Wednesday, 2015-07-15

*** lixinhui has joined #senlin00:02
*** lixinhui has quit IRC00:07
*** Qiming has quit IRC00:08
*** lixinhui has joined #senlin00:29
*** lixinhui has quit IRC00:34
openstackgerritMerged stackforge/senlin: Getting Started Doc (5)
*** lixinhui has joined #senlin00:56
*** jruano has joined #senlin01:02
*** Qiming has joined #senlin01:04
*** Qiming_ has joined #senlin01:05
*** lixinhui has quit IRC01:08
*** Qiming has quit IRC01:09
openstackgerritQiming Teng proposed stackforge/senlin: Remove tox and requirements hacking
*** lixinhui has joined #senlin01:24
*** Yanyanhu has joined #senlin01:25
*** lixinhui has quit IRC01:29
*** Yanyanhu has quit IRC01:29
*** Yanyan has joined #senlin01:34
*** jruano has quit IRC01:35
*** lixinhui has joined #senlin01:51
*** lixinhui has quit IRC01:56
*** lixinhui has joined #senlin02:18
openstackgerritMerged stackforge/senlin: Remove tox and requirements hacking
*** lixinhui has quit IRC02:27
openstackgerritMerged stackforge/python-senlinclient: Updated from global requirements
openstackgerritMerged stackforge/senlin: Updated from global requirements
*** lixinhui has joined #senlin02:45
*** admin__ has joined #senlin02:52
*** lixinhui has left #senlin02:52
*** Qiming_ has quit IRC02:55
openstackgerritYanyan Hu proposed stackforge/senlin: Clean the init progress of some Senlin drivers
openstackgerritYanyan Hu proposed stackforge/senlin: Remove initilization of auth and session attrs in some drivers
*** admin__ has quit IRC03:30
*** lixinhui has joined #senlin05:28
*** Qiming has joined #senlin05:59
openstackgerritCindia-blue proposed stackforge/senlin: enable scheduler_hints when nova server create
Yanyanhi, Qiming, found the reason of the error that occured during 'list' operation06:16
Yanyanthe 'params' argument is changed in latest patch in sdk06:16
Yanyanseems they made a decision after discussion...06:18
Qimingokay, so we may need to revise our driver code06:28
Yanyanwill make a change on both driver and client side06:29
openstackgerritCindia-blue proposed stackforge/senlin: enable scheduer_hints
Qiminghave you seen the link shared by Julio?06:34
Yanyannot yet, where did he share it?06:34
Qimingon mailinglist, forwarding it to you all06:35
Yanyanok, let me have a look06:35
Yanyanok, thanks06:37
Yanyansaw the mail in maillist06:37
Qimingthe second link doesn't work for me06:42
YanyanI'm downloading this doc, but very slow06:43
Yanyanwill cost about 20mins06:43
openstackgerritYanyan Hu proposed stackforge/senlin: Using openstacksdk function call in keystone_v3 driver
openstackgerritYanyan Hu proposed stackforge/python-senlinclient: Using list query arguments as a single dict input parameter
QimingYanyan, with your change ^, we will need to use the master head of openstacksdk?06:50
Yanyanemm, that's true, so I guess we may need to block these patches until sdk provides a new release06:51
Yanyane.g. 5.106:51
QimingI cannot recall when, but there was an anncouncement from Brian that they are not accepting new resources to the sdk until they are finishing their 1.0 release06:52
Qiminga 1.0 release sounds like an official one to me06:52
Yanyanok, so maybe that will be a 'stable' one we want to rely on when we do our own release :)06:53
YanyanI will finish another patch that revises the query argument usage in other drivers and then block all these three patches until sdk updates its release version06:55
openstackgerritYanyan Hu proposed stackforge/senlin: Using openstacksdk function call in keystone_v3 driver
Yanyanwhy they remove this constrains07:07
Qiming:) a single contributor project now qualifies07:07
Yanyaninsteresting :)07:07
Yanyanthe one who is PTL, core member and also the only contributor ;)07:08
lixinhuidoes openstack connection can provide novaclient equal fucntions?07:36
lixinhuiI am checking its support to server_group07:36
lixinhuiis in limits07:37
lixinhuibut novaclient can support whole server_group calls07:37
