Friday, 2017-01-06

*** liusheng has joined #openstack-mogan00:40
liushengzhenguo: the tempest patch merged failed :(00:46
*** wanghao has joined #openstack-mogan00:48
*** zhenguo has quit IRC01:13
*** little has joined #openstack-mogan01:15
openstackgerritliusheng proposed openstack/python-nimbleclient: The --nic parameter should be required
*** liusheng has left #openstack-mogan01:24
*** liusheng has joined #openstack-mogan01:25
*** kevinz has joined #openstack-mogan01:26
*** zhenguo has joined #openstack-mogan01:31
*** liusheng has quit IRC01:34
*** liusheng has joined #openstack-mogan01:35
*** zhangyang has joined #openstack-mogan01:37
liushengzhenguo: For tempest, it is weird, the console logs seems show the server has been deployed sucessfully, but the job run timeout01:39
liushengzhangyang: I have checked the tempest's log, it takes more than 10min between POST /instances/ and last GET01:40
liushengzhenguo: the status is stucked in 'building'01:40
openstackgerritzhangjl proposed openstack/nimble: TrivialFix: Remove the duplicate _destroy_network method
*** yuntongjin has joined #openstack-mogan01:48
*** Kevin_Zheng has joined #openstack-mogan01:50
zhenguoliusheng: you mean we should bump the job timeout?01:51
liushengzhenguo: I am not sure, but it is unreasonable the server stucked in 'building' more than 10 mins01:52
zhenguoliusheng: yes, let me check the log01:53
liushengzhenguo: according the console log, seems the server has been deploy sucessfully, don't found the network issue we saw before01:53
zhenguoliusheng: you mean the baremetal console log?01:53
liushengzhenguo: yes01:53
zhenguoliusheng: currently the tempest job timeout is what01:54
liushengzhenguo: seems 600s01:54
zhenguoliusheng: seems ironic's is 800s01:55
liushengzhenguo: I am not sure01:55
zhenguoliusheng: I remember when adding the job we set it to 800, but infra guys suggest we should make it to 600 at first01:55
*** yuntongjin has quit IRC01:56
zhenguoliusheng: I find some errors in screen-mogan-engine logs02:02
liushengzhenguo: ok, let me check02:04
zhenguoliusheng: in wait_for_active, TypeError: can't compare offset-naive and offset-aware datetimes02:04
zhenguoliusheng: after status goes to active and do instance.save02:05
liushengzhenguo: oh, seems a bug of mogan-engine ?02:05
zhenguoliusheng: oh, I  merged a patch before this02:06
liushengzhenguo: related with the error ?02:06
zhenguoliusheng: maybe it's because this
zhenguolet me test with it again02:06
liushengzhenguo: ok, let me update my env and run it locally02:08
zhenguoliusheng: K02:08
*** yuntongjin has joined #openstack-mogan02:08
liushengzhenguo: seems it has the  issue, it occured in my env02:14
zhenguoliusheng: oh, sorry02:14
liushengzhenguo: the Provisioning State is active when I run ironic node-list02:15
liushengzhenguo: hah, maybe we really need this tempest test as a gate02:16
zhenguoliusheng: hah, this tells us how important the tempest is, hah02:16
liushengzhenguo: aha02:16
*** zhangjl has joined #openstack-mogan02:35
zhenguoliusheng: not sure why that patch will impact it, seem bacause timezone issue for sqlachamy02:41
liushengzhenguo: seems like the format of timestamp ?02:41
zhenguoliusheng: yes, seems it is02:42
zhenguoliusheng: not sure why this is related to the patch adding instance aborted,02:49
zhenguoliusheng: seems because versioned object
openstackLaunchpad bug 1469120 in Cinder "Cannot save dates in Versioned Objects" [Medium,Fix released] - Assigned to Gorka Eguileor (gorka)02:49
liushengzhenguo: looking the but description, looks like a same issue02:50
zhenguoliusheng: yes02:50
zhenguoliusheng: hah, Ironic also face the same problem
liushengzhenguo: any new version of lib cause this ?02:59
zhenguoliusheng: no only because we do after obj.refresh02:59
liushengzhenguo: so how about borrow the ironic's fix ?03:06
zhenguoliusheng: yes, I'm testing with it now03:07
