02:00:46 <zhenguo> #startmeeting mogan
02:00:47 <openstack> Meeting started Thu Jun 22 02:00:46 2017 UTC and is due to finish in 60 minutes.  The chair is zhenguo. Information about MeetBot at http://wiki.debian.org/MeetBot.
02:00:48 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
02:00:50 <openstack> The meeting name has been set to 'mogan'
02:00:52 <wanghao> hi all
02:00:55 <zhenguo> o/
02:01:05 <liusheng> o/
02:01:09 <shaohe_feng> zhonghua: sorry, I want to ping zhenguo
02:01:15 <zhenguo> lol
02:01:26 <shaohe_feng> o/
02:01:34 <zhonghua> :)
02:02:01 <zhenguo> litao__, zhangyang, Xinran: are you here?
02:02:06 <zhangyang> o/
02:02:32 <zhenguo> as always, the agenda:
02:02:32 <zhenguo> #link https://wiki.openstack.org/wiki/Meetings/Mogan#Agenda_for_next_meeting
02:02:38 <zhenguo> #topic announcements and reminders
02:02:41 <Xinran> o/
02:03:06 <zhenguo> Sydney Summit call for presentations open until July 14
02:03:12 <zhenguo> please help to prepare some if you to time
02:03:24 <zhenguo> s/to/got
02:05:07 <zhenguo> We got a presentation to introduce mogan during the OpenStack Days China next month, maybe that could be one also for Sydney Summit :D
02:06:15 <zhenguo> ...
02:06:42 <zhenguo> seems you guys don't quite like this topic, let's move on, hah
02:06:48 <liusheng> zhenguo: that will be cool, beside introduction, do we have other things that is proper to display ?
02:07:17 <zhenguo> liusheng: I think HFBank got a good topic for this
02:07:30 <zhenguo> zhangyang: wdyt?
02:07:33 <liusheng> zhenguo: hah, mogan+cloudboot ?
02:08:01 <zhenguo> liusheng: yes, if they also run mogan in production before that, it's really cool
02:08:08 <zhangyang> zhenguo: e... i've no idea now...
02:08:16 <zhenguo> zhangyang: hah
02:08:35 <liusheng> zhenguo: does the China Days persentation will include a hands-on ?
02:08:49 <shaohe_feng> liusheng: HFBank  topic on China day?
02:08:57 <zhenguo> liusheng: no
02:09:03 <zhenguo> shaohe_feng: no, for summit
02:09:08 <liusheng> yes
02:09:23 <liusheng> may a hands-on will be cool
02:09:25 <zhenguo> we an prepare with HFBank guys
02:09:50 <zhenguo> liusheng: hah, not sure if the network is good enough
02:10:01 <zhenguo> liusheng: but I will try, still not preparing the slides
02:10:22 <shaohe_feng> liusheng: really good news for mogan.
02:10:40 <liusheng> zhenguo:that is true, a video is alternative
02:10:47 <zhenguo> shaohe_feng: and sure a presentaion with intel guys about valence
02:10:56 <liusheng> hah
02:10:59 <zhenguo> liusheng: sure
02:11:04 <wanghao> zhenguo: do you ensure that Mogan can have a project on board?
02:11:08 <litao__> i am here
02:11:13 <shaohe_feng> zhenguo: sure, valence
02:11:21 <zhenguo> wanghao: still not sure
02:11:27 <zhenguo> wanghao: but seems no rush
02:11:30 <wanghao> zhenguo: okay
02:11:41 <zhenguo> litao__: o/
02:12:23 <shaohe_feng> liusheng: the HFBack's topic is seleced?
02:12:45 <zhenguo> wanghao, litao__: and Fiberhome guys sure have a good views of why Mogan, so we can prepare a presentation as well
02:12:55 <liusheng> shaohe_feng: no, we just hoped there will be on and will be selected. lol
02:13:07 <zhenguo> shaohe_feng: no, I suggest we prepare a topic
02:13:26 <wanghao> zhenguo: yeah, we can talked about it.
02:13:33 <zhenguo> wanghao: greate
02:14:08 <zhenguo> ok, we can discuss the details later
02:14:18 <litao__> zhenguo: greate
02:14:28 <zhenguo> another thing I want to mention here is the Placement service
02:14:52 <zhenguo> we are migrating to placement for resources tracking..
