Wednesday, 2015-12-02

*** zhenguo has joined #senlin00:45
openstackgerritlvdongbing proposed openstack/senlin: Move release note of 'remove-trigger-support' to right location
*** pm90_ has joined #senlin01:01
*** Qiming has joined #senlin01:02
openstackgerritMerged openstack/senlin: Move release note of 'remove-trigger-support' to right location
*** pm90_ has quit IRC01:24
*** Yanyanhu has joined #senlin01:29
openstackgerritjunxu proposed openstack/senlin: Store the actual UUID of obj_id in Senlin database for webhook create
openstackgerritlvdongbing proposed openstack/senlin: Pass environment variables of proxy to tox
*** elynn has joined #senlin01:54
openstackgerritMerged openstack/senlin: Pass environment variables of proxy to tox
*** lixinhui_ has joined #senlin02:34
openstackgerritxu-haiwei proposed openstack/python-senlinclient: Fix profile update parameters passing error
openstackgerritjunxu proposed openstack/senlin: Store the actual UUID of obj_id in Senlin database for webhook create
*** Qiming has quit IRC03:10
*** Qiming has joined #senlin03:10
openstackgerritMerged openstack/senlin: Store the actual UUID of obj_id in Senlin database for webhook create
*** elynn has quit IRC03:55
*** lixinhui_ has quit IRC04:01
*** jruano has quit IRC04:22
*** yuanying has joined #senlin04:42
*** lixinhui has joined #senlin05:01
openstackgerritxu-haiwei proposed openstack/senlin: Check new_profile before updating node
*** elynn has joined #senlin05:06
openstackgerritYanyan Hu proposed openstack/senlin: Revise action api interfaces used in functional test
openstackgerritQiming Teng proposed openstack/senlin: Fix status code returned from API
openstackgerritYanyan Hu proposed openstack/senlin: Revise action api interfaces used in functional test
*** elynn has quit IRC05:42
*** elynn has joined #senlin05:53
openstackgerritxu-haiwei proposed openstack/python-senlinclient: Fix profile update parameters passing error
openstackgerritxu-haiwei proposed openstack/senlin: Check new_profile before updating node
*** elynn_ has joined #senlin06:04
*** elynn has quit IRC06:04
openstackgerritxu-haiwei proposed openstack/senlin: Check new_profile before updating node
*** elynn_ has quit IRC06:18
*** elynn has joined #senlin06:19
*** pm90_ has joined #senlin06:20
openstackgerritMerged openstack/python-senlinclient: Fix help message error for node_create
openstackgerritMerged openstack/senlin: Check new_profile before updating node
*** gongysh has joined #senlin06:33
openstackgerritMerged openstack/senlin: Fix status code returned from API
Yanyanhuhi, Qiming, are you around?06:52
YanyanhuI found after applying this patch, the resp code of some API interfaces are still old one when I did some tests06:53
Yanyanhue.g. node create will return 200 rather than 20206:53
Yanyanhunot sure if it's my local env's problem06:54
Qimingread this:
Yanyanhuoh, I see06:55
Qimingit will take several patches to fix all of the return code issue07:00
Qimingthere are cases we will need to customize the header ('Location') field of the Reponse object07:00
openstackgerritQiming Teng proposed openstack/senlin: Reorganize documentation
Yanyanhuhi, Qiming, I think maybe we can change all status code at the first step since the code about 'success' could confuse the user if the actual status_code is different07:06
Yanyanhuor we can leave a comment there07:06
Qimingwhat do you mean?07:07
Qimingusers won't see the 'success' variable07:07
*** zhenguo has quit IRC07:07
Qimingactual status code can vary, it could be a 401, 404, 400 for sure07:08
Qimingwhat we are changing here is only about the status_code for success07:08
Yanyanhuoh, I mean the developer or some guys who want to check the code by themselves07:08
Yanyanhuyes, I mean the first time people see the code, they could feel xxx_create will return 20207:08
Qimingthis doesn't make sense to you?07:08
Yanyanhubut actual resp code is still 20007:09
Qimingthat is something I'm working on07:09
Qimingthey will never see 200 for xxx_create when patches are all in07:09
