Wednesday, 2017-01-11

*** zhangyang has quit IRC00:11
*** wanghao has joined #openstack-mogan00:43
*** little has joined #openstack-mogan00:53
*** yuntongjin has joined #openstack-mogan01:25
*** kevinz has joined #openstack-mogan01:28
*** zhangjl has joined #openstack-mogan01:47
zhenguomorning mogan01:50
zhenguoGerrit downtime on Thursday 2017-01-12 at 20:00 UTC, we will finish renaming then, lol01:51
zhangjlgood news01:51
*** zhangyang has joined #openstack-mogan02:07
* zhenguo away for lunch03:56
zhenguohi all, do you know what's the difference between state and status?06:14
zhenguoliusheng: hi, after a second this maybe it's better to introduce a state machine for instance state, leveraging check_instance_state like nova will be not that clear, and nova also plan to introduce sm, as we are new we can do that at beginning06:23
zhenguosecond think06:23
liushengzhenguo: yes, agree, not everything of Nova is reasonalbe06:24
zhenguoliusheng: only a state machine for instance sate, which can control states transitions. power_state and task_state just show the status06:24
zhenguoliusheng: hah06:24
zhenguoliusheng: I still a bit confused about status and state06:24
liushengzhenguo: just need to ping Kevin_Zheng06:25
liushengzhenguo: he is not in workplace now06:25
zhenguoliusheng: ok06:26
zhenguoliusheng: ironic has a provision_state, target_provision_state, power_state, target_power_state, just these06:27
liushengzhenguo: so many..06:27
liushengzhenguo: lol06:27
zhenguoliusheng: indeed, hah06:27
zhenguoliusheng: we will have a power_state, task_state and state06:28
zhenguoliusheng: task_state will include the ing status of power_state and state like powering_on, spawning.06:29
liushengzhenguo: I am thinking about how about merged tast_state and state ?06:29
zhenguoliusheng: you mean only have a state?06:30
liushengzhenguo: state will includes middle state and fixed state06:30
zhenguoliusheng: yes06:30
liushengzhenguo: yes, just a primitive thought06:31
liushengzhenguo: the state can be active, stopped, booting, stopping06:31
zhenguoliusheng: seems ok, as I find some duplicate with task_state and state06:31
Kevin_Zhengwhats up06:31
zhenguowe are discussing about task_state and state06:32
liushengKevin_Zheng: we need a Nova expert now, lol06:32
Kevin_Zhengone sec06:32
zhenguoliusheng: seems it's more clear to just using one state, hah06:36
zhenguoliusheng: we can use a state maching to control the states transition06:37
Kevin_Zhengyou can check this out06:37
Kevin_Zhengit is a mapping between vm_state+task_state and Status06:38
liushengzhenguo: yes there isn't overlap between persistent state and processing state06:38
zhenguotoo messy06:38
zhenguoKevin_Zheng: we plan to just use one instance state, wdyt?06:39
liushengzhenguo: may we need to designe a map firstly to show state transition06:40
zhenguoliusheng: yes, that could be done with the state machine defination06:40
zhenguowe don't need to show that specific states like nova task_state06:41
liushenge.g. when a Nova instance is in 'building' state, the task state maybe in scheduling, networking, spawing. but seems building is easy to understood by end users.06:45
Kevin_ZhengI think it is ok, because for baremetal instances06:48
Kevin_Zhengpower_state means status?06:48
zhenguopower_state will be power on or power off06:51
zhenguoliusheng: yes, and nova task_state always just present in 1 sec or less06:52
liushengzhenguo: yes06:54
zhenguoand when building instance it's really rare to see all task states, so I doubt whether we need that, and we have a task flow now06:54
zhenguotask flow can be useful to track specific task states transition06:55
liushengzhenguo: we need to design all the state transitio to see how many state/tast_state we will have06:58
zhenguoliusheng: ok, will draft a patch soon06:58
liushengzhenguo: coool!06:59
zhenguoliusheng: within a state machine, hah06:59
*** yuntongjin has quit IRC07:20
*** yuntongjin has joined #openstack-mogan07:20
*** kevinz has quit IRC07:40
*** kevinz has joined #openstack-mogan07:41
zhenguoliusheng: after repo renamed, seems we need to upload all inprogress patches to the new repo07:46
