Tuesday, 2017-05-16

*** dixiaoli has joined #senlin00:54
*** dixiaoli has quit IRC00:55
*** dixiaoli has joined #senlin00:55
*** elynn has joined #senlin02:06
*** yuanying has quit IRC02:51
openstackgerritMerged openstack/senlin-dashboard master: Updated from global requirements  https://review.openstack.org/46449003:13
*** shu-mutou has joined #senlin03:24
*** yuanying has joined #senlin03:58
*** yuanying has quit IRC04:23
*** yuanying has joined #senlin04:23
*** yuanying has quit IRC05:18
*** yuanying has joined #senlin05:19
*** yuanying has quit IRC05:19
*** yuanying has joined #senlin05:20
*** yuanying has quit IRC06:06
openstackgerritQiming Teng proposed openstack/senlin master: Change requirement assertion to 'must'  https://review.openstack.org/46494306:38
*** yuanying has joined #senlin06:59
*** dixiaoli has quit IRC08:04
openstackgerritQiming Teng proposed openstack/senlin master: Change requirement assertion to 'must'  https://review.openstack.org/46494308:26
*** dixiaoli has joined #senlin08:28
*** chenyb has joined #senlin08:31
openstackgerritQiming Teng proposed openstack/senlin master: Trivial: Use correct words in inequality assertions  https://review.openstack.org/46497608:36
*** shu-mutou is now known as shu-mutou-AWAY08:44
*** dixiaoli has quit IRC09:15
*** dixiaoli has joined #senlin09:16
*** Guest67875 has joined #senlin09:33
openstackgerritQiming Teng proposed openstack/senlin master: Add NodeAdoptPreviewRequest  https://review.openstack.org/46500009:33
*** Guest67875 is now known as XueFeng09:36
*** elynn has quit IRC09:46
*** elynn has joined #senlin09:47
*** elynn has quit IRC09:47
*** elynn has joined #senlin09:48
*** XueFeng has quit IRC09:51
*** XueFeng has joined #senlin09:53
*** elynn has quit IRC10:05
*** dixiaoli has quit IRC10:14
*** chenyb has quit IRC10:16
*** chenyb has joined #senlin10:33
*** chenyb has quit IRC11:27
*** chenyb has joined #senlin12:09
*** chenyb has quit IRC12:54
*** elynn has joined #senlin12:56
Qimingnot sure we will have a weekly meeting today12:59
Qimingsince most of us have just returned from a long week of summit activities13:00
Qimingelynn, haiwei, XueFeng ?13:00
*** chenyb has joined #senlin13:01
elynnI have nothing to update.13:01
Qimingtwo things from my side13:02
*** Ruijie_ has joined #senlin13:02
Qimingimproving node adoption, just proposed a patch, separating the preview request from adopt request13:02
elynnSaw your email that this meeting was cancelled, if you have sth we could attend this meeting.13:03
Qiminganother thing is trying to fix py35 gate jobs: https://review.openstack.org/46456813:04
Qimingother than those two items, I don't have a lot to update13:04
Ruijie_https://review.openstack.org/#/c/463276/  please help review if you have time :)13:05
Ruijie_Didn't find a better way other than this one to do health check when processing scaling actions.13:06
Qimingruijie, I checked that patch a few days ago13:06
Qimingit looks ... ugly13:07
QimingI was thinking about taking a step back13:07
Qimingwhat is the usage scenario?13:07
Ruijie_yes Qiming, a little bit .. for the current model ..13:08
*** XueFengLiu has joined #senlin13:08
Qimingso .. the requirement is this, (trying to sort it out), when you are scaling in a cluster13:09
Ruijie_the decision should be made excatly after scaling policy and before deletion and other policies13:09
Qimingyou want to make sure the nodes' health status updated before removing nodes13:09
Qimingno matter how many nodes you want to remove eventually, the need is about making sure senlin knows the exact state of each node, right?13:10
Ruijie_yes Qiming13:10
Qimingthen how about this ...13:10
Qimingadd an optional parameter to the scale in api13:11
Qimingdo health check if 'health_check' is specified, before triggering the cluster scale in action13:11
Ruijie_Qiming, you mean in engine.cluster_action.execute()/_execute() or even more ahead ?13:13
Qimingbefore triggering that action13:13
QimingI cannot envision a case where health check operation itself needs a policy, so it can be thrown out of the action execution logic13:14
Qiminghow about here: http://git.openstack.org/cgit/openstack/senlin/tree/senlin/engine/service.py#n134413:17
Qimingyou know what13:18
Qimingthat brings me to another thought13:18
Qimingmaybe we should leave this to users13:18
Qimingif they want to do a health check before scaling in a cluster13:18
Qimingthey should do cluster_check call before cluster_scale_in call13:18
Qimingtwo calls13:19
QimingI have seen another patch that adds health check to the cluster_recover api, which is ... not consistent with our current design13:20
Ruijie_actually, that makes sense ...13:20
Qimingmerging two calls into one will always create a lot headaches13:20
openstackgerritQiming Teng proposed openstack/senlin master: Add NodeAdoptPreviewRequest  https://review.openstack.org/46500013:23
Qimingwe cannot, by any means, ensure that nodes that were active remain active between the time it was checked and the time cluster is scaled in13:23
Qimingafter health check, a node can still fail13:24
Ruijie_after all, we want to provide primitive functions but not the services13:24
Qimingright, that is another way to put it13:24
Ruijie_I mean to be atomic13:25
Qimingif we can find a way to solve a user problem easily, we can do it13:25
Qimingfor the scaling operation in an unstable environment, we can promise nothing13:25
Ruijie_okay Qiming, that makes sense13:26
QimingI have seen you trying this in many ways, none of them makes us feel comfortable, but still, many thanks13:27
Ruijie_so the work will be: 1. improve cluster_check (lb faiure need to be handled), 2. remove cluster_check login in cluster_recover, 3. improve cluster_recover to check desired_capacity13:27
Ruijie_np Qiming, that's my job :)13:28
Qimingitem 2 came from a patch13:28
Ruijie_yes Qiming, I sae that "check" parameter13:29
Qimingthis one: https://review.openstack.org/#/c/462802/13:29
Ruijie_that may change the right value to wrong, like the NS_WARNING which was set by lb_policy13:30
QimingI didn't think about it for long time13:31
Qimingit is acceptable, but ... I'm a little bit concerned about the direction we are going13:31
*** elynn has quit IRC13:36
*** XueFengLiu has quit IRC13:38
*** XueFengLiu has joined #senlin13:38
chenybhttps://review.openstack.org/#/c/464027/ please help review if you have time :)13:59
*** chenyb has quit IRC14:23
*** chenyb has joined #senlin14:23
*** chenyb has quit IRC14:27
*** dixiaoli has joined #senlin14:28
*** dixiaoli has quit IRC14:28
*** dixiaoli has joined #senlin14:29
*** XueFengLiu has quit IRC14:30
*** XueFengLiu has joined #senlin14:30
*** dixiaoli has quit IRC14:36
*** Ruijie_ has quit IRC15:17
*** XueFengLiu has quit IRC16:15
*** openstackgerrit has quit IRC18:17

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!