Qimingthey will only see the status_code set to the value of 'success' if the call succeeds07:09
Qimingthe actual response code is still 200 because the whole change has not yet landed07:10
Yanyanhuyes, understand. Just feel maybe we should let user know the status_code adjustment is still in progress before all patches are in07:10
Qimingno users will know that07:11
Qimingonly developers07:11
Qimingso far it is not impacting any code logic07:11
QimingI was doing this very very carefully in case it will break existing code07:11
Qimingat the moment, 'success' is there just as any garbage07:12
Qimingit is not used, why should any developer care?07:12
Qimingas developers ourselves, we know we are in the middle of some changes07:12
Yanyanhuok, make sense07:13
QimingI wasn't pushing in more patches in this thread (API revision) because I want to avoid dependencies07:16
QimingI'm now working on publishing the docs07:16
Qimingwhen that thing gets straighten up, and this "baby step" is reviewed and merged, I will switch back to this thread07:17
*** pm90_ has quit IRC07:17
Yanyanhuno problem. publishing doc should be the one have higher priority07:18
QimingI have to switch between several threads ...07:19
Yanyanhuyea, we really need more hands...07:19
Qimingstill cannot find a guide how to publish docs07:21
Qimingsigh ...07:21
YanyanhuI found only less than 10 projects have doc published here
Yanyanhuis there any limitation?07:23
Yanyanhuoh, 11 ones07:23
Yanyanhubut I saw Manila's doc07:23
Qimingokay, we are talking about different things again07:24
Yanyanhuit's also new for openstack07:24
Qimingyou are talking about API docs07:24
QimingI'm talking about developer docs07:24
Yanyanhuabout doc for developers from different openstack projects, didn't see anything officially before...07:26
openstackgerritMerged openstack/python-senlinclient: Fix profile update parameters passing error
openstackgerritMerged openstack/python-senlinclient: Updated from global requirements
Qimingin additional to understand how to organize the docs, how to publish them07:31
QimingI need to find out how to get 'Clustering' doc listed here07:31
Qimingasking questions on openstack-api channel is a waste of time07:36
Qimingno one is answering questions there07:36
elynnFind an old bug to publish docs, maybe it could help you.
openstackLaunchpad bug 1374073 in stevedore "need to publish docs to" [High,Fix released] - Assigned to Doug Hellmann (doug-hellmann)07:42
openstackgerritYanyan Hu proposed openstack/senlin: [Tentative]Disable multi-engine to debug functional test
elynnAnd a new merged patch to publish docs.
Qimingnot sure if this is all that is required: openstack-server-publish-jobs07:45
Qimingthere are also openstack-publish-jobs07:45
Qimingthere are jobs about gating07:45
elynnSubmit a patch to project-config first and I think other cores will help you review ;)07:47
Qimingyes, that is the last resort07:47
Qimingnot sure if this is it:07:48
elynnI think openstack-server-publish-jobs and python-jobs need to add to layout.yaml and openstack-publish-jobs need to add to projects.yaml07:48
elynnI guess...07:49
Qimingthen that is all07:49
Qimingcheck patch #25223007:49
elynnJust saw it, let's wait for jenkins...07:50
YanyanhuIt's HUGE!08:19
Qimingthe whole API08:20
Qimingthere are things to be changed too08:20
xuhaiweithis is your excellent job this afternoon?08:20
Qimingnot excellent at all08:20
xuhaiweidon't want to review it :(08:20
elynnQiming: you are superman!08:20
Qimingthings to do regarding the api doc08:22
xuhaiweishould we put it there?08:23
xuhaiweithe location08:23
Qiming1) version_info response is not an object 2) response status codes are not accurate 3) some response bodies will be changed08:23
Qimingthat is the central location for all api refs08:24
Qimingthe result will be published here:
xuhaiweiwhen having this samples, it's easy to judge whether the api has bugs08:25
Qimingthere is a huge effort to migrate doc composition from WADL to RST,  using Swagger08:25
Qimingbut I'm not sure when that will land08:25
