the design approach. would be nice to investigate the code 13:50:52 okay, will check in the code tomorrow 13:51:07 hoho~ 13:51:16 #topic open discussions 13:52:04 the error lixinhui_ has reported is similar like this https://bugs.launchpad.net/python-senlinclient/+bug/1472557 13:52:05 Launchpad bug 1472557 in python-senlinclient "cluster, profile and node can't be updated" [Undecided,New] 13:52:05 Launchpad bug 1472557 in python-senlinclient "cluster, profile and node can't be updated" [Undecided,New] 13:52:05 one thing above my head is about the talk proposal 13:52:06 Launchpad bug 1472557 in python-senlinclient "cluster, profile and node can't be updated" [Undecided,New] https://launchpad.net/bugs/1472557 13:52:52 yes, need to get started on an outline and some initial content 13:53:09 haiwei_, will look at it later, thanks for reporting 13:53:27 and anyone has decided to join the summit? 13:53:36 we need a POC implementation of a container cluster 13:54:08 it won't be very difficult if you are using CoreOS guest image 13:54:35 I had some experiments before, though not a cluster environment 13:54:37 but we need a driver for this? 13:54:52 it is just a nova server 13:55:08 with some extra metadata 13:55:17 oh, you mean lxc supported by nova 13:55:21 understand 13:55:36 someone need to dive into magnum code to find out what else do they need from a VM to run containers 13:55:49 no, not lxc 13:56:25 magnum is building bays using nova servers (VMs) and then create containers inside the VMs 13:56:57 ah i see, so just some extra nova server data 13:57:12 yes, that could be a quick experiment 13:57:32 you need some k8s or docker packages in the image 13:57:37 i can take a look. when do you need it by? 13:57:55 with user_data and cloud-init? 13:58:04 jruano, the earlier the better 13:58:11 yes, yanyanhu 13:58:15 ok 13:58:27 ill see if i can get to it later today 13:58:33 thanks, jruano, ping me if you need some help on starting this 13:58:38 np 13:59:00 #action Julio to experiment with containers 13:59:08 okay, time's up 13:59:17 let's go to #senlin 13:59:23 thank you all 13:59:28 #endmeeting