openstackgerritliusheng proposed openstack/nimble: Trivial-Fix: replace 'nimble' with 'mogan' in two places
*** Guest35757 has quit IRC03:08
*** luyao has joined #openstack-mogan03:11
luyaozhenguo, I have a question,How to enable mogan plugin in devstack?enable_plugin mogan git://
zhenguoluyao: yes03:32
zhenguoluyao: but you can use enable_plugin xxx git://
zhenguoluyao: the repo name is still nimble instead of mogan03:33
zhenguoenable_plugin mogan git://
luyaozhenguo,ok, let me try03:34
openstackgerritZhenguo Niu proposed openstack/nimble: Fix instance save after refresh error
zhenguoliusheng: this can fix the tempest issue03:36
zhenguoliusheng: please try03:36
*** yuntongjin has quit IRC03:36
* zhenguo brb03:37
liushengzhenguo: ok, let me try, thanks!03:37
liushengzhenguo: it works, cool!03:46
zhenguoliusheng: haha03:47
zhenguoliusheng: just copy ironic03:47
liushengzhenguo: hah03:47
openstackgerritMerged openstack/nimble: Trivial-Fix: replace 'nimble' with 'mogan' in two places
openstackgerritMerged openstack/nimble: Fix instance save after refresh error
*** wanghao_ has joined #openstack-mogan04:39
*** wanghao has quit IRC04:41
*** wanghao has joined #openstack-mogan04:41
*** wanghao_ has quit IRC04:44
*** zhangjl has left #openstack-mogan04:51
openstackgerritzhangjl proposed openstack/nimble: TrivialFix: Remove the duplicate _destroy_network method
*** yuntongjin has joined #openstack-mogan05:11
openstackgerritMerged openstack/nimble: Add tempest tests for instance creation
*** yuntongjin has quit IRC06:02
zhenguoliusheng: it's merged, lol06:05
liushengzhenguo: thanks ! lol06:06
zhenguoliusheng: thanks for the persistent!06:07
liushengzhenguo: np :)06:07
*** liudong has joined #openstack-mogan06:09
*** liudong has quit IRC06:11
*** yuntongjin has joined #openstack-mogan06:20
zhenguohi all, please check this again
zhenguofeel free to update it06:22
*** wanghao_ has joined #openstack-mogan06:30
*** wanghao has quit IRC06:33
*** zhangjl has joined #openstack-mogan06:39
*** yuntongjin has quit IRC06:42
openstackgerritMerged openstack/nimble: TrivialFix: Remove the duplicate _destroy_network method
shaohe_fengzhenguo: little:
shaohe_fengzhenguo: little: what do you think of the personality and user_data ?07:14
zhenguoshaohe_feng: hey, I see that section in the doc07:19
zhenguoshaohe_feng: is that also leverage config drive?07:20
zhenguoshaohe_feng: seems only admin can access that data07:20
zhenguoshaohe_feng: hi, I would suggest you split your patch into some seperate parts, we can add one type first, wdyt?07:22
*** yuntongjin has joined #openstack-mogan07:22
shaohe_fengzhenguo: OK.07:41
zhenguoshaohe_feng: thanks, and maybe it can split into some small tasks, hah07:42
shaohe_fengzhenguo: little: not sure the scenario for "user_data" and "personality", seem they can do the same thing.07:42
shaohe_fengzhenguo: just "user_data" is a string, and "personality" is file.07:43
zhenguoshaohe_feng: personality can specify the path07:43
shaohe_fengfile path.07:43
shaohe_fengzhenguo: yes. so wo support these 2?07:43
zhenguoshaohe_feng: they are both string07:43
zhenguoshaohe_feng: according to the doc, seems the difference is personality has a path and only admin can access it07:44
zhenguoshaohe_feng: seems it's not high priority thing, wdyt07:44
zhenguoshaohe_feng: if you want to inject something to the instance, you can use inject file, hah07:45
zhenguoshaohe_feng: you would like to support all thing with config drive in one patch?07:46
shaohe_fengzhenguo: I'm doing them in one patch, now.07:46
shaohe_fengzhenguo: I can split it.07:46
zhenguoshaohe_feng: maybe we can just support inject_file at first, config_drive + inject_file, all other things can do it later07:46
shaohe_fengzhenguo: OK. but the nova create API, has no inject_file param, the inject_file is come from "user_data" or "personality"07:48
shaohe_fengzhenguo: so how do we define our api ?07:48
