13:00:42 #startmeeting senlin 13:00:43 Meeting started Tue Mar 14 13:00:42 2017 UTC and is due to finish in 60 minutes. The chair is yanyanhu_. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:44 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:47 The meeting name has been set to 'senlin' 13:00:48 Qiming: Error: Can't start another meeting, one is in progress. Use #endmeeting first. 13:00:59 oops 13:01:16 oops 13:01:32 :) 13:01:32 thanks yanyan 13:01:33 hi, Qiming, sorry a little bit later. the network is unstable, have to use vpn to connect 13:01:39 Qiming, please go ahead : ) 13:01:51 alright 13:01:52 my pleasure :) 13:01:55 meeting agenda posted 13:02:04 #link https://wiki.openstack.org/wiki/Meetings/SenlinAgenda#Weekly_Senlin_.28Clustering.29_meeting 13:02:32 basically the things on the list is about work items update, pike goal tunings, summit pre 13:02:45 pls add items if any 13:03:23 let's get started 13:03:37 #topic pike work items 13:03:40 #link https://etherpad.openstack.org/p/senlin-pike-workitems 13:04:21 api - profile_only 13:04:43 444100 is in now 13:05:10 API test just added: https://review.openstack.org/#/c/445298/ 13:05:24 found a serious bug in version negotiation 13:05:42 need more eyes on it 13:06:02 there could be similar bugs in other request objects 13:06:13 which problme? 13:06:22 problem 13:06:32 see the about patch: https://review.openstack.org/#/c/445298/ 13:06:56 ok 13:07:54 the obj_make_compatible is not deleting the primitive keys properly 13:08:07 if 'profile_only' in primitive['senlin_object.data']: 13:08:07 del primitive['senlin_object.data']['profile_only'] 13:08:10 right 13:08:23 moving on to next topic? 13:08:29 move 13:08:32 feature rich nova server 13:08:53 I guess current patches are in now 13:09:02 both 438891 and 432587 are in 13:09:14 got merged this afternoon 13:09:32 let's see if there are more patches coming in 13:09:33 Need some tuning and address Qiming's -1 :) 13:09:46 I think it's done for now 13:09:51 you can leave those to your backport work 13:09:55 And I plan to move it to nova.server 13:10:07 piece by piece 13:10:09 I believe the top priority is to make it work 13:10:16 +1 13:10:30 Actually xinhui did a demo internal 13:10:32 and make sure the summit sessions are well prepared 13:10:39 successfully 13:11:24 Use this vdu profile together with monitor from vmware, to create a NFV scenario . 13:11:33 nice! 13:11:55 moving on 13:12:03 engine improvement -- cluster check 13:12:17 no update this week 13:12:21 are we working on self-destroyable actions? 13:12:28 okay 13:12:32 node adopt 13:12:38 prepare for the meetup 2017.3.18 13:12:46 proposed a patch: 13:13:04 https://review.openstack.org/444817 13:13:24 great 13:13:30 this is what we are expecting 13:13:32 there are things very tricky 13:13:44 personality 13:13:55 is an attribute you cannot get back from a server 13:14:21 networks, we cannot get the original request if it was created from port ids 13:14:38 and, port id, even extracted, doesn't seem that useful 13:14:58 with elynn's suggestion, I'm only extracting network name 13:15:27 Yes 13:15:32 the worst thing is about user_data 13:15:38 we really need them back 13:15:46 I always thought we can not get user_data... 13:15:53 nova api microversion 2.3 added this support 13:15:57 Since I didn't find a rest API for it... 13:16:18 however, the attribute is renamed to OS-EXT-SRV-ATTR:user_data ... 13:16:18 okay, good to know 13:16:29 yes, user_data is important for further scaling operation 13:16:30 and it is limited to admin users, as of now 13:16:45 talked to nova-api group, they are onto it. 13:17:08 they won't fix the name issue in near future, but they will fix the privilege restriction 13:17:22 sounds good 13:17:45 from sdk side, brian came up with a hack: https://review.openstack.org/#/c/444933/ 13:17:56 it is not an easy choice to make 13:18:08 things are always complicated when we are talking about nova 13:18:44 until that is released, we cannot get user_data back 13:19:06 another todo is about creating snapshot when adopting a server node 13:19:18 this is made optional 13:19:38 will need to work on profile level and then engine -> api -> sdk -> client path 13:19:47 moving on 13:19:58 RDO 13:20:01 ok 13:20:20 sorry, this week rdo is also no update 13:20:54 okay, pls keep pushing 13:20:59 next, health management 13:21:05 Will 13:21:17 xinhui is not in? 13:21:27 seems so 13:21:39 she has been back in last week 13:21:52 alright, we have got a talk about enterprise monitoring + senlin + mistral 13:22:04 that story has to be made ready 13:22:09 yes, there is a presentation on it 13:22:43 maybe we should move fencing support back to TODO list? 13:22:49 it is not a trivial job 13:22:59 maybe another meet in vmware :P 13:22:59 for the summit preparation 13:23:19 em, xinhui called me the other day 13:23:34 talking about having senlin make placement decisions 13:23:48 because we are already there, with quite a few policies 13:24:04 agree 13:24:04 placement policy? 13:24:15 her question was about how to make it work with the new placement api from nova 13:24:35 I don't have a vision on that either 13:24:42 Xinhui is busy recent days and she is sick.. 13:24:58 I'm not sure how the placement API is positioned, where it is heading 13:25:05 so nova has new api for advanced placement operations? 13:25:20 if so, that could be helpful 13:25:30 is it very ambitious to cover all aspects of placement decision? including network and storage factors? 13:25:44 sure 13:25:47 that is ambitious 13:26:07 or is it only about host resources and many also other multi-az, multi-region scheduling? 13:26:08 especially for network 13:26:08 yanyanhu_, yes, nova has a placement API in nova codes now. 13:26:30 it is mandatory if an devstack installation now 13:26:56 anyway, let's see what will happen 13:26:57 I guess recent days she will get more details of this requirement, and later we could figure out how we fit in this requirement. 13:27:50 at the same time, I really don't want to waste time duplicating features from another project 13:28:09 okay, so we can stick to our own agenda 13:28:54 yes, we won't reinvent the wheels. 13:28:58 we have less than one month to release p-1 13:29:06 ... 13:29:08 +1 13:29:15 Yes 13:29:29 Just need to fill the gap. 13:29:34 we hope we can may operators life easier, not more painful 13:29:53 #link https://releases.openstack.org/pike/schedule.html 13:30:11 moving on? 13:30:15 ok 13:30:17 okay 13:30:26 I guess no progress on client functional tests? 13:30:38 though the gate job (non-voting) is ready 13:30:47 Yes:) 13:30:59 okay 13:31:15 policy improvement, wrt LB policy and CLUSTER_RECOVER 13:31:31 em, this can be removed .. 13:31:38 done? 13:31:57 or ... still some minor improvements needed? 13:32:03 may need to improve the data 13:32:06 After meetup will do all things hard 13:32:09 :) 13:32:37 okay, thx 13:32:39 yes Qiming, still some improvements needed, will look forword to it 13:33:01 leaving that item there for a while 13:33:18 okay 13:33:22 NFV support is on the wishlist 13:33:41 got no news from Haiwei, although we have a talk on that 13:33:59 Xinhui's live demo is for that session, elynn ? 13:34:21 For haiwei's session? 13:34:37 No, xinhui's demo is for internal review. 13:34:39 haiwei is busy, and he said he will do the patch in tacker 13:35:00 But could do some modify and be prepare to other presentation. 13:35:04 okay 13:35:17 no matter what we are talking about 13:35:18 And then we prepare for the session 13:35:38 it has to be in the master, for users to try out 13:36:00 Cause it's not purely openstack, including some vmware software. But definitely for NFV. 13:36:32 okay, then the openstack part has to be ready 13:36:55 and possibly an alternative to vmware software as a generic solution would be great 13:37:17 there is nothing wrong to present vendor specific solution 13:37:27 Yes, that will be great 13:37:27 :) 13:37:37 but a generic one would be more useful to most audience 13:37:46 We can find a day to let her show us again. 13:37:53 cool. 13:38:10 Until she is not busy and recover from cough . 13:38:12 anything else on pike work items, the etherpad? 13:38:29 write a article for senlin[XueFeng Finished 13:38:30 I'm suffering too from coughing lately 13:38:32 great job from xuefeng 13:38:58 +1 13:38:59 would be nice if we have an English version, :) 13:39:00 Qiming, take a good rest 13:39:24 cannot reset, man 13:39:32 understand... 