Wednesday, 2015-09-02

*** Qiming has quit IRC00:07
openstackgerritMerged stackforge/senlin-dashboard: Updated from global requirements
*** zhenguo has joined #senlin00:30
*** xuhaiwei has joined #senlin00:39
*** Qiming has joined #senlin01:06
*** lawrancejing has joined #senlin01:33
*** mathspanda has joined #senlin01:36
*** jdandrea has quit IRC01:55
*** Yanyanhu has joined #senlin01:55
branwthe new version senlin,senlin create-profile with nova_server.spec,the spec add some item that must specify. such as type,version,properties..., i don't know type, version, properties should be, and the spec item "name" can not recognizable01:58
Qimingyes, branw01:58
branwanyone give me a new spec file example01:59
Qimingdid you check the senlin/examples/profiles dir?01:59
branwi am in the dir01:59
branwif the spec item type specify profile-type like os.nova.server, then senlin  create profile  shoud remove -t02:02
Qimingbranw, the client side change is there for approval02:03
openstackgerritMerged stackforge/python-senlinclient: Revise support to profile-create/show
Qimingit was checked in yesterday, and just approved today, ^02:03
*** LiuWei has joined #senlin02:05
openstackgerritMerged stackforge/senlin: Updated from global requirements
Yanyanhuhi, Qiming, about that bug caused by incorrect context usage in Action, I will try to propose a patch to differentiate the assignment of context inpurt params for those two different cases first.02:12
Yanyanhulike we discussed yesterday02:12
Yanyanhubut I found a problem that the target param is now the ID of target object02:13
Yanyanhuso should we change it to target object?02:13
*** Yanyanhu has quit IRC02:16
*** Yanyanhu has joined #senlin02:17
Yanyanhujust dropped02:19
Qimingpass an object reference over RPC?02:23
Yanyanhuno, just pass an object when create Action instance in engine service module02:24
Yanyanhulike this02:25
Yanyanhupass cluster rather than when creating ClusterAction instance02:26
*** elynn has joined #senlin02:26
Qimingthat's fine02:26
*** weli__ has joined #senlin02:31
*** LiuWei has quit IRC02:31
openstackgerritMerged stackforge/senlin: Sample profiles with type and version keys
*** lkarm has joined #senlin02:58
*** lkarm has quit IRC03:02
*** weli__ has quit IRC03:02
*** LiuWei has joined #senlin03:02
LiuWeihi,when I execute cluster create action,cluster create failed ,log prints that start_time in action is null.03:31
LiuWeihere is my log errors:
*** weli__ has joined #senlin03:31
*** weli__ has quit IRC03:34
*** LiuWei has quit IRC03:36
*** LiuWei has joined #senlin03:38
LiuWeicompute just reboot03:39
LiuWeican someone give me some tips?03:39
elynnThis might be a bug...05:10
LiuWeiIs that you also meet this question?05:12
Yanyanhuhi, LiuWei, this is a bug05:22
Yanyanhuwe are now trying to fix it05:22
Yanyanhubut since it is related to the refactoring of context usage in Action progress, it will take for a while to resolve this problem05:23
Yanyanhuno problem :)05:23
*** lkarm has joined #senlin05:40
*** lkarm has quit IRC05:45
*** mathspanda has quit IRC06:01
openstackgerritQiming Teng proposed stackforge/senlin: Unit test case for keystone v3 driver
*** Qiming_ has joined #senlin06:12
Qiming_Test message from phone06:14
Qiming_It works06:15
Yanyanhuhi, Qiming_, about refactoring context usage in action object, I found we have no place to store the 'is_admin' value if no context is passed to Action.__init__...06:20
Yanyanhuunless we use target to pass it06:20
Yanyanhuthis is not a correct way06:22
Qiminglet me try propose a fix soon06:26
Qimingpls check in your current fix regarding the new code in __init__06:26
Yanyanhuyou mean propose the patch I'm working on about contex?06:27
QimingI'm wondering if there have been already a lot of code on it06:32
Yanyanhuso maybe I should hold my patch since it still doesn't work06:33
Yanyanhujust don't want to make thing worse :)06:35
openstackgerritMerged stackforge/senlin: Add 'singleton' property to base policy
Qimingit is gonna be a huge revision again07:12
openstackgerritYanyan Hu proposed stackforge/senlin: Add functional test for cluster scale in/out
openstackgerritYanyan Hu proposed stackforge/senlin: Revise policy_type functional test
openstackgerritMerged stackforge/senlin: Fix functional test for policy type listing
openstackgerritMerged stackforge/senlin: Revise policy_type functional test
*** lawrancejing has quit IRC07:46
*** lkarm has joined #senlin07:52
*** lkarm has quit IRC07:56
QimingYanyanhu, I am afraid the bug of concurrency is actually caused by the list data type08:04
Yanyanhuyou mean the list data type provided by python?08:05
Qimingthis one08:06
Yanyanhuhow it happened08:07
Qimingit is a layer between what a threads sees from in-memory data and the db data08:07
Yanyanhuok, so just like a cache?08:07
QimingI heard complaints from heat team on this, it was copied to Heat and later deleted due to some data problems and it was not that useful in Heat anyway08:08
Yanyanhubut we use it for storing dependents08:09
Yanyanhuem, can we control the construction and destruction of these List objects?08:14
Qimingnot sure08:15
Qimingstill digging08:16
YanyanhuI was writing functional test for cluster resize and I think I found a bug which breaks the size contraints of cluster08:17
Yanyanhuwill fix it08:17
Qimingbe careful though08:17
Qimingthat part is getting complicated too08:17
Yanyanhugot it08:17
Yanyanhuah, my mistake, the default value of strict param is False, not True08:33
Qimingit is really weird08:50
QimingI even tried overriding the context of the derived actions08:50
Yanyanhuyou mean actually the two actions in cluster_action and node_action are the same?08:51
Yanyanhuin the test08:51
Qimingtrying to mock that08:52
Qimingactions are dispatched through RPC08:54
Qimingthe dispatcher/scheduler is creating new context instance?08:54
Qimingthe context is not passed over08:54
openstackgerritYanyan Hu proposed stackforge/senlin: Fix a bug in _wait_for_dependents method of ClusterAction
Yanyanhulet me check the code08:54
YanyanhuI guess the dispatcher/scheduler won't create a new context08:55
Yanyanhuit just used the one in service module I think08:56
Yanyanhuthe context created in engine should be passed to start_action method of Dispatcher class08:57
Qimingthat is a misunderstanding09:03
Qimingunder current framework, there is no way to pass the context or DB session from "parent" action to "child" action09:05
Qimingthe reason:09:05
Qimingthis line is still using the "parent" action's context:
Qimingbut when the Dispatcher (subclass of service.Service) receives this request:
Qiminga new context is created by the service.Service framework09:07
Qimingthey are not the samething09:07
Yanyanhuoh, so oslo_service.Service adds context decorator for all memeber methods?09:08
Yanyanhusince we didn't add it explicitly09:08
Qimingsome kind of that09:09
YanyanhuQiming, but I found oslo_service.server.Services class isn't decorated09:10
Yanyanhuso weird09:10
Qimingthe context parameter09:11
QimingI didn't say it comes from decorator09:11
Yanyanhuso you mean the context object is reconstructed when it was passed through the rpc09:13
Yanyanhuthe one sent in rpcclient side is not the same one that received in server side09:13
Yanyanhuhmm, I also doubted that transmitting an 'object' through rpc might be incorrect09:15
Yanyanhuno matter whether olso messaging handle the construction and destruction correctly, the object in client side and the one in server side should not be the same one09:16
*** elynn has quit IRC09:22
*** Yanyanhu has quit IRC09:23
*** LiuWei has quit IRC09:43
*** branw has quit IRC10:41
openstackgerritQiming Teng proposed stackforge/senlin: Fix cluster status refresh error
*** Qiming has quit IRC10:54
*** Qiming_ is now known as Qiming10:54
*** Qiming_ has joined #senlin11:49
openstackgerritQiming Teng proposed stackforge/senlin: Fix cluster status refresh error
*** lkarm has joined #senlin12:27
openstackgerritMerged stackforge/senlin: Fix a bug in _wait_for_dependents method of ClusterAction
*** jdandrea has joined #senlin13:13
*** zhenguo has quit IRC14:24
openstackgerritQiming Teng proposed stackforge/senlin: Delete dead code in sdk module
openstackgerritMerged stackforge/senlin: Fix cluster status refresh error
openstackgerritMerged stackforge/senlin: Remove half-baked profile types
openstackgerritMerged stackforge/senlin: Revise doc to reflect latest changes
openstackgerritMerged stackforge/senlin: Unit test case for keystone v3 driver
*** Qiming_ has quit IRC15:13
openstackgerritMerged stackforge/senlin: Delete dead code in sdk module
*** lkarm has quit IRC16:41
*** lkarm has joined #senlin16:42
*** lkarm has quit IRC16:44
*** lkarm has joined #senlin16:44
*** lkarm has quit IRC17:00
*** lkarm has joined #senlin17:37
*** lkarm has quit IRC17:55
*** lkarm has joined #senlin18:06
*** lkarm has quit IRC18:34
*** lkarm has joined #senlin18:35
*** lkarm has quit IRC20:29
*** lkarm has joined #senlin20:40
*** lkarm has quit IRC21:25
*** Qiming_ has joined #senlin23:26
*** elynn has joined #senlin23:35

Generated by 2.14.0 by Marius Gedminas - find it at!