zhenguoshaohe_feng: oh, sorry, if so, maybe user_data first?07:48
zhenguoshaohe_feng: as vm users may be used to this way, I think we should not change, wdyt?07:49
shaohe_fengzhenguo: OK. and we will both support user_data and personality at last.07:49
zhenguoshaohe_feng: inject file will transfer to user_data, right?07:51
zhenguoshaohe_feng: we don't support to specify a inject file throuth the API, and only CLI or UI can use inject_file, right?07:51
shaohe_fengzhenguo: seems  nova CLI transfer file to user_data or personality07:54
zhenguoshaohe_feng: ok, we can follow nova, just support user_data is ok at first step07:54
zhenguoshaohe_feng: and you may need to rebase you code after the renaming, hah07:56
shaohe_fengzhenguo: OK, will do it.07:57
zhenguoshaohe_feng: thanks!07:58
openstackgerritZhenguo Niu proposed openstack/nimble: Add a looping call to wait for power action completed
* zhenguo brb08:00
*** liudong has joined #openstack-mogan08:04
*** zhangjl1 has joined #openstack-mogan08:13
*** zhangjl has quit IRC08:16
zhenguoshaohe_feng: wrt get power state, do you think it's better to cache all instances' pwer state instead of fetching from ironic during every get instance API08:17
*** Kevin_Zheng has quit IRC08:31
*** zhangyang has quit IRC08:31
*** zhenguo has quit IRC08:31
*** openstackgerrit has quit IRC08:31
*** yuntongjin has quit IRC08:31
*** liusheng has quit IRC08:31
*** Jeffrey4l_ has quit IRC08:31
*** kevinz has quit IRC08:31
*** liudong has quit IRC08:31
*** luyao has quit IRC08:31
*** zhangjl1 has quit IRC08:31
*** wanghao_ has quit IRC08:31
*** little has quit IRC08:31
*** shaohe_feng has quit IRC08:31
*** yuntongjin has joined #openstack-mogan08:38
*** kevinz has joined #openstack-mogan08:38
*** zhangjl1 has joined #openstack-mogan08:38
*** liudong has joined #openstack-mogan08:38
*** wanghao_ has joined #openstack-mogan08:38
*** luyao has joined #openstack-mogan08:38
*** Kevin_Zheng has joined #openstack-mogan08:38
*** zhangyang has joined #openstack-mogan08:38
*** liusheng has joined #openstack-mogan08:38
*** zhenguo has joined #openstack-mogan08:38
*** little has joined #openstack-mogan08:38
*** shaohe_feng has joined #openstack-mogan08:38
*** openstackgerrit has joined #openstack-mogan08:38
liushengzhenguo: I can get this message, lol08:38
*** Jeffrey4l_ has joined #openstack-mogan08:38
*** yuntongjin has quit IRC08:38
*** yuntongjin has joined #openstack-mogan08:38
zhenguoam I still alive? seems something wrong with the irc08:40
liushengzhenguo: I can see you lol08:41
zhenguoliusheng: lol08:43
liushengzhenguo: the irc is unstable08:43
zhenguoliusheng: am I still alive?08:48
liushengzhenguo: I can see this message08:48
*** yuntongjin has quit IRC08:50
zhenguoliusheng: seems it's ok now08:53
liushengzhenguo: hah08:54
liushengzhenguo: did you see my messages ?08:54
zhenguoliusheng: something wrong with the irccloud08:54
zhenguoliusheng: yes, it's ok now08:54
zhenguoliusheng: do you remember why ceilometer introduce that?08:55
liushengzhenguo: seems they think it is a matter of course08:57
zhenguoliusheng: hah,08:57
zhenguoliusheng: I will test that later, seems it's ok08:58
liushengzhenguo: someone filed a bp, and sbmit code, the cores review it08:58
zhenguoliusheng: ok08:58
zhenguoliusheng: when using moganclient, I find somthing can improve08:59
liushengzhenguo: thanks, but it is don't need merged quickly, I also need to test it, the functionality is ok, but the tests with pecan seems hard to apply the pipline08:59
zhenguoliusheng: yes, that's my concern09:00
liushengzhenguo: I need to modify the part2 patch09:00
liushengzhenguo: we used  return pecan.testing.load_test_app(self.app_config) to start a app for testing, seems it hard to apply pipline to the testing app09:01
shaohe_fengzhenguo: it's time consume to get the power stat?09:01
liushengzhenguo: what's improvement of moganclient ?09:02