Qimingif the api-site has started to block any new api docs, at least they will tell us where is the 'new location'08:26
Qimingthe current documentation has some defects08:29
Qimingwe are not clearly documenting the status_code for failures08:29
openstackgerritShu Muto proposed openstack/python-senlinclient: Remove py26 support
xuhaiweiin fact, I am a little afraid of that08:30
Qimingthere is another proposal here:
Qimingit is not preventing us from adding the existing api docs to api-site08:32
Qimingalthough in future, a reST/YAML style API doc is preferred08:33
openstackgerritxu-haiwei proposed openstack/senlin: Webhook creation refactor
Yanyanhuhi, xuhaiwei, around?08:42
YanyanhuI remember I said you met an error like this one when creating cluster?08:42
Yanyanhucluster_create_4909a4b5[2e1e61b7-cc05-430c-9217-d8aead4b232f] CLUSTER_CREATE - FAILED: CLUSTER_CREATE [2e1e61b7] timeout08:42
Yanyanhudid you know how to reproduce it.08:43
YanyanhuI found this error happened in gate side08:43
openstackgerritMerged openstack/senlin: Revise action api interfaces used in functional test
Yanyanhubut just can't reproduce it locally08:44
xuhaiweihi Yanyanhu08:44
xuhaiweiyes, that's what I met yesterday08:45
xuhaiweithat happened when I tried to create a cluster, but after the Error happened, the cluster created soon08:46
Yanyanhuso you just met this error once or it always happened08:46
xuhaiweilet me have a try now08:46
xuhaiweiI created 3 clusters, neither has this error08:48
xuhaiweinone of them08:48
Yanyanhudifficult to locate this error08:49
Yanyanhuwill wait for the functional test result in gate side08:49
xuhaiweimaybe a bug, maybe not08:51
xuhaiweithe default timeout value is how long?08:55
xuhaiweishould be longer than we see?08:55
*** lixinhui has quit IRC08:57
Yanyanhuxuhaiwei, it's 3600 seconds08:57
xuhaiweiso timeout should not happen?08:57
xuhaiweithat is a bug08:58
elynnMaybe race condition?08:58
Yanyanhuso I guess could be some bugs when calculating action lapsed time08:58
Yanyanhuelynn, you mean?08:58
elynnIs that because we have multi workers and some kind of race when acquire resources?08:59
openstackgerritxu-haiwei proposed openstack/senlin: Webhook creation refactor
elynnThen a deadlock happened?09:00
Yanyanhuhmm, I disabled multi-engine tentatively when executing functional test09:00
Yanyanhuso theoretically, that shouldn't happen...09:01
QimingI'd suggestion nailing down the problem step by step09:01
Qimingfirst of all, I am seeing a lot profile still in use errors09:01
Qimingthat means the functional test is not structured properly09:01
Qimingevery test_xyz methods are in theory executed in parallel09:01
YanyanhuQiming, that is because profile can't be cleaned correctly since they are still being used09:01
Qimingwhy they are still in use?09:02
Yanyanhusince cluster created failed and cluster is in error status09:02
Qimingif the cluster is in error status, why attempt to delete profile?09:02
Yanyanhuso related profile can't be deleted as expected when a test case is done09:02
xuhaiweido you ever meet this situation that every single test case is succeeded, but when test them all, some of them will fail ?09:03
Qimingxuhaiwei, yes09:03
Qimingthat happened a long time ago in Heat09:03
Yanyanhuhmm, currently, profile creation and deletion are put into setup and teardown method of test class09:03
xuhaiweithat's mock's problem?09:03
Qimingthe fix was to give each temporary stack a dedicated name09:03
xuhaiweiI met this problem in senlin client test09:03
Qimingor else some test cases will try delete stacks created by other test cases09:04
Yanyanhubut based on the log of devstack-gate, all functional test are run serially09:04
Yanyanhuno in parallel09:05
Yanyanhuhmm, I saw the log in both senlin-api and engine09:06
Qimingmaybe the logs are concatenated?09:09
Qimingdon't ever trust I/O results09:09
Qimingthose things are buffered at every layers09:09
Yanyanhuhmm, that is possible. I think a thing need to do is given a random name for all resources created during test progress09:10