02:15:22 <zhenguo> do you have any concerns?
02:15:38 <liusheng> yes, I am trying a PoC firstly, but sorry for my slow progress :(
02:15:56 <zhenguo> liusheng: hah, it's really a big task
02:16:18 <shaohe_feng> yes, big task
02:16:21 <shaohe_feng> no rush
02:16:32 <litao__> zhenguo: If we put it in placement, the old scheduler will be abandoned?
02:16:52 <zhenguo> litao__: schduler will also be there
02:17:05 <zhenguo> litao__:  we just move resources tracking to placement
02:17:15 <zhenguo> litao__: and all filters will be removed
02:17:18 <shaohe_feng> placement just for resource track
02:17:28 <litao__> zhenguo:  yes, i understand
02:17:45 <zhenguo> yes, and I think placement is more advanced the the current design
02:18:00 <zhenguo> instead of list all nodes and check every filter to see if the node is passed
02:18:08 <shaohe_feng> so mogan claim resouce from placement
02:18:16 <zhenguo> placement will just use a complex sql to query the required nodes
02:18:19 <zhenguo> it's move fast
02:18:26 <zhenguo> ys
02:19:05 <shaohe_feng> and who will report the resource status to palcement?
02:19:17 <zhenguo> shaohe_feng: mogan
02:19:30 <zhenguo> shaohe_feng: mogan resource tracker will update resouces to placement
02:19:41 <zhenguo> and mogan scheduler will set allocation info there
02:19:58 <shaohe_feng> zhenguo: got it.
02:20:04 <zhenguo> with placement, we don't save baremteal resources in mogan db
02:20:22 <liusheng> yes
02:20:29 <litao__> But Currently the placement locates in nova, so mogan must be with nova?
02:20:48 <zhenguo> litao__; we can just enable placement service
02:20:58 <zhangyang> zhenguo: then how will we manage aggregates
02:21:00 <liusheng> litao__: placement has itself endpoint
02:21:01 <zhenguo> litao__: placement is in a different stack
02:21:08 <litao__> zhenguo: OK
02:21:14 <zhenguo> zhangyang: leverage placement aggregates
02:21:18 <liusheng> litao__: we can treat it a indendent service
02:21:31 <liusheng> s/indendent/independent
02:21:39 <litao__> liusheng: i do
02:21:40 <zhenguo> zhangyang: every node is a resource provider there, we can call placement api to make nodes aggregates
02:21:57 <zhenguo> zhanyang: in mogan flavor we can pass aggregates info with extra specs to placement
02:22:14 <shaohe_feng> zhenguo: placement has a Proposition for every resource update. most distribute system use this way for its consistency
02:22:46 <zhangyang> zhenguo: oh, i see, mainly use aggregates for server group usage.
02:22:52 <zhenguo> shaohe_feng: yes
02:23:01 <zhenguo> zhangyang: yes
02:23:10 <zhenguo> zhangyang: that could be easy to management
02:23:36 <zhenguo> ok, we can continue to discuss this in the next topic
02:23:45 <zhangyang> zhenguo: i see, need to learn more about placement
02:23:54 <zhenguo> zhangyang: hah,
02:24:08 <zhenguo> zhangyang: it's easy for you guys
02:24:36 <zhenguo> ok, not much of announcements, anyone else have a thing here?
02:24:49 <zhenguo> a, the contribution
02:24:58 <zhenguo> #link http://stackalytics.com/report/contribution/mogan/30
02:25:14 <zhenguo> seems the review stats is better than last week
02:25:18 <shaohe_feng> zhenguo: seems the theory of place is simple, use fold to manage the resouces
02:25:22 <liusheng> hah
02:25:29 <wanghao> yes
02:25:49 <shaohe_feng> zhenguo: you know fold can be recursive
02:26:33 <zhenguo> shaohe_feng: placement also introduce a notion of nested resources, we can discuss in the next topic task-tracking, hah
02:26:45 <shaohe_feng> zhenguo: it also support tag, like gmail,
02:27:11 <shaohe_feng> zhenguo: yes, the notion is fold
02:27:25 <zhenguo> shaohe_feng: we can set traits to resource providers like tag
02:27:34 <zhenguo> and filter with them
02:27:41 <zhenguo> ....