liushengzhenguo: ok, the gerrit won't handle the patches automatically ?07:47
zhenguoliusheng: not sure, seems they can, according to the email Existing reviews, project watches, etc, for these projects will all be carried over.07:48
zhenguoliusheng: they will do this rename tomorrow, hah07:49
liushengzhenguo: so let's wait!07:49
zhenguoliusheng: hah07:49
zhenguoliusheng: powering on will go to active state and powering of will be stopped, right?07:51
liushengzhenguo: seems not exactly,07:53
zhenguoliusheng: I find in nova active means vm is running07:54
liushengzhenguo: I remember there is other state after instance end powering off07:56
zhenguoliusheng: I find a STOPPED state for vm07:57
zhenguoliusheng: maybe we should also introduce a maintenance state, which will block other state transiation07:58
liushengzhenguo: you mean two set of states ?08:00
zhenguoliusheng: no, just one staet08:00
liushengzhenguo: which one state ?08:00
zhenguoliusheng: just instance state, I mean it can be set to maintenance08:01
liushengzhenguo: oh, I see, you mean like ironic08:01
zhenguoliusheng: yes, as baremtal is not like vm, sometimes we need to do some maintenance08:01
*** yuntongjin has quit IRC08:02
littleIn nova, the status is the combination of vm_state and task_state, you can read the _STATE_MAP defination in nova :\nova\api\openstack\common.py08:18
zhenguolittle: thanks, let me check08:19
littleI think the status is readable for human08:21
zhenguolittle: you mean we should keep the current status field?08:21
littleyes, I think we can learn from nova08:24
zhenguolittle: seems OK08:24
zhenguoliusheng: so we don't need to replace status with state, wdyt?08:24
zhenguoas nova will transfer it to status in the response body08:25
zhenguonova's state management seems too complex08:25
littleyes, a bit08:25
littlewe can simple it08:26
littlestatus and task_state08:26
zhenguolittle: we plan to get rid of task_state08:26
zhenguolittle: only leave one state/status field08:27
littlewhat is the defference of state and status?08:27
zhenguoas status is a combination of vm_state and task_state, so we can just use a status08:27
zhenguonot sure the difference08:28
zhenguobut we can keep to use status08:28
zhenguofollow nova way08:28
littleyes, we no need to use combination of vm_state and task_state, just use status08:29
littlethis is simple08:29
zhenguoyes, and it's clear08:29
zhenguowe will introduce a state machine to control the states transiation as well08:29
zhenguoliusheng, shaohe_feng, wanghao, zhangjl: wdyt?08:30
zhangjlzhenguo: IMO, same as nova maybe is better08:31
*** l4yerffeJ has joined #openstack-mogan08:31
zhenguozhangjl: but nova's seems not design that well08:31
zhenguozhangjl: and they also have a plan to introduce state machine08:32
zhenguozhangjl: I will sent a draft patch by today, you can help review it08:33
zhangjlmaybe, this is helpful #link:
zhenguozhangjl: yes, will borrow the fsm from ironic, hah08:35
zhangjlzhenguo:that`s good08:36
wanghaozhenguo: agree with keeping only one status in mogan, since we can change the status to show what task_state it is.08:36
zhenguowanghao: yes, and we have a taskflow to track the specific states transitions in log08:37
zhenguozhangjl: btw, for the api doc, we plan to setup a http server ourselves.08:37
zhangjlzhenguo:good news08:38
zhenguozhangjl: hah08:38
wanghaozhenguo:  yes,  and also support to introduce the maintance status as god sheng said, that's helpful to baremetal host.08:39
zhenguowanghao: ok, thanks08:39
wanghaozhenguo: Cinder also has this status FYI.08:39
littlezhenguo: where is the spec docs?08:39
*** l4yerffeJ has quit IRC08:40
zhenguolittle: currently not, after renaming tomorrow, I will initialize the mogan-specs08:40
littlezhenguo: good08:40
zhenguowanghao: you mean volume has a maintenance status?08:40
*** l4yerffeJ has joined #openstack-mogan08:40
wanghaozhenguo: yes08:42
zhenguowanghao: hah, ok will add that.08:43
liushengwanghao: I am not god sheng, lol08:43
wanghaoliusheng: yes, you're08:44
zhenguowanghao: agree08:44
wanghaozhenguo:  In migrate_volume, user can lock volume, that means user can't operate this volume in mirgrating process, the volume will be maintenance.08:44