zhenguoliusheng: yes, and seems it's not high priority thing, wdyt?09:02
liushengzhenguo: yes09:02
zhenguoliusheng: we can focus on other tasts which should be done in O09:02
zhenguoliusheng: so how about hold on a while for that?09:02
liushengzhenguo: ok09:02
zhenguoliusheng: thanks09:03
openstackgerritwanghao proposed openstack/nimble: [WIP]Quota support in Mogan(part 1)
zhenguoshaohe_feng: yes09:03
zhenguoshaohe_feng: but a seperate periodic task to fetch all power state also time consuming09:03
zhenguoshaohe_feng, liusheng: but I think keeping user's request performance is more important, wdyt?09:04
liushengzhenguo: why it is time-consuming ?09:05
zhenguoliusheng: you mean for API request?09:06
*** zhangjl1 has quit IRC09:06
zhenguoliusheng: currently if you get instance info, it will call ironic to get power_state.09:06
liushengzhenguo: you are talking about geting power state ?09:06
zhenguoliusheng: yes09:06
*** wanghao_ has quit IRC09:06
*** wanghao has joined #openstack-mogan09:07
liushengzhenguo: why it is time consuming to call ironic to get power_state ?09:07
liushengzhenguo: just query db of ironic ?09:08
zhenguoliusheng: need a REST API call to ironic09:08
liushengzhenguo: oh09:08
liushengzhenguo: how about record power_state in mogan's db ?09:09
zhenguoliusheng: it's another option, but need to sync with ironic as well09:09
zhenguoliusheng: but I'm OK with save it to db09:09
liushengzhenguo: just treat ironic as a libvirt, lol09:10
openstackgerritZhenguo Niu proposed openstack/nimble: Add a looping call to wait for power action completed
zhenguoliusheng: hah, yes09:10
zhenguoliusheng: and for set power state, I think we should  change our CLI to 'openstack baremetal server reboot/on/off '09:10
zhenguoliusheng: wdyt?09:10
liushengzhenguo: reboot is ok, but on/off looks weird09:11
zhenguoliusheng: I find it's not convient to use openstack baremtal server power --power-state xx server09:11
liushengzhenguo: ok09:11
zhenguoliusheng: let me check what's nova used09:11
liushengzhenguo: start/stop in Nova09:11
*** liudong has quit IRC09:11
zhenguoliusheng: that's more better, hah09:12
liushengzhenguo: we just stole it ? :P09:12
zhenguoliusheng: and how about change to use 'server boot' instead of 'server create'09:12
zhenguoliusheng: yes09:12
liushengzhenguo: yes, I am ok09:12
zhenguoliusheng: as openstack users already been used that way09:13
liushengzhenguo: ok, will try to change it09:13
zhenguoliusheng: big thanks09:14
shaohe_fengzhenguo: openstack baremetal server reboot/shutdown/poweron?09:15
zhenguoshaohe_feng: reboot/start/stop like nova, hah09:15
liushengzhenguo: another problem I found when I debuging tempest, currently when we deleting a server, the status is 'deleting' but the server is deleted soon while it is still in doing some clenup in ironic09:15
liushengzhenguo: we may need to wait the cleanup completed then delete the server09:15
zhenguoliusheng: yes, we need to wait for that09:16
zhenguoliusheng: oh, now09:16
zhenguoliusheng: when it's going to a deleting state we can just change instance deleted09:16
zhenguoliusheng: we don't need to wait ironic, as the resouce cleanup it's not related to mogan09:17
liushengzhenguo: yes, we need to wait the cleanup in rionic completed and then marked the instance deleted09:17
zhenguoliusheng: sorry, I mean no09:17
liushengzhenguo: that may cause many issues09:17
zhenguoliusheng: why09:18
shaohe_fengzhenguo: start/stop is OK.09:18
liushengzhenguo: we cannot query the instance from mogan, and we asume it has free the node in ironic, but actually it don't09:18
zhenguoliusheng: actually we use a wait unprovision state now09:19
liushengzhenguo: e.g. we have 3 nodes in ironic, if we have created 3 servers and then deleted them, and then we create the 4th server, but it will fail09:19
zhenguoliusheng: if the node is going to a status indicates that its cleaning, we should not wait09:19
zhenguoliusheng: yes, that's ok09:20
