Wednesday, 2015-12-16

*** xuhaiwei_ has quit IRC00:02
*** pratikma_ has quit IRC00:02
*** pratikmallya has joined #senlin00:07
*** Qiming has quit IRC00:16
*** xuhaiwei_ has joined #senlin00:57
*** xuhaiwei_ has quit IRC01:10
*** gongysh has joined #senlin01:28
*** Qiming has joined #senlin01:36
openstackgerritlvdongbing proposed openstack/senlin: Remove senlin.conf.sample
*** pratikmallya has quit IRC01:38
*** xuhaiwei_ has joined #senlin01:46
*** zhenguo has joined #senlin01:47
*** Yanyanhu has joined #senlin01:49
xuhaiwei_when is the spring festival vocation in China this year?01:51
Qimingvacation 2016.02.07-2016.02.1301:52
*** elynn has joined #senlin01:52
Yanyanhuspring vacation?01:54
xuhaiwei_want to hear your available time01:55
*** elynn_ has joined #senlin01:57
*** elynn has quit IRC01:58
openstackgerritlvdongbing proposed openstack/senlin: Make health_mgr_opts auto generate to config sample
Qimingxuhaiwei_, I think the Jan 11 week would be okay02:16
xuhaiwei_it is ok for me Qiming02:17
Qimingbefore that, it will be the New Year's day, some may go vacation02:18
Qimingafter that week, some may go vacation for the Spring Festival02:18
*** Liuqing has joined #senlin02:22
openstackLaunchpad bug 1517353 in senlin "Node can't be created for KeyError: 'server'" [Undecided,New] - Assigned to Haiwei Xu (xu-haiwei)02:34
Qimingthis is still valid bug?02:34
xuhaiwei_no valid02:37
xuhaiwei_fixed by SDK side02:38
openstackgerritMerged openstack/senlin: Make health_mgr_opts auto generate to config sample
*** lixinhui has joined #senlin03:07
Qiminglixinhui, I'm marking bug #1525897 as an invalid report03:15
openstackbug 1525897 in senlin "sqlalchemy api crash for Conext has no attributes " [Undecided,Invalid] - Assigned to Cindia-blue (miaoxinhuili)03:15
*** yuanying has quit IRC03:22
lixinhuiOkay, Qiming03:25
openstackgerritQiming Teng proposed openstack/senlin: Add blue-green deployment as a feature request
*** yuanying has joined #senlin04:07
*** elynn_ has quit IRC04:12
*** elynn_ has joined #senlin05:00
*** Liuqing_ has joined #senlin05:35
*** Liuqing has quit IRC05:38
*** pratikmallya has joined #senlin05:55
*** pratikmallya has quit IRC07:11
QimingYanyanhu, seems the transaction isolation problem still persists07:24
QimingI have revised all MutableList data type back to basic strings07:25
Yanyanhuok, will do more investigation about it based on this reverting07:25
Qimingif it is not solving the problem, switching back to basic strings won't help07:26
Qimingthe most visible impact is about performance07:26
Yanyanhuyes, I guess so07:26
Qimingit is ..... very slow now07:26
Yanyanhumore overhead from converting string to list and back07:27
QimingI'm abandoning this change07:31
Qimingit is not helping, sigh ...07:32
Yanyanhureally a issue takes headache...07:32
Qiminganother attempt: I'll try explicitly add action objects to sessions07:33
Yanyanhuyou mean?07:34
Qimingwhen doing mark_succeed07:35
Qimingwe are changing the fields of several actions together07:35
Qimingsome of them may get missed if not explicitly added to the current session07:35
Qimingthis could be improved by just adding a line: session.add(action)07:35
Qimingwill try that approach07:36
Yanyanhuactually I'm not very clear about this part. I thought eachtime an action is queried in a session, it will be added to this session by default.07:36
Qimingyes, that is the depended action07:37
Yanyanhunot sure this is a correct understanding07:37
Qimingbut for the dependent action, they are not automatically added sometimes07:37
Yanyanhuoh, I see what you mean07:38
Yanyanhuyes, we should make a try07:38
Qimingthat is why NODE_DELETE has been successfully marked as SUCCEEDED but cluster action is not updated correctly07:39
Yanyanhuyes, we met that issue before07:40
Qimingit is funny now07:41
QimingI have killed the tox functional test07:41
Qimingrestarted senlin-engine07:41
Qimingnow I'm seeing that senlin-engine automatically starts running things07:42
Yanyanhufunctional test is still running at backend?07:42
Yanyanhuseems crtl+z can't stop it07:43
Yanyanhusometimes, have to ps and kill the thread07:43
Qimingem, short ids are really convenient07:45
Qiming$ senlin profile-delete fc4 713 341 1d8 02a 65707:45
QimingProfile deleted: ['fc4', '713', '341', '1d8', '02a', '657']07:45
YanyanhuI never tried so short id07:46
Qimingyou should07:46
Yanyanhuusually use id consists of 8 letters07:46
Qimingit is really convenient07:46
Qimingif 'fc4' is too short, you will get warnings, not errors07:46
Yanyanhuadministrators will definitely like it :)07:47
Qimingyes, moving mouse to do copy is not fun07:47
Yanyanhuyes, especially when you have to scroll you screen07:48
QimingI think I'm now close to a better, cleaner solution07:48
Qimingfunctional tests are passing one by one07:48
Qimingthere is still an error, but not related to this damn thing07:49
openstackgerritQiming Teng proposed openstack/senlin: Fix action dependency problem
QimingYanyanhu, patch #258308 is the fix08:03
Qimingstill missing two places08:04
Yanyanhuwill check it08:04
openstackgerritQiming Teng proposed openstack/senlin: Fix action dependency problem
Qimingnow hopefully complete08:04
*** R_lixh has quit IRC08:06
*** tianc_ has quit IRC08:06
xuhaiwei_Qiming, about using lambda to replace the inner method, it's Ok for the source, but it is still difficult to test08:07
Qimingyou will need to mock some objects08:07
Qimingor else the lambda will fail08:08
Yanyanhuhi, Qiming, about 258308, do we need to refresh action in _mark_failed_action and _mark_cancelled_action?08:08
QimingYanyanhu, looking08:08
Yanyanhulike what we are doing in _add_action_dependency08:08
*** yuanying has quit IRC08:09
QimingYanyanhu, I'm inclined to avoid doing refresh()08:09
Qimingit is kind of breaking the atomicity of transactions08:10
Qimingbut not 100% sure08:10
QimingYanyanhu, check this:
Yanyanhuunderstand, it looks like an operation that around the default consistent level08:12
Qimingyes, so I hate the 'refresh()' hack08:12
Yanyanhuthus could make the behavior unpredictable08:12
Qimingeventually we may need to understand oslo.db better08:13
Qimingand use the enginefacade.writter kind of annotation08:13
Qimingor we need to consider distributed locks, like the one summarized here:
Yanyanhuhmm, distributed lock is really useful and also really complicated...08:14
Qiminganyway, we have TOOZ item in FEATURES.rst08:15
Qimingwe won't drop this ball08:15
QimingI'm hoping that currently we have a not-perfect-but-working solution08:17
Qimingbased on a simple glimpse on this: line 122-13208:18
Qimingwe really need to do a session.add(someobject)08:18
Qimingthe enginefacade thing could be useful for better control of using contexts08:19
Yanyanhuhmm, will look through this model to better understand the way oslo.db control the session08:19
*** lixinhui has quit IRC08:37
openstackgerritxu-haiwei proposed openstack/python-senlinclient: Don't show show_deleted column in node-list when not needed
openstackgerritQiming Teng proposed openstack/senlin: Implement purge_deleted command
openstackgerritMerged openstack/python-senlinclient: Don't show show_deleted column in node-list when not needed
openstackgerritxu-haiwei proposed openstack/python-senlinclient: Replace inner methods with lambda function
Qimingsigh ....09:26
Qimingproblem still not fixed09:26
Yanyanhuseems so...09:27
QimingI'm suspecting that this patch is the root cause:
Yanyanhuallow_async could be true by default for read transaction after applying this patch?09:35
Qimingdon't know09:35
Qimingwe don't have problems before09:36
QimingI don't think we have updated sqlalchemy09:36
*** zhenguo has quit IRC10:00
*** Liuqing_ has quit IRC10:01
*** Liuqing has joined #senlin10:02
*** Liuqing has quit IRC10:02
*** Qiming has quit IRC10:05
*** Yanyanhu has quit IRC10:27
*** Liuqing has joined #senlin10:40
openstackgerritxu-haiwei proposed openstack/python-senlinclient: Unit test for /v1/ part3
*** elynn_ has quit IRC10:58
*** Qiming has joined #senlin11:15
*** pratikmallya has joined #senlin11:24
openstackgerritMerged openstack/senlin: Implement purge_deleted command
*** pratikmallya has quit IRC11:44
*** pratikmallya has joined #senlin13:09
*** zhenguo has joined #senlin13:16
*** cschulz has joined #senlin13:18
cschulzHi, Has anyone created Heat resources that represent Senlin objects such as OS::Senlin::Cluster?13:19
*** pratikmallya has quit IRC13:24
Qimingcschulz, yes13:38
cschulzIs there a repository I can clone to get those?  They're not in
Qimingthey are still under development14:22
Qimingonly when the code is reviewed and merged can you get it from github14:22
*** Liuqing has quit IRC14:24
Qimingcschulz, even when cluster resource type is merged, it is not a recommended way to do autoscaling in heat14:28
cschulzBut we are evaluating Senlin and it doesn't make much sense unless you can use it to create clusters which are part of a Heat stack.14:30
Qimingwhy is that?14:30
*** Liuqing has joined #senlin14:31
Qimingsenlin creates clusters of heat stacks, and does autoscaling with more knobs (i.e. verbs)14:31
cschulzIf, as in our case, you need to build a multi-tiered solution, you need Heat orchestration.14:31
Qimingwith Heat, the only way to make change to your autoscaling solution is to do stack-update14:31
cschulzAnd if some part of that stack needs to be autoscaled, you need a cluster.14:31
cschulzAh, but you need to think beyond that.14:32
cschulzWith heat I should be able to create a stack that includes some alarms which get connected to a Senlin cluster.14:32
cschulzTo do the autoscaling.14:33
Qimingyes, in the long run, we have reasons to support senlin clusters in Heat14:33
Qimingright, with heat, you will create ceilometer alarms to trigger auto-scaling14:34
cschulzYes, and from our auto-scaling standpoint, Senlin isn't very compelling unless we can build Heat stacks which include Senlin clusters for auto-scaling.14:34
*** Liuqing has quit IRC14:36
cschulzI just really got my head around this stuff yesterday.  And immediately came to the conclusion that what I needed was OS::Senlin::Cluster resources so I could build my stacks.14:36
*** Liuqing has joined #senlin14:37
Qimingokay, won't try convince you then, :)14:38
cschulzLooking at the Blueprint for Senlin::Cluster, I think an important aspect is missing.  A cluster should have as one property a list of policies to be attached.14:38
Qimingif you need helps on this, we are always here, :)14:38
Qimingcschulz, makes sense14:39
cschulzWe were ready to go off and create our own Senlin::Cluster resource till I asked the question this morning.14:40
cschulzIs there somewhere we can get the code that has been built?14:40
Qiminghaha, elynn is from IBM too14:40
cschulzOK thanks14:40
Qimingyou can do a 'git review -d 247429' to get a copy of the code under review14:41
cschulz'review' is not a valid command for 'git'14:50
Qimingwell... you don't have git-review installed14:53
Qimingtry this? git fetch refs/changes/29/247429/8 && git checkout FETCH_HEAD14:53
Qimingor: git pull refs/changes/29/247429/814:53
cschulzgot it.  I'm new to Gerrit stuff.14:58
*** Qiming has quit IRC15:16
*** zhenguo has quit IRC15:49
*** Liuqing has quit IRC15:50
-openstackstatus- NOTICE: Gerrit will be offline for a software upgrade from 17:00 to 21:00 UTC. See:
*** pratikmallya has joined #senlin16:33
-openstackstatus- NOTICE: Gerrit is offline for a software upgrade from 17:00 to 21:00 UTC. See:
*** ChanServ changes topic to "Gerrit is offline for a software upgrade from 17:00 to 21:00 UTC. See:"17:06
*** pratikmallya has quit IRC17:27
*** pratikmallya has joined #senlin19:33
*** ChanServ changes topic to "IRCLog: | Bugs: | Review:,n,z"21:02
-openstackstatus- NOTICE: Gerrit has been upgraded to 2.11. Please report any issues in #openstack-infra as soon as possible.21:02
*** Qiming has joined #senlin23:43
*** pratikmallya has quit IRC23:54

Generated by 2.14.0 by Marius Gedminas - find it at!