02:27:48 <shaohe_feng> zhenguo: only two ways to manage resouce, fold and tag
02:27:58 <zhenguo> shaohe_feng: yes,
02:28:12 <shaohe_feng> zhenguo: here recursive I means nested
02:28:36 <shaohe_feng> fold = nested
02:28:40 <shaohe_feng> tag is flat,
02:28:43 <zhenguo> shaohe_feng: yes, in fact, I more like a nested resouce for baremetal instead of just one custom resouce class
02:29:37 <shaohe_feng> zhenguo: yes. I saw most resource manage tools use nested
02:29:37 <zhenguo> let's continue talk about about the contribution
02:29:41 <shaohe_feng> OK
02:29:44 <zhenguo> shaohe_feng: hah
02:30:04 <zhenguo> again, I want to mention here again, review quantities/qualities is the most important aspect of considering one to be a core reviewer.
02:30:30 <zhenguo> hope you guys can keep reviewing
02:30:46 <liusheng> yes, and we need more cores, lol
02:30:52 <zhenguo> sure
02:31:02 <zhenguo> more cores and more diversty
02:31:03 <shaohe_feng> yes.
02:31:28 <zhenguo> Our 'review hour' seems not running very well, we should focus one some high priority things, maybe need to prepare a ehterpad every time
02:31:37 <zhenguo> wdyt?
02:32:27 <zhenguo> to help us focus one some important patches/specs, and we can fully discuss during the hour
02:32:54 <litao__> zhenguo: yes
02:33:06 <wanghao> zhenguo: sounds good
02:33:19 <liusheng> yes, make sense
02:33:37 <litao__> zhenguo: more efficiency
02:33:46 <zhenguo> so, start from today's review hour?
02:34:00 <shaohe_feng> sounds good.
02:34:15 <zhenguo> the most things I like to be reviewed is the specs
02:35:01 <zhenguo> ok, anything else?
02:35:30 <zhenguo> if not, let's move on
02:35:38 <zhenguo> #topic task tracking
02:35:50 <zhenguo> #links https://etherpad.openstack.org/p/MoganWhiteBoard
02:36:04 * zhenguo gives folks a few minutes to review
02:37:43 <zhenguo> hi guys, please remember to update the tasks to help wanghao construct the weekly report, thanks
02:38:01 <wanghao> yeah, thank guys
02:38:55 <zhenguo> about the new flavor task, I think it's almost done, and with this bp, I don't plan to move to use resources for scheduling
02:39:10 <zhenguo> will wait for placement got get rid of the old way of scheduling
02:40:43 <zhenguo> shaohe_feng: we can  continue the discussion about placement here, hah
02:40:56 <wanghao> so flavor is not working for scheduling temporary
02:41:15 <zhenguo> wanghao: no, we also use the flavor name for scheduling now
02:41:26 <zhenguo> just added a resources filed, but not use it
02:41:40 <zhenguo> it's prepared for placement
02:41:50 <zhenguo> no need to change it again and again
02:42:01 <zhenguo> make sense?
02:42:04 <liusheng> it probably the currently new flavor model can be merged fistly, right ?
02:42:27 <zhenguo> liusheng: yes
02:42:52 <liusheng> ok, great
02:43:08 <wanghao> zhenguo: ok, that make sense
02:43:22 <zhenguo> after moving to placement, we will only select nodes based on the resources
02:43:39 <zhenguo> will get rid of all filters
02:44:14 <wanghao> placement now can get the node information from ironic now?
02:44:14 <zhenguo> which will make us lose the ability of avaliability zone
02:44:38 <zhenguo> wanghao: no, we get from ironic then update it to placement
02:44:40 <liusheng> may we can also keep the az filters currenly ?
02:44:55 <wanghao> zhenguo: ah, I see
02:45:00 <zhenguo> liusheng: all resources move to placement
02:45:02 <liusheng> zhenguo: Nova also keep many filters now
02:45:08 <liusheng> zhenguo: yes
02:45:11 <zhenguo> liusheng: and how to save the az info with resource provider
02:45:25 <zhenguo> liusheng: they will get rid of it in future
02:45:33 <liusheng> zhenguo: yes, I know...
02:45:35 <zhenguo> liusheng: and nova also save compute_nodes info with ti
02:45:40 <liusheng> zhenguo: I mean for now.