zhenguoliusheng: as deleting ironic node will take long time09:20
zhenguoliusheng: especially in real evn, as we have to erase disks09:20
liushengzhenguo: so it effect anything ?09:20
zhenguoliusheng: that may take many hours09:20
zhenguoliusheng: users should not wait for that,09:21
liushengzhenguo: users don't need t wait09:21
liushengzhenguo: why need to wait ?09:21
zhenguoliusheng: hah09:21
liushengzhenguo: the deletion is a cast call, we just set the status of the instance in deleting when ironic doing cleanup09:22
zhenguoliusheng: we set it to deleting if the cast call sent successfully09:23
liushengzhenguo: yes, I mean the "deleting" should be continued until ironic complete cleanup09:23
zhenguoliusheng: yes, but I don't think so09:24
liushengzhenguo: so how to solve the problem I said09:24
zhenguoliusheng: if you are a end user, when you delete your instance, it will finished in about 8 hours, how would you feel09:24
zhenguoliusheng: you mean node cleanup failed?09:24
zhenguoliusheng: currently when node goes to NOSTATE,CLEANING,CLEANWAIT,CLEANFAIL,AVAILABLE, we will set instance to be deleted09:25
liushengzhenguo: no, if you have 3 nodes, and you created 3 servers on them, then you delete them all, you do "sever list", it return empty, but you cannot create the 4th server09:25
zhenguoliusheng: yes, that's OK, as you don't have resources which can be used09:26
zhenguoliusheng: we are cloud, no need to care about the resouces used up09:26
zhenguoliusheng: and users should not know how many nodes we have09:27
liushengzhenguo: hmm... addionally, we cannot get anything about the instance after deleted from mogan api, but actually, it may still exist in ironic09:27
zhenguoliusheng: no, the node has dissociated with the instance09:28
liushengzhenguo: but that may cause issues if we import quota09:28
*** wanghao has quit IRC09:28
zhenguoliusheng: no, the node is already not related to the node09:28
liushengzhenguo: oh, I see09:29
zhenguoliusheng: hah, we can jsut assue ironic is a black box09:30
liushengzhenguo: you mean the instance info has been cleaned from the node ?09:30
zhenguoliusheng: exactly09:30
zhenguoliusheng: yes09:30
zhenguoliusheng: it's already not related to the instance, why we still need to wait09:30
liushengzhenguo: oh, may I was wrong, I thought it still releted to the instance09:32
zhenguoliusheng: hah,09:32
liushengzhenguo: I will test more, hah09:33
zhenguoliusheng: thanks09:34
zhenguoliusheng: do you familiar with partition in nova?09:34
liushengzhenguo: no :(09:34
zhenguoliusheng: hah09:34
liushengzhenguo: may you can ping Kevin_Zheng09:35
liushengzhenguo: lol09:35
liushenghave a nice weekend :)09:35
zhenguoliusheng: thanks09:36
zhenguolittle: hi09:46
littleI am away this aternoon09:46
zhenguolittle: no problem, hah09:46
littleI think user_data and personility should be considered a feature to develop09:47
littlesometimes, inject files is very useful09:48
littlego home, hah09:51
zhenguolittle: yes, we should first support user data09:51
zhenguolittle: have a nice weekend:)09:52
zhenguoseems you guys can always go home on time after leaving huawei, lol09:55
zhenguoshaohe_feng: are you still around?09:56
zhenguoshaohe_feng: do you remember this ?09:56
littlehah, leaving huawei is very cool, bye09:57
zhenguohah, bye10:00
* zhenguo disappeared10:00
*** little has quit IRC10:03
openstackgerritliusheng proposed openstack/nimble: Versioned notification: Add suport for instance.create.start (WIP)
*** kevinz has quit IRC10:13
*** Jeffrey4l_ has quit IRC11:14
*** Jeffrey4l_ has joined #openstack-mogan11:14
*** kevinz has joined #openstack-mogan13:33
*** kevinz has quit IRC13:51
*** liusheng has quit IRC14:47
*** Jeffrey4l_ has quit IRC18:39
*** Jeffrey4l_ has joined #openstack-mogan18:39
*** Jeffrey4l_ has quit IRC21:35
*** Jeffrey4l_ has joined #openstack-mogan21:47

Generated by 2.14.0 by Marius Gedminas - find it at!