lixinhuisuch as I wanna get one server_group by id07:37
lixinhuiI donot know how to reach this by openstacksdk connections07:38
openstackgerritOpenStack Proposal Bot proposed stackforge/senlin: Updated from global requirements
Qiminglixinhui, I'm trying to find an email from the mailinglist07:50
Qimingit was sent by Brian07:50
openstackgerritYanyan Hu proposed stackforge/senlin: Using openstacksdk function call in keystone_v3 driver
Yanyanhi, Qiming, just realized that we actaully don't need to revise the driver implementation in Senlin service since the change in openstacksdk only influence the interface of resource call but not the function call07:58
Qiminglixinhui, not sure if the 'PUT' or 'PATCH' actions are supported07:59
Qimingit is not well documented07:59
QimingYanyan, okay07:59
Qimingthat is good news07:59
Yanyanso just proposed a change to keystone driver which is still using resource call for some cases07:59
Yanyanyes :007:59
lixinhuiwhat does that mean?08:01
lixinhuishould we propose some function add to openstacksdk or08:01
lixinhuijust format/request type problem08:02
Qimingwe have to do this via sdk08:02
Qimingokay, I found the link:
Qimingthat is bad news08:03
Qimingneed to check out their agenda08:04
lixinhuican we drive them? :-)08:06
lixinhuiI can move ahead to policy/plugin works for our demo but depend on server_group/scheduler_hints at nova layer08:07
QimingI don't think pushing them is gonna change anything08:27
lixinhuiI see08:27
QimingSince the email was sent a week ago, I think they are drawing pretty near to their release08:27
lixinhuioh, I did not notice that.08:30
lixinhuianyway, I can try patach way08:30
Qimingyes, you can try this sdk patch and continue the senlin patch08:31
Qimingsigh, network is very slow ...08:31
*** jdandrea has quit IRC08:32
openstackgerritMerged stackforge/senlin: Updated from global requirements
QimingYanyan, have you solved the tox env problem?09:06
YanyanQiming, yes, it is caused by my low virtualenv version09:09
Yanyaneach version virtualenv has different pip version09:09
Yanyanso after I updated my virtualenv to 1.3.0, problem is solved :)09:10
Yanyan1.1.0 virtualenv only support 6.1 version pip09:10
Yanyansorry, virtualenv 13.009:10
Qimingholy shit09:12
Qimingpypi is not accessible09:12
Qimingcannot install oslo.utils09:12
Yanyanour network is really poor...09:13
Qimingit was blocked by the firewall09:13
Yanyangreat firewall?09:13
Qimingeach time after the so called network device upgrade, our network connection becomes more unstable09:17
Qimingwell, I am not using venv09:20
Qimingbut I don't get why tox still run pip 6.1.1, not 7.1.009:21
Yanyanthat's weird09:21
YanyanI found some other people asked this question, but no answer was given...09:21
YanyanI searched in bing09:21
Yanyanthat tox used incorrect pip version09:22
Qimingwhich version of tox are you using?09:23
Yanyanlet me check09:24
YanyanI think it's 1.7.009:24
Qimingtox --version09:24
Qimingpbr --version09:24
Qimingpip --version09:25
Yanyan0.11.0 pbr09:25
QimingI have updated to 1.3.009:25
Yanyan7.1.0 pip09:25
Yanyanoh, if pbr=1.3.0, I can't run senlin-engine correctly09:25
Yanyanso I always manually downgrade it to 0.1109:26
Yanyanpbr version conflict will be given if pbr version is 1.3.0 in my env09:26
Yanyanold problem...09:28
Qiminglifeless gave me an answer on heat channel09:35
Yanyanso even you don't use virtual env, you still need to update it?09:37
Qimingyes, tox is using it09:38
Qiming13.1.0 is recommended09:38
Qimingearlier versions won't understand the environment constraints newly added09:38
Yanyanoh, right09:38
YanyanI forget this part09:38
Yanyantox will use virtualenv09:39
Qimingnow ....09:39
Qimingthe last problem, recreating a tox venv for testing, that will take time09:39
Qimingwhy we were born in this rediculous country ...09:39
Yanyanwe can't choose ;p09:40
Yanyandecades ago09:40
Qimingby the way, I'm adding user/project/domain fields to policy object09:40