Yanyanhucurrently, the same name are using in different test cases09:10
Yanyanhuthe second time test passed09:11
Yanyanhuhi, xuhaiwei, just saw your patch about webhook rework
YanyanhuI think incorrect service could be returned if only service type is used to filter the list result.09:20
Yanyanhusince Keystone allows creating multiple services with the same service_type09:20
Yanyanhuthey give an example about this in API doc09:21
Yanyanhuthis place, line 6909:21
xuhaiweifrom my test, the 'services' contains all the services09:22
xuhaiweiand each one has this format:  attrs={u'description': u'OpenStack Telemetry Service', u'links': {u'self': u''}, u'type': u'metering', u'enabled': True, u'id': u'6bcb49a47b7b41e581c9f3516077bc0d', u'name': u'ceilometer'09:23
xuhaiweithis is for ceilometer09:23
Yanyanhuyes, that's because in our current env, each service has unique service_type09:23
Yanyanhubut we can't guarantee that I guess09:24
Yanyanhuyou know that Monasca doesn't mark their own service type to 'metering' to avoid conflict with ceilometer09:24
xuhaiweiyou mean the 'type' could not be unique?09:24
Yanyanhue.g. two different services in Openstack are all 'metering' service, they compete with each other09:25
xuhaiweiso we may need to test 'type' and 'name'?09:25
Yanyanhuyes, I guess so09:25
xuhaiweiok, that makes sense09:25
xuhaiweiI will fix it later09:25
Yanyanhuthanks, other parts are ok for me09:26
*** zhangguoqing has joined #senlin09:30
QimingYanyanhu, I don't thing there are two metering services in OpenStack09:36
Qiminga more authorative information source would be keystoneauth09:37
Qimingmaybe we need to check their assumptions on this, instead of being unnecessarily cautious09:37
Yanyanhuyes, currently only ceilomter's service type is metering, Monasca's service type is monitoring09:38
Yanyanhuagree that we should get the answer from the community09:38
Qimingin real openstack deployments, if there are more than one services bearing the same type, it would cause a lot trouble09:39
Qimingfor example, in Heat, in Murano, ...09:39
Qiminganother argument is that openstack sdk is talking to services using service type09:40
YanyanhuJust beacuse keystone allows to do this, so I guess we need to be aware of this possible issue.09:41
Qimingif there are two services using the same type, it will break in a minute09:41
xuhaiweicurrently when try to get keystone service, only 'identity' type is provided09:41
Qimingbecause it requires an admin role?09:42
xuhaiwei{'service_type': 'identity', 'interface': 'public', 'version': None, 'region_name': None, 'service_name': None}09:43
xuhaiweionly these information09:43
Yanyanhuso I guess we'd better get an answer from keystone or the community09:44
Qimingxuhaiwei, you may need an admin role to get the full list09:44
Qimingxuhaiwei, check this:
Qimingalthough I think that is stupid09:45
xuhaiweiso when creating a webhook, we have a 'admin' role?09:46
Yanyanhu(openstack) service list09:46
Yanyanhu| ID                               | Name       | Type           |09:46
Yanyanhu| 11af2389070449f98f294c5cf2446a5f | service1   | compute        |09:46
Yanyanhu| 4a203daa7df745f2bbb21bad0fe4302e | service2   | compute        |09:46
YanyanhuI created two fake compute services in my local env09:47
xuhaiweithat's true, you have two compute node, i think09:47
Yanyanhujust used openstackclient to do this test09:49
xuhaiweianyway, I made a mistake in that patch09:50
xuhaiweithe exception handling logic is wrong09:50
xuhaiweiwe should keep the original type09:50
xuhaiweisorry for that09:50
QimingYanyanhu, then you try heat and see if it is possible to run a thing?09:50
Qimingtoday, most service catalog discovery is done based solely on service_type09:52
Yanyanhuyes, that's true09:53
xuhaiweiI want to go home09:57
Yanyanhuprepare to leave too09:57
xuhaiweiso the conclution is?09:58
Yanyanhumaybe throw a question to mailing list?09:58
QimingI'm okay with the name as an additional filter09:58
Qimingbut I'm curious how useful that would be09:58
