02:02:46 #startmeeting mogan 02:02:47 Meeting started Thu Mar 9 02:02:46 2017 UTC and is due to finish in 60 minutes. The chair is zhenguo. Information about MeetBot at http://wiki.debian.org/MeetBot. 02:02:48 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 02:02:51 The meeting name has been set to 'mogan' 02:02:59 o/ 02:03:04 o/ 02:03:07 o/ 02:04:00 wanghao, litao, are you here? 02:04:10 here 02:04:30 ok, let's start 02:04:36 as always, the agenda: 02:04:44 #link https://wiki.openstack.org/wiki/Meetings/Mogan#Agenda_for_next_meeting 02:04:58 seems the agenda has not changed for weeks, lol 02:06:08 ok, let's jump in 02:06:15 #topic announcements and reminders 02:06:38 We have landed the patch that makes ironic a driver for mogan, so it's not a requirement anymore 02:07:09 maybe we need to remove it from our requiremetns, but should hold on for a while before cloudboot driver ready 02:07:33 that makes many conflicts :) 02:07:54 hah yes 02:08:25 no better way to avoid that 02:08:32 indeed 02:09:20 and a lots of following up refactors will coming, hah 02:09:28 more conflics will come 02:10:03 I plan to make the taskflow and driver interfaces more general for every provisioning tools 02:11:32 zhangyang: how's everything about the cloudboot driver? 02:12:42 zhenguo: we are now designing the driver. 02:12:57 zhenguo: will finish it in a few weeks. 02:13:37 zhangyang: ok, and I will refactor the base driver and the workflow by this week 02:14:17 another thing is the Valence integartion 02:14:42 I drafted a spec for Valence integartion. need to align with Valence guys more. 02:14:55 #link https://review.openstack.org/#/c/441790/ 02:16:09 lgtm 02:16:22 hah thanks 02:16:50 liusheng: about the flavor refactor, do you have some suggestions? 02:17:45 I tried some baremetal servers provider and find out that seems only one flavor can not meet the requirements 02:17:49 zhenguo: TBH, I don't know the requirements of Valance or Cloudboot, so maybe wen should consider based on the two drivers 02:18:33 yes, but for general baremetal flavors, seems need to consider more 02:18:57 like the current instance type(flavor) seems only can include basic informations like cpu, ram 02:19:07 zhenguo: yes, maybe for Valence, we need to specify cpu, disk, ram ? 02:19:12 zhenguo: not sure 02:19:19 but for network interfaces and disks, maybe need some other hints 02:19:35 yes, for valence we need that basic properties 02:20:36 could we design a flavor model more abstract, and user can define the properties, like glance image. 02:20:50 zhenguo: does ironic only need to set the flavor name, e.g. small to the node ? 02:21:14 wanghao: yes we should 02:21:25 liusheng: not only ironic, 02:21:41 we should figure out a general way, but not that easy 02:22:10 when creating a server you may need to specify multi --nic and --disks or raid config or others 02:22:34 zhenguo: I mean ironic only, not only ironic. lol 02:23:15 liusheng: ironic for us is a provision tool 02:23:16 zhenguo: the association between flavor and ironic nodes is only the flavor name ? 02:23:47 liusheng: yes, currently it is 02:24:08 but the when we said flavor, seems it's for scheduler right? 02:25:04 currently scheduler filters take into account of instance_type and port_type to select a baremetal node 02:25:41 zhenguo: I think so 02:25:48 we may need to design a more general node model 02:26:31 we can split a node into board(including ram, cpus..), nics, and disks, and maybe more devices 02:26:54 and the flavor also need to splict into a few types like what we currently do 02:27:14 like instance_type(maybe need to renamed), port_type, and disk_type, and so on. 02:28:17 zhenguo: so this implementation is more like Nova's or Cinder's. lol 02:28:23 and we collect all these types specified with the create instance request, then combine it to a big flavor for scheduler 02:29:07 we have a complex flavor than Nova or Cinder. haha 02:29:11 liusheng: does nova have these devices model 02:29:41 zhenguo: not so much 02:29:46 yes, we really need a flavor like that to meet the requrement, 02:30:36 hah, we need to discuss this more, I would like to do more refacors before we continue 02:31:16 as we already have the basic functinalities, a good architecture can make us more lightweight 02:31:28 I think maybe it is like this: Mogan_flavor={'hardware': {'cpu':xxx, ram:'xxx', 'nics':xxx}, type:{'instance_type': xxx, port_type:xxx,....}} 02:32:06 maybe add one more extra_spec:{xxx:xxx} 02:32:07 wanghao: yes, need a spec for this 02:32:37 zhenguo: yes, we can discuss it more detail 02:32:49 wanghao: and you can bring fiberhome's requirements here to help us be more mature with the refactor 02:33:19 zhenguo: sure 02:33:47 hah ok, I will draft a spec for that after finished the driver refactoring 02:34:34 ok, anyone elase have thing for announcements and reminders? 02:35:23 let's move on 02:35:27 #topic task tracking 02:35:43 #link https://etherpad.openstack.org/p/mogan-task 02:36:44 I have took over the config drive task 02:37:47 I think we can only support configdrive, as metadata server need more works for other projects and there should not be two metadata servers in one env 02:37:55 zhenguo: you are so capable 02:38:25 are you kidding me, haha 02:38:42 noop 02:39:10 hah, but anyhow, the config driver will not be done before the refactoring finished 02:39:38 I have tested that a partition appeared on the cirros instance, 02:39:44 zhenguo: :) so we don't have plan to add metadata server in current period ? 02:40:00 liusheng: no only current period, hah 02:40:08 maybe forever 02:40:09 zhenguo: hah 02:40:29 configdrive can meet all requirements hah 02:40:51 sounds cool 02:40:55 metadata server need a long path to maintain 02:41:14 mogan instance->neutron-metadata-agent->metadata server 02:41:17 zhenguo: Quota bp can be finished in next week, after that I can be part in other work. 02:41:29 yes 02:41:35 wanghao: cool 02:41:48 zhenguo: and I hope I can finish the console job next week 02:41:49 liusheng: and it requires network 02:41:58 liusheng: hah cool 02:42:03 zhenguo: the PoC has been tested ok 02:42:18 with shellinabox proxy? 02:42:22 zhenguo: yes 02:42:52 ok, thanks, that's really help 02:43:14 zhenguo: np 02:43:30 I will try to finish the driver refactoring by this week, to avoid more conflicts for you guys, hah 02:44:01 zhenguo: you first :) 02:44:16 hah, refactor first 02:45:10 oh, just want to ping luyao 02:45:19 come back again, 02:45:32 seems your client is not stable luyao 02:46:26 are you still work on the instance faults? seems there should be a following up patch to make it work 02:47:16 ok, will ping her afterwards 02:47:19 let's mvoe on 02:47:28 #topic open discussion 02:47:40 who's got a thing here? 02:48:26 bani, are you around? 02:49:42 a new comer ? 02:49:56 yes, I think he is from NEC 02:50:03 cool 02:50:50 ok, if there's no one want a thing to be disuccsed here, we can end the meeting 02:51:15 thanks everyone for joining the meeting! 02:51:20 ok bye 02:51:25 #endmeeting