13:39:39 really busy recently 13:39:47 So we need a English version? 13:39:50 got some instructions to share your workload, just this afternoon 13:40:09 Qiming, yes, I also got the call... 13:40:14 need more discussion on it tomorrow 13:40:17 ok 13:40:22 you know, about the handover 13:40:28 okay 13:40:28 which will come soon 13:40:30 moving on 13:40:42 handover what? 13:40:46 #topic pike goals 13:40:53 QiMing 13:40:57 elynn, you will get it soon :P 13:41:06 senlin? 13:41:28 we talked about this very briefly last week 13:41:41 ironic container adoption health improve ;do some use cases 13:41:49 okay...bless it's a good news 13:41:51 My option in pike 13:42:14 great. 13:42:23 more thoughts? 13:42:57 container is hot now maybe 13:43:00 ironic has a bp, container is there for improvement, adoption is already started, health improvement needs update from xinhui 13:44:02 Yes 13:44:05 XueFeng, when ironic support work has started, please help ensure it is tracked in the workitems etherpad 13:44:13 in my mind, both ha and container are very important. For HA, we need a practical use case, for container, we need deployment and orchestration support 13:44:19 And maybe some stories container in baremetel 13:44:47 I mean something like k8s deployment(or at least something like compose) 13:44:53 container on baremetal would be next step when both components are stable 13:45:07 right 13:45:19 I'm thinking of a cluster config property 13:45:33 Ok 13:45:46 introducing yet another abstraction is gonna make things difficult to understand 13:45:48 Recently, some guys put into zun, and I talk with them 13:46:17 Agree with you, Qiming 13:46:34 zun is good if it can really make the container runtime easy to use/manage 13:47:09 still a lot of things to do in the HA space 13:47:20 more thoughts? 13:47:23 And we need to let them know senlin and zun 13:47:31 yes, if it is reliable for managing container, will be really helpful 13:47:33 Which is different 13:48:01 I think that team is aware of both projects' mission 13:48:24 I'd like to see closer integration happen earlier than later 13:48:54 So we can integrate zun+senlin 13:48:57 Right? 13:49:03 why not? 13:49:10 Yes 13:49:32 #topic boston summit prep 13:50:05 I'm only included in one of the talks accepted, the senlin + mistral one 13:50:18 so that will be my top priority 13:50:43 I believe xinhui is involved in all three talks? 13:50:57 Yes 13:51:10 maybe she will be the best driver for this 13:51:26 will sync with her offline 13:51:33 :) you add a lot work to her. 13:51:50 or better yet, online, when she stops by the #senlin channel 13:51:57 congratulate senlin:accepted three sessions in bostion summit. 13:52:12 elynn, I think she enjoys driving things 13:52:31 sorry just dropped 13:52:34 or, she may want to offload that to you, elynn 13:52:42 /s/bostion/boston 13:52:48 Qiming, I believe what you say. 13:53:01 :D 13:53:06 HaHa 13:53:17 I will handle all trivial things :) 13:53:36 alright, people need some believes, no matter what he/she is working on 13:53:58 #topic open discussions 13:54:24 since we cannot talk about talks without xinhui joining us 13:54:33 moving on to open discussions 13:54:58 Will you go to boston summit Qiming ? 13:55:03 what is the exact date of summit 13:55:15 May 5-11 13:55:20 May 08 - May 12 13:55:23 May 8-11 13:55:25 https://review.openstack.org/#/c/445298/ 13:55:35 \o/ 13:55:39 I love pep8 errors 13:56:04 LoL 13:56:24 :D 13:56:31 :) 13:56:43 it took me almost 3 hours to narrow this down to a pep8 error, :) 13:57:28 anything else, gentlemen? 13:57:34 nope from me 13:58:04 no~ 13:58:22 Hi all, I think container is important in pike cycle:) 13:58:25 thanks for joining, I have saved one minute for you, remember it 13:58:33 #endmeeting 13:58:56 yanhan has to end the meeting 13:59:52 network connection is terrible... 14:00:03 hi, Qiming, has the meeting finished? 14:00:14 if so, I will end the meeting 14:00:37 ok, seems so 14:00:40 #endmeeting 14:01:37 yanyanhu: maybe you need to rename to yanyanhu_ ? 14:01:39 #endmeeting