Yanyanhuor add a comment here to delete it someday?09:58
Qimingrecall that the other day we are trying to get a 'compute' service named 'senlin'09:59
Qimingdon't know why we ended up such a weird condition09:59
YanyanhuI guess the filter didn't work as expected?10:00
Yanyanhusince it is passed to sdk's 'services' function call10:00
xuhaiweithe original filter didn't work , so I fixed it10:00
Qimingyes, my feeling was that we are introducing some problems to fix10:01
Qimingbut I'm fine with it if we are also solving this at the sdk side10:01
xuhaiweiin my opinion, the service_type should be unique in the whole openstack environment10:02
xuhaiweibut making a double check is also not wrong10:03
Yanyanhumaybe some bugs in sdk side10:03
Yanyanhuanyway, will make a test on services function call10:03
Yanyanhuusing those fakes services :)10:03
Qimingit was really a torture -->
Qimingthere are so many unnecessary json format checks10:06
QimingI tried that, Yanyanhu10:07
Qimingthat thing worked10:08
Qimingso I am not strongly against adding service name check10:08
QimingI was simply saying that it may be not necessary, and it seems revealing bugs in sdk10:09
Yanyanhuok, actually I also tested it locally before adding that code10:09
Yanyanhuyes, so maybe we just add TODO for removing it in future10:09
xuhaiweithe patch is coming10:10
YanyanhuI'm also curious whether TC will allow the service type conflict if a new project propose application for joining bigtent :P10:10
Qimingpatch #252245 passed!10:10
Qimingthe first checklinks gate is not caused by this patch10:10
Yanyanhuthat means one project are yelling to another project 'I'm your competitor' :)10:11
openstackgerritxu-haiwei proposed openstack/senlin: Webhook creation refactor
xuhaiweisince OpenStack is a whole family, it should not allow two project has the same type10:14
YanyanhuI also think so10:15
YanyanhuI think the follower won't choose the same service type as an existing service10:16
openstackgerritQiming Teng proposed openstack/senlin: Reorganize documentation
xuhaiweiok, time to go10:17
xuhaiweisee you tomorrow10:17
Qimingme too10:17
Yanyanhuwill leave too10:18
Yanyanhusee U guys10:18
*** Qiming has quit IRC10:22
*** Yanyanhu has quit IRC10:23
*** elynn has quit IRC10:24
*** yuanying has quit IRC10:57
*** Qiming has joined #senlin11:28
*** lixinhui_ has joined #senlin11:30
*** Qiming has quit IRC11:34
*** pm90_ has joined #senlin12:10
*** lixinhui_ has quit IRC12:32
*** Qiming has joined #senlin12:56
openstackgerritMerged openstack/senlin: Webhook creation refactor
*** Qiming has quit IRC13:08
*** elynn has joined #senlin13:41
openstackgerritMerged openstack/python-senlinclient: Remove py26 support
*** pm90__ has joined #senlin13:48
*** pm90_ has quit IRC13:50
*** pm90__ has quit IRC14:35
*** elynn has quit IRC14:42
openstackgerritMerged openstack/senlin: Add more configurable parameters to wait_for_server
*** pm90_ has joined #senlin15:00
*** pm90__ has joined #senlin15:02
*** pm90_ has quit IRC15:06
*** pm90__ is now known as pratikmallya15:12
*** zhangguoqing has quit IRC15:34
*** xuhaiwei has quit IRC15:34
*** gongysh has quit IRC15:34
*** pratikmallya has quit IRC15:34
*** jdandrea has quit IRC15:35
*** zhangguoqing has joined #senlin15:40
*** gongysh has joined #senlin15:42
*** xuhaiwei has joined #senlin15:42
*** pratikmallya has joined #senlin15:59
*** pratikmallya has quit IRC16:09
*** pratikmallya has joined #senlin16:34
*** pratikmallya has quit IRC16:45
*** pratikmallya has joined #senlin16:46
*** pratikmallya has quit IRC17:06
*** rebase_ has joined #senlin18:20
*** pratikmallya has joined #senlin19:06
*** jdandrea has joined #senlin19:59
*** jdandrea has quit IRC21:54
*** jdandrea has joined #senlin22:13
*** pratikmallya has quit IRC22:38
*** yuanying has joined #senlin23:00
openstackgerritxu-haiwei proposed openstack/senlin: Make asynchronous action return 202
openstackgerritMerged openstack/senlin: Reorganize documentation

Generated by 2.14.0 by Marius Gedminas - find it at!