Qimingthe reason is that we have actions targeting at them09:41
Qimingand for webhook to work properly, we have to check if the requester is the "owner" of the targeted object or the "admin"09:41
Qimingand we cannot do this via policy.json, because we don't know which policy scope will be used09:42
Yanyanso actually if an entity can talk with other services, it need these three properties09:42
Qimingmaybe we should add these fields to profile as well09:42
Qimingnot necessarily09:42
Qimingwhen a requester A wants to create a webhook targeting a policy, with action 'POLICY_ATTACH'09:43
QimingI have to check if A is the owner or an admin09:43
Qimingthis is necessary because we have the "webhooks:create" door widely open09:43
Qimingthis has to be done manually09:44
Yanyanactually, when we try to talk with other services to create some extra resources when doing attach_policy, should we use the user/domain/project of policy to do the request?09:45
Yanyanthe same problem09:45
Qimingyou will check the target the poicy is attached to09:46
Qimingusually a cluster09:46
Qimingand use the cluster's user/project/domain info09:46
Qimingactually, we may need to check if the policy and the cluster are owned by the same user before trying the attach operation09:47
Yanyanok, why we don't use the user info in context?09:47
Qimingnot sure we should be so strict in that09:48
QimingI want to remove context from calls09:48
Qimingcontext info is bound to a transaction/interaction with the service09:49
Qimingbut the object will be owned by certain user/project09:49
Qimingthose are immortal info09:49
Yanyanso the context is used to do the top-half authentication while user/domain/project info of the owner will be used in the bottom-half progress09:50
Yanyanthe context decided whether a user can access some senlin resources, e.g. policy, cluster09:50
Qimingthere is no context, we created it09:50
Yanyanwhile the owner of cluster decided whether it can talk with nova and create a instance09:51
Qimingwe use trust there09:51
Yanyanyes, just currently, it is the context09:51
Yanyanowner user/domain/project + trust_id will be used to create sdk connection09:52
Qimingwe are passing context info everywhere, while we have decided those things should be done with a trust created in the first place09:53
Qimingthis is making the whole code base a lot difficult to maintain09:54
Yanyanso currently, context will be removed after the request workflow reached engine09:54
Yanyanand the following operations will be done without context09:54
Qimingoh, shit, the mock-1.1.3 hash code is wrong on pypi, another stupid error09:55
Qimingthe only use of it for us, is to interact with DB09:55
Qimingthat can be resolved some day I believe09:55
Yanyanso I said after we queried object(cluster, node, policy)  back from DB, we don't need context :)09:56
Yanyanem, if DB don't need it, we can remove it from API layer I guess09:56
Qimingsenlin will then operate on behalf of those object owners09:57
Qimingthat is possible09:57
Yanyanwill leave, have a call at 8:00 pm...09:57
*** Yanyan has quit IRC10:00
*** Yanyan has joined #senlin10:00
*** Yanyan has quit IRC10:05
openstackgerritQiming Teng proposed stackforge/senlin: Add user/project/domain to policy objects
*** Qiming has quit IRC10:49
*** openstack has joined #senlin11:39
*** Qiming has joined #senlin11:54
*** Qiming has quit IRC12:26
*** Qiming has joined #senlin12:26
openstackgerritQiming Teng proposed stackforge/senlin: Temporarily unmask nova exception
*** lixinhui_ has joined #senlin13:14
openstackgerritMerged stackforge/senlin: Temporarily unmask nova exception
*** jdandrea has joined #senlin13:37
*** lixinhui_ has quit IRC14:24
*** lixinhui_ has joined #senlin14:25
*** Qiming has quit IRC16:37
*** htruta_ has joined #senlin21:30
*** htruta_ has quit IRC22:05
*** jruano has joined #senlin22:10
*** htruta_ has joined #senlin22:50
*** htruta_ has quit IRC23:13

Generated by 2.14.0 by Marius Gedminas - find it at!