14:18:09 <Li_Liu> #startmeeting openstack-cyborg
14:18:10 <openstack> Meeting started Wed Oct 24 14:18:09 2018 UTC and is due to finish in 60 minutes.  The chair is Li_Liu. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:18:11 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:18:13 <openstack> The meeting name has been set to 'openstack_cyborg'
14:18:21 <Li_Liu> #topic Roll Call
14:18:29 <Sundar> #info Sundar
14:18:34 <Li_Liu> #info Li Liu
14:18:42 <wangzhh> #info wangzhh
14:18:56 <Coco_gao> #info Coco_gao
14:19:08 <Sundar> Thanks, Li
14:19:14 <Li_Liu> #topic DB schema work
14:19:33 <Sundar> #link https://docs.google.com/document/d/1XLQtvyGJeEgo3ztBQiufWLF-E7S7yGLaYrme8iUPtA0/edit
14:19:42 <Sundar> Re-posted the link
14:20:04 <Li_Liu> as we discuss earlier, Sundar will open a Storyboard and put DB tasks there.
14:20:23 <Sundar> Li_Liu: Yes.
14:20:41 <Sundar> This is the same schema that was presented by Zoom a couple of weeks ago
14:20:43 <Coco_gao> Thanks to break down the work
14:21:03 <wangzhh> And we should start with the new DB schema first.
14:22:21 <Sundar> Welcome. The next big thing will be the driver-agent API. Some of the APis are already covered in https://review.openstack.org/#/c/608624/
14:22:55 <Sundar> That is an important spec, because that is the core of the Cyborg implementation for Nova-Cyborg interaction. Can you all review it?
14:23:09 <Li_Liu> shall we start with a spec for the new DB schema?
14:23:55 <wangzhh> Li_liu: I think so.
14:24:28 <Sundar> Li: OK. Does anybody else want to write that spec?
14:25:06 <Li_Liu> hmm
14:25:13 <Li_Liu> I might be able to help
14:25:20 <wangzhh> Sundar, OK, could u help to add me as a reviewer.  https://review.openstack.org/#/c/608624/
14:25:27 <Li_Liu> just send all the stuff you got so far, Sundar
14:26:14 <Sundar> wangzhh: Done
14:26:25 <Sundar> Li_Liu: That ink is the main one
14:26:29 <Sundar> *link
14:27:03 <Li_Liu> ok, I will find some time at night to work on it
14:27:18 <Sundar> Li_Liu: Great, thanks
14:27:44 <Sundar> What would be covered in the spoec, apart from tables?
14:28:26 <Sundar> Are there versioning/upgrade issues, performance aspects (defining index etc.)?
14:29:06 <Li_Liu> well, shall we give a newer version of the DB?
14:29:32 <Li_Liu> I though we agreed earlier, we don't need to worry about the upgrade for this change
14:30:00 <Sundar> Yes, but upgrade from Stein to T may require versioning
14:30:30 <Sundar> We can implement in phases but don;t we need some plan in Stein for upgrade to T?
14:31:09 <Li_Liu> ok, sure. I will keep that in mind
14:31:27 <wangzhh> Yes, but we need explain it. Use doc or other way. Maybe someone use the old db schema.
14:31:43 <Sundar> Also, I was told that upgrade from Rocky to Stein should also be a rolling upgrade. We may need to check what that takes.
14:32:24 <Sundar> Does any of us have experience with OpenStack upgrades?
14:34:26 <Coco_gao> I think that will be fine since cyborg not intergrate with nova in Rocky, Rolling upgrade means you upgrade software while not breaking the VM's work.
14:34:31 <Li_Liu> I know how to write script for migrating DB
14:34:42 <Li_Liu> just in case it's needed
14:35:53 <Sundar> Li_Liu: Good to know. Also, we need to address database performance considerations -- creating indexes, joins/query optimizatons, etc. I did stuff that like that a decade ago but not with SqlAlchemy.
14:36:44 <Sundar> IMHO, documenting the tables in a spec is one thing. Going beyond that to address upgrades and performance will give us all  a sense of comfort :)
14:37:17 <Coco_gao> That's nice to have.
14:37:55 <Sundar> Also, we may need to set rules/guidelines like, only the conductor can access the db, not the agent. Otherwise we have upgrade issues.
14:39:01 <Sundar> In the future, may be the API server can also access the db -- others do that for scalability apparently. May be we should keep that for the future till we understand upgrade issues fully
14:41:24 <Coco_gao> Sundar, may be we can mention that at beginning of your spec on db schema , just in case we forget some day.
14:42:23 <Li_Liu> I will try to upload the draft asap so that you guys can provide all the comment there :P
14:42:43 <Sundar> Li_Liu: Sounds good.
14:43:13 <wangzhh> Thx, Li.
14:43:53 <Li_Liu> npnp
14:44:02 <Sundar> The driver-agent API and discovery probably need a new spec. Some of the APIs are already in the spec I mentioned, but that doesn;t cover the driver report structