Wednesday, 2017-06-14

*** wanghao has joined #openstack-mogan00:24
*** litao__ has joined #openstack-mogan00:52
*** litao__ has quit IRC00:53
*** litao__ has joined #openstack-mogan00:53
zhenguomorning mogan00:54
litao__morning mogan00:55
liushengzhenguo: the patch to fix the issue, sorry for my fault00:55
zhenguoliusheng: hah, thanks00:56
litao__zhenguo: I want to add limit and sort functions in list operations, wdyt?00:56
zhenguolitao__: sure, please go ahead00:57
litao__in BP or Bug?00:57
zhenguolitao__: bug is enough00:57
zhenguolitao__: not conflict with Xinran's patch to add filters, right?00:58
litao__maybe  a little, but i  will rebase it00:58
liushengzhenguo, litao__ should we need to consider searchlight ?00:59
zhenguoliusheng: if so, it worth a BP :D01:00
zhenguoliusheng, litao__: but not sure if it's easy to integrate with SL, need to add more notifications?01:00
liushengzhenguo: we don't need to implement filter and pagination/sort  functionalities01:00
liushengzhenguo: yes01:00
liushengzhenguo: if we consider SL01:01
zhenguoliusheng, litao__: if there's no block for us to integrate with SL, I think that's the right approach01:01
* litao__ zhenguo: so you suggest to use SL?01:02
liushengNova have decided to avoid adding more pagination/sort functions01:02
litao__ so you suggest to use SL?01:02
zhenguolitao__, liusheng: seems that's the right approach01:03
zhenguolitao__: we got a SL core here, you can ping Kevin_Zheng, if you want to know more about it01:04
litao__zhenguo: ok, thanks, i will ping him for more details if needed01:04
zhenguolitao__: ok01:04
liushengalso need to let Xinran know this01:05
litao__so i will summit a BP firstly.01:05
zhenguolitao__: ok01:05
liushenglitao__: maybe need a bp in SL, not in Mogan01:05
liushenglitao__: Mogan just need to add more notification implementations01:05
zhenguoliusheng, litao__: let's wait for Kevin_zheng01:06
litao__zhenguo: OK01:07
zhangyangzhenguo: sorry the last day i was off, i think we do not have requirement for port type now.01:09
zhenguozhangyang: ok, thanks for the feedback01:09
zhenguohi all, I plan to drop port type support provisionally, wdyt?01:10
openstackgerritZhenguo Niu proposed openstack/mogan master: Change to use server metadata instead of extra
Kevin_Zhengcannot write plugin in SL for non-official project01:19
Kevin_Zhengbut we can do out-of-tree first01:19
*** liujiong has joined #openstack-mogan01:20
zhenguoKevin_Zheng: you mean add a SL plugin in mogan repo?01:20
zhenguois there some sample for us01:20
liushengis the requirement urgent ?01:21
Kevin_Zhengno you could add some private file and a conf01:21
Kevin_Zhengall projects in sl are plugins01:21
zhenguooh, if so, we can wait to be official, no rush01:21
Kevin_Zhengjust files, you add your file and enable it01:21
liushengKevin_Zheng: does plugin loaded by stevedore entry points ?01:22
litao__Kevin_Zheng:  Add files in mogan or SL?01:22
zhenguolitao__: SL will not accept non-offical plugin, so it shouldbe in mogan01:23
liushengzhenguo: maybe not in Mogan, in a independent repo01:23
zhenguoliusheng: that's too complex01:23
liushengzhenguo: not very complex01:23
liushengzhenguo: there is a tool to generate basic frame01:24
zhenguoliusheng: you need to add a repo, just for a Sl plugin?01:24
liushengzhenguo: repo can put in github temporary01:24
litao__liusheng: need more workload?01:25
zhenguoliusheng: why not just place it in mogan repo like other devstack plugins01:25
zhenguoliusheng: then we copy the files to SL and run when setup devstack env01:26
liushengzhenguo: just personal thought but, the plugin will be put into SL, not in Mogan, not similar with Mogan01:26
liushengzhenguo: it is very easy to put it in independent repo01:26
liushengzhenguo: I am ok to put it in Mogan tree01:27
zhenguoliusheng: I know it's easy, but should that be under openstack namespace01:27
zhenguoliusheng: if not, I think it's a bit odd01:27
liushengzhenguo: lol, I may think it in Mogan is  a bit odd :D01:28
zhenguoliusheng: and I'm not quite sure if we should integrate with SL if it's not accept non-official project now01:28
zhenguoliusheng: maybe integrate with Sl is more odd01:28
liushengso I just asked is this requirement urgent ?01:29
zhenguoliusheng: not sure, if HFBank want to run it in production, maybe list with pagination is required?01:30
zhenguozhangyang: wdyt?01:30
litao__zhenguo: I think it is necessary01:30
litao__zhenguo: in production01:30
zhenguomaybe we can just do it ourselves if it's necessary01:31
zhenguountil we are afficial, wdyt?01:31
litao__you mean we don't do it currently?01:33
zhenguolitao__: no, I mean we don't need SL currently01:33
zhenguolitao__: just do as your fist proposal01:33
litao__yes , maybe01:35
litao__zhenguo: so i will do as traditional method01:36
zhenguolitao__: yes,01:36
zhenguoliusheng: wdyt?01:36
liushengzhenguo: I may prefer we put effort on other things if the requirement is not urgent :(01:37
zhenguoliusheng: so, let's wait zhangyang or liudong 's feedback to see01:38
litao__zhenguo: yes01:38
liushengwe have many things need to do, which help to become official project01:39
zhenguoyes, but not everyone full involved in01:39
zhenguoI think it's hard to control01:40
zhenguoas you may want to do more things, but there's no task, as all task has owners01:40
zhenguoIf this happening, please feel free to take over anyone from me, lol01:41
liushengtests, review, docs, every things make Mogan more like a official project is urgent, hah01:42
zhenguoliusheng: sure01:42
zhenguoeveryone can find more things to do here01:42
liushengzhenguo: do we have the required items/problems which required to become official project we disscussed ?01:45
zhenguoliusheng: no,01:45
liushengzhenguo: maybe it is better also to paste it in the etherpad01:45
zhenguoliusheng: not hard requirement,01:46
liushengzhenguo: last meeting, I remember you have mentioned about 10+ items01:46
zhenguoliusheng: but one big thing is the community don't know what Mogan is, lol01:46
liushengzhenguo: hah01:46
zhenguoliusheng: that's all help to make others know what we are01:47
zhenguoliusheng: docs, mails, presentations...01:47
liushengzhenguo: yes, but more than these, we have other things to do help01:48
zhenguoliusheng: and we must follow the 4 opens01:48
liushengzhenguo: how aobut sending a meeting summary to ML after weekly meeting ?01:48
zhenguowe need to discuss in public more01:48
zhenguoliusheng: yes, wanghao will do that01:49
zhenguoliusheng: not meeting summary, just present our task update01:49
zhenguoto let the community aware what we are doing01:49
liushengzhenguo: oh, thanks wanghao, I remember that01:49
liushengzhenguo: yes, really need that01:49
zhenguoliusheng: we will send it weekly after meeting, so please remember to update the ehterpad01:50
liushengzhenguo: ok01:50
zhenguoliusheng: As we are all in a same TZ, I would like to have a 'review hour' every day, wdyt?01:52
liushengzhenguo: hah, sounds good01:53
zhenguoliusheng: at least core must attend, and as there are only 3 cores currently, if one is absent, we can't land patch01:53
zhenguoliusheng: need to get more cores01:53
zhenguoliusheng: will discuss during tommorow's meeting01:54
liushengzhenguo: yes, sure01:54
* zhenguo brb01:58
litao__: I summited a bug :, maybe we can add unit tests in spare time02:00
openstackLaunchpad bug 1692767 in devstack "[devstack] keystone service fails to start after reboot" [Undecided,In progress] - Assigned to Kirill Zaitsev (kzaitsev)02:00
* litao__ sorry, the true link is
openstackLaunchpad bug 1697813 in Mogan "Add unit tests for mgaon" [Undecided,New]02:02
wanghaolitao__: cool02:08
wanghaozhenguo: ping02:08
wanghaozhenguo: I see your comments in manage node spec.  For image, we can get it directly from driver,  but for flavor I think user should specify it since it's a resource in Mogan.02:10
openstackgerritMerged openstack/mogan master: Tempest: Add tests for keypairs functionalities
zhenguowanghao: but how can users know the resource class that the node should be?02:12
zhenguowanghao: only operators know that, so i think operators need to prepare that, they need to set resource class to the node first, then that node can be managed by mogan02:12
zhenguolitao__: thanks02:13
wanghaozhenguo: yes,  for Ironic,  that should be in the process of adopting node in Ironic.02:13
wanghaozhenguo: and operators need to create a flavor in Mogan,  and user use it to manage node.02:14
wanghaozhenguo: or your idea is Mogan will create the flavor automatically when adopting the node02:14
zhenguowanghao: yes, ironic can adopt any nodes, but for mogan, we can keep the process simple02:14
wanghaozhenguo: so in my understand,  your way is that user didn't need to care the flavor or image.02:15
zhenguowanghao: we can only care about what nodes we can manage, if the node resources not associate with a mogan flavor, we can't adopt it.02:15
zhenguowanghao: yes, I would perfer operators prepare all things before02:16
zhenguowanghao: but for images, maybe we don't care, if ironic can't provide that information, we can adop it with the server image field empty02:16
wanghaozhenguo: so that means Mogan will check if the existing flavor match the resource class, right?02:17
wanghaozhenguo: yes,  that could be02:17
zhenguowanghao: yes, they can include in the manageable node list, but when we do adopt action, we need to check the flavor02:18
wanghaozhenguo: ok I see,  so the rules for manageable node are three: 1. without instance_uuid in Ironic, 2 active stats 3, have resource class02:19
zhenguowanghao: I think so, need to see others' opinion on that part :D02:19
wanghaozhenguo: cool,02:20
zhenguowanghao: hah02:20
wanghaozhenguo: basically I think it's fine,  since that depend on driver's implemention.02:20
wanghaozhenguo: you can implement it as you want.02:20
zhenguowanghao: yes, every driver want to support adopting, they need to provide such things02:21
wanghaozhenguo: yes,  We can just throw a example in Spec,  in fact, I think this spec don't care how driver implement it.02:22
zhenguowanghao: yes,02:22
litao__zhenguo: Check flavor is in order to do other  operations such as rebuilding after adoption?02:23
wanghaozhenguo: so the rules will be more flexible for drivrs.02:23
zhenguolitao__: we need a flavor to show the server detail hardware profile02:24
wanghaolitao__: maybe,  but I think more important thing is to match the service in Mogan and node in Ironic.02:24
zhenguowanghao: yes, but they must adhere to the rules mogan defined :D02:25
wanghaozhenguo: of course02:25
wanghaozhenguo: not every node could be managed by Mogan.02:26
zhenguowanghao: yes, we only mange nodes fully prepared by operators02:26
wanghaozhenguo: right02:27
zhenguowanghao: we need to add a docs for operators later02:27
wanghaozhenguo: exactly02:32
wanghaozhenguo: like Ironic adopt process02:32
zhenguowanghao: yes02:32
wanghaozhenguo: do you think in adopting API,  flavor_uuid is needed?02:36
openstackgerritZhenguo Niu proposed openstack/mogan master: Remove unneeded *encoding: utf-8*
zhenguowanghao: no, it's not needed02:36
openstackgerritliusheng proposed openstack/mogan master: Ensure the servers deleted completely in tempest's resource_cleanup
openstackgerritliusheng proposed openstack/mogan master: Tempest: add tests for server states interfaces
wanghaozhenguo: since I think is there a problem that may there are many flavors have same resource class02:37
zhenguowanghao: aha, that's a problem02:37
wanghaozhenguo: yes, you know that Mogan can check the resource class match the flavor when getting the manageable nodes.02:38
zhenguowanghao: let's think more about whether a flavor should associated more than one resoure class02:38
wanghaozhenguo: but if there are many flavors matched,  how Mogan to choose one?02:39
zhenguowanghao: I think we should limit it as one flavor should be associated with one resouce class02:39
zhenguowanghao: will update my spec02:39
wanghaozhenguo: but consider this case,  same resource class, different extra_specs?02:40
zhenguowanghao: oh, good point02:40
wanghaozhenguo: since resource class is just a string that operator set it.02:40
wanghaozhenguo: it's very possible.02:41
zhenguowanghao: yes, I understand02:41
zhenguowanghao: so please add flavor_uuid back to adopting API :D02:41
wanghaozhenguo: emm,  okay02:41
zhenguowanghao: manageable nodes only provide resouce class instead of flavor uuid02:42
wanghaozhenguo: yes02:42
zhenguowanghao: users need to choose one flavor with the resouces class, that's the right approach, thanks02:42
wanghaozhenguo: so we need user to decide which flavor you want to use.02:42
openstackgerritliusheng proposed openstack/mogan master: Tempest: Add test for node listing api
zhenguowanghao: ok02:42
zhenguolitao__: hi, I updated this according your comments02:45
openstackgerritwanghao proposed openstack/mogan-specs master: Manage existing nodes in Mogan
litao__zhenguo: great02:50
* zhenguo brb03:09
openstackgerritliusheng proposed openstack/mogan master: Tempest: add tests for server states interfaces
openstackgerritliusheng proposed openstack/mogan master: Rename the rpc_server/rpc_flavor variable with db_server/db_flavor
openstackgerritZhenguo Niu proposed openstack/mogan master: Remove unneeded *encoding: utf-8*
shaohe_fengzhenguo: I'm coming back04:32
shaohe_fengzhenguo: from Shanghai.04:32
zhenguoshaohe_feng: welcome back!04:47
openstackgerritZhenguo Niu proposed openstack/mogan master: Following up patch of 6673dbb88fd35d8491248fc94bdc6365d352703a
shaohe_fengzhenguo: thanks.05:08
liushengzhenguo: same error occured in  rebuilding tempest test cases06:21
zhenguoliusheng: you can run it successfully in you local env?06:22
liushengzhenguo: yes06:22
liushengzhenguo: Jun 14 03:45:12.825925 ubuntu-xenial-osic-cloud1-s3700-9301861 ironic-conductor[11001]: ERROR ironic.drivers.modules.agent [-] node c471693e-ed12-4ce2-b2ba-b396fa40fb4e command status errored: {u'message': u'Command execution failed: Unable to stat device /dev/vda2 after attempting to verify 3 times.', u'code': 500, u'type': u'CommandExecutionError', u'details': u'Unable to stat device /dev/vda2 after attempting to verify 3 times.'}06:23
liushengzhenguo: werid error06:23
zhenguoliusheng: I saw this error on my patch yesterday, but recheck fixed it06:25
liushengzhenguo: really, I only found the error in my patch06:25
liushengzhenguo: which one ?06:25
zhenguoliusheng: aha, sorry, it's your patch log :D06:32
*** shaohe_feng has quit IRC07:08
*** shaohe_feng has joined #openstack-mogan07:15
zhenguoliusheng: the purpose of flavor.disabled
liushengzhenguo: thanks for the reference, let me take a look07:53
liushengzhenguo: got it, so maybe we should also add the limitation that dont' allow delete flavor there is server using it ?07:57
zhenguoliusheng: yes, I think so, in this senario, they should disable the flavor07:58
liushengzhenguo: just asked zhenyu, Nova don't have the api yet, not sure is there any plan08:00
zhenguoliusheng: hah, we can do first08:00
liushengzhenguo: hah sure08:01
zhenguoliusheng: we provide the ability to disable flavor, then don't allow users delete flavor if it's used, seems the right approach08:02
liushengzhenguo: not sure, if we need to temporarily skip the rebuilding test case and file a bug, wdyt ?08:02
zhenguoliusheng: yes, but you can keep the patch open08:03
liushengzhenguo: dont' allow users delete the flavor may shouldn't depend on the disable/enable08:03
zhenguoliusheng: and for rebuilding, the currently implementation is simple, we need to allow specify image and configdrive when rebuilding08:03
zhenguoliusheng: you mean check disabled first before deleting?08:04
liushengzhenguo: yes, I just wondered why we dont' support specifying a new image in rebuilding lol08:04
zhenguoliusheng: hah08:05
zhenguoliusheng: that need to be improved08:05
liushengzhenguo: no, maybe when a flavor still in use by a server, we shouldn't delete it. but we can disable it to forbiden being use by new server08:06
liushengzhenguo: as your reference described, wdyt ?08:06
zhenguoliusheng: for flavors, maybe we can introduce a new 'reference' field, when claiming a server with that, the reference++ and when deleting server we make reference--, and check if reference before deleting flavor08:06
zhenguoliusheng: you mean automatically disable the server if deleting not allowed08:07
liushengzhenguo: or filter query servers by flavor id before deleting flavor?08:07
liushengzhenguo: no disable mannually08:07
zhenguoliusheng: query servers may time consuming08:08
liushengzhenguo: but make logical simple08:08
zhenguoliusheng: how about the 'reference' field proposal08:08
liushengzhenguo: hmm, I am not sure if we will support "resize". hah08:09
zhenguoliusheng: hah, sure no resize support, but maybe valence can make it possible, not sure08:09
liushengzhenguo: sure ? the "resize" may not only relate the cpu, ram, disks08:10
zhenguoliusheng: not quite sure08:11
liushengzhenguo: oh, but seems we cannot support, we also cannot support migration, hah08:11
zhenguoliusheng: after boot from volume supported, maybe we can support migration08:11
zhenguoliusheng: for diskless server08:12
liushengzhenguo: hah, think too much08:12
zhenguoliusheng: hah08:12
zhenguoliusheng: ironic will support boot from volume soon...08:13
liushengzhenguo: according to the referece, seems disable operation performed manually08:13
liushengzhenguo: hope that08:13
zhenguoliusheng: yes, we need admins disable the server themselves08:14
zhenguoliusheng: and if they issue a delete request for a server which used, then just return error08:14
liushengzhenguo: is this wrong ?08:15
zhenguoliusheng: yes, notallowed, hah08:16
zhenguoliusheng: return forbbiden08:16
liushengzhenguo: oh, I see08:16
zhenguoliusheng: with error, the flavor is been using by some servers08:16
zhenguoliusheng: and please take a look at this which gets rid of the extra specs table/object08:18
zhenguoliusheng: for all such things we just use one PATCH method, which make things more clear08:18
liushengzhenguo: yeah, I am looking08:18
zhenguoliusheng: but for flavor access, maybe also need a separated endpoint.08:19
liushengzhenguo: because the access projects is a list of project ids, right ?08:20
zhenguoliusheng: because we don't show it to users, not like extra specs08:21
liushengzhenguo: got it08:22
openstackgerritliusheng proposed openstack/mogan master: Tempest: add tests for server states interfaces
liushengzhenguo: I just skip the rebuilding test case ^^08:25
zhenguoliusheng: cool08:26
liushengzhenguo: do you still have following patches after the above patch ?08:34
zhenguoliusheng: no, hah08:56
liushengzhenguo: I have commented08:56
zhenguoliusheng: thanks, will check08:57
zhenguoFYI, we got a presentation selected on OpenStack Days China next month.08:58
openstackgerritZhenguo Niu proposed openstack/mogan master: Following up patch of 6673dbb88fd35d8491248fc94bdc6365d352703a
zhenguoliusheng: hi, wrt , IIRC, metadata is a reserved name in DB model, you can help to confirm it09:29
liushengzhenguo: if so, how about named it server_metadata ?09:33
zhenguoliusheng: you mean change the db field name?09:33
zhenguoliusheng: or change all things to server_metadata09:34
zhenguoliusheng: server.server_metadata seems odd09:34
liushengzhenguo: hah09:34
liushengzhenguo: yes, seems, the metadata is a reserved name, it  is not blocking09:38
liushengzhenguo: you can just update others I commented09:38
zhenguoliusheng: ok, thanks, and I think that's the reason why ironic node use extra, hah09:38
liushengzhenguo: hah09:39
liushengzhenguo: the description field in your new flavor model also don't allow to be update ?09:40
zhenguoliusheng: I think so, but we can discuss09:41
zhenguoliusheng: we will show the description with deployed server09:41
zhenguoliusheng: if it changed, may make users confusing09:41
*** liusheng has quit IRC09:42
*** liusheng has joined #openstack-mogan09:42
liushengzhenguo: ok, make sense to me09:43
*** wanghao has quit IRC09:43
openstackgerritZhenguo Niu proposed openstack/mogan master: Change to use server metadata instead of extra
zhenguoXinran: will you take over luyao's patch ?09:47
liushengzhenguo: I just given your a little -l for your flavor spec patch :-D09:49
zhenguoliusheng: hah, got it, will check09:49
Xinranzhenguo, yes09:49
liushengzhenguo: btw, do you know where to get the private key of the keypair of our physical env ?09:50
zhenguoliusheng: I just import my puglic key09:51
zhenguoyou can use that keypair directly on our env09:51
liushengzhenguo: ok got it09:51
zhenguoXinran: ok, thanks, you can go ahead and drop port_type support09:51
*** liujiong has quit IRC09:58
openstackgerritZhenguo Niu proposed openstack/mogan-specs master: New flavor for baremetal servers
* zhenguo away10:00
Xinranzhenguo,  so just call ironic vif_attach/detach is enough?10:07
*** litao__ has quit IRC12:03
*** openstackgerrit has quit IRC13:18

Generated by 2.15.3 by Marius Gedminas - find it at!