02:46:03 <zhenguo> liusheng: but I don't want to save the resources to mogan db and placement db, it's duplicate
02:46:16 <zhenguo> maybe we can leverage traits for az
02:46:26 <liusheng> zhenguo: oh, got it, I see
02:46:33 <liusheng> zhenguo: yes, sure
02:46:37 <zhenguo> we set a az trait to every resource provider(node)
02:46:55 <shaohe_feng> zhenguo: do you means tag
02:46:58 <liusheng> zhenguo: yes
02:47:09 <shaohe_feng> zhenguo: tag with a AZ name
02:47:11 <zhenguo> shaohe_feng: similar to tag
02:47:15 <liusheng> shaohe_feng: similar things
02:47:23 <zhenguo> shaohe_feng: in placement notion, it's traits
02:48:22 <zhenguo> also seems there's a os-traits library including all standard information
02:48:46 <liusheng> zhenguo: not sure, how to treat the az list api
02:49:02 <liusheng> zhenguo: can mogan api and then can placement api ?
02:49:03 <shaohe_feng> zhenguo: yes, use stander notion, traits. haha
02:49:10 <zhenguo> liusheng: get all resouce providers from nova...
02:49:22 <zhenguo> shaohe_feng: yes
02:49:35 <liusheng> zhenguo: a bit weird
02:49:50 <zhenguo> liusheng: yes
02:50:01 <zhenguo> liusheng: but not sure how to hanle it gracefully
02:50:25 <zhenguo> in fact, I also think set az info with every baremetal node is odd
02:50:37 <liusheng> zhenguo: ok, it is not important
02:51:02 <liusheng> zhenguo: what's your idea ?
02:51:22 <zhenguo> liusheng: not sure, for az, I think one ironic should be in one az
02:51:32 <zhenguo> liusheng: so one mgoan-engine?
02:51:42 <zhenguo> maybe with multi engine services
02:52:12 <liusheng> maybe combining with nodes aggregate like nova does
02:52:59 <zhenguo> sounds good
02:53:43 <zhenguo> need to discuss more, will upate the placement spec
02:54:06 <liusheng> zhenguo: thanks for your spec :D
02:54:42 <zhenguo> liusheng: hah, please feel free to update it, you are the primary assigner
02:55:05 <liusheng> zhenguo: lol
02:55:12 <zhenguo> hah
02:55:42 <liusheng> btw, please you guys help to review the rest tempest tests
02:56:02 <zhenguo> sure
02:56:06 <liusheng> all the api interfaces now have temptest tests to cover
02:56:17 <zhenguo> liusheng: or maybe we can save az info in mogan cache?
02:56:30 <zhenguo> just for user's listing
02:57:10 <zhenguo> or maybe cache all resource providers
02:57:14 <liusheng> zhenguo: how it goes if restarting services ?
02:57:29 <zhenguo> restarting service will fetch that again
02:57:36 <liusheng> zhenguo: sounds duplicated things... hah
02:57:43 <zhenguo> hah
02:58:00 <liusheng> zhenguo: let's disscuss after placement landed
02:58:04 <zhenguo> as we also want to provide node list
02:58:10 <zhenguo> liusheng: ok
02:58:14 <liusheng> zhenguo: oh, yes
02:58:25 <zhenguo> now just focus on how to migrate to placement
02:58:43 <zhenguo> oh, it's almost time
02:58:48 <wanghao> zhenguo: could this be a spec to explain what we want to do?
02:58:53 <shaohe_feng> zhenguo: what's the AZ will be for?
02:58:58 <zhenguo> wanghao: sure
02:59:00 <liusheng> zhenguo: I also asked zhenyu if placement have itself client, he also has no idea
02:59:03 <zhenguo> wanghao: will be in the placement spec
02:59:04 <shaohe_feng> zhenguo: one AZ means one cluster?
02:59:10 <wanghao> zhenguo: ok
02:59:31 <zhenguo> shaohe_feng: yes, so maybe we don't need to set it to every node
02:59:46 <zhenguo> maybe an ironic should provide a az
03:00:13 <zhenguo> oh, it's time
03:00:18 <zhenguo> thanks y'all
03:00:22 <zhenguo> #endmeeting