* liusheng have to escape08:44
littleI just mistook 'god sheng' to 'gou sheng', sorry08:47
liushenglittle: :(08:48
zhenguowanghao: you mean when users lock the volume it will go to maintenance state?08:48
zhenguowanghao: we should also support lock instance08:48
zhenguoI find nova also has a option to disable instance terminate08:49
zhenguolittle: lol08:50
liushengzhenguo: which one ?08:50
zhenguoliusheng: instance has a filed named disable_terminate08:51
zhenguoliusheng: it will first check this when deleting instance08:51
zhangjlzhenguo:shall we lock instance to disable terminate instead of add a filed?08:52
wanghaozhenguo: yes, in cinder, when user uses migrate operation, he has a option choice to lock volume.08:53
liushengzhenguo: Kevin told me seems the attribute is useless..08:53
zhenguozhangjl: need to dig08:53
zhenguowanghao: ok, thanks08:53
liushengzhenguo: it is always False, and not way to set it to true08:53
liushengs/not way/no way08:53
zhenguoliusheng: hah, if so we can just lock the instance to disable terminate08:53
zhenguoliusheng: and it's more generic08:53
wanghaozhenguo:  for mogan, I agree it's a lock API to change instance to maintance.08:54
zhenguozhangjl: yes, seems that's better08:54
zhenguowanghao: ok08:54
wanghaozhenguo: Like vmware, in Vcenter,  user can set host to maintance.08:54
zhenguozhangjl: if you are interested in locking instance, maybe you can take this task, hah08:54
zhangjlzhenguo:let me try08:55
liushengzhenguo: vmware expert!08:55
zhenguowanghao: yes, we should support maintenance08:55
liushengzhangjl: wanghao08:55
zhenguozhangjl: thanks08:55
zhenguohah, ok so things become clear, we will keep to use status and not introduce task_state, need to add a state machine to control states transitions08:58
*** l4yerffeJ has quit IRC09:01
*** l4yerffeJ has joined #openstack-mogan09:01
zhenguozhangjl: for locking instance, you can first adding a new task here , and maybe register a bp for that09:07
zhenguozhangjl: thanks09:08
*** wanghao has quit IRC09:28
*** l4yerffeJ has quit IRC09:47
*** l4yerffeJ has joined #openstack-mogan09:47
*** l4yerffeJ has quit IRC10:08
*** l4yerffeJ has joined #openstack-mogan10:08
*** zhangjl has quit IRC10:13
*** little has quit IRC10:16
*** kevinz has quit IRC10:27
zhenguoseems cinder doesn't have a state_file, just using hard code for status11:32
*** kevinz has joined #openstack-mogan12:06
*** yuntongjin has joined #openstack-mogan12:31
*** kevinz has quit IRC12:31
*** kevinz has joined #openstack-mogan12:32
*** _liusheng has joined #openstack-mogan12:38
*** kevinz has quit IRC12:49
*** kevinz has joined #openstack-mogan12:50
Kevin_ZhengStill working on states13:03
*** little has joined #openstack-mogan13:25
*** little has quit IRC13:42
*** little has joined #openstack-mogan13:43
*** Jeffrey4l_ has quit IRC13:45
*** little has quit IRC13:47
*** yuntongjin has quit IRC13:51
*** little has joined #openstack-mogan13:55
*** little has quit IRC14:11
*** little has joined #openstack-mogan14:12
*** little has quit IRC14:16
*** kevinz has quit IRC14:31
*** _liusheng has quit IRC14:47
*** little has joined #openstack-mogan14:49
*** little has quit IRC14:54
*** little has joined #openstack-mogan14:56
*** little has quit IRC15:13
*** little has joined #openstack-mogan15:13
*** little has quit IRC15:18
*** Jeffrey4l has joined #openstack-mogan15:40
*** little has joined #openstack-mogan15:54
*** little has quit IRC16:11
*** little has joined #openstack-mogan16:11
*** little has quit IRC16:15
openstackgerritZhenguo Niu proposed openstack/nimble: Introduce a state machine for instance status
*** little has joined #openstack-mogan16:46
*** little has quit IRC16:51
*** little has joined #openstack-mogan16:55
*** little has quit IRC17:12
*** wxy has quit IRC17:14
*** wxy has joined #openstack-mogan19:12
*** l4yerffeJ has quit IRC20:08
*** l4yerffeJ has joined #openstack-mogan20:09
*** l4yerffeJ_ has joined #openstack-mogan21:34
*** Jeffrey4l_ has joined #openstack-mogan21:34
*** l4yerffeJ has quit IRC21:35
*** Jeffrey4l has quit IRC21:35

Generated by 2.14.0 by Marius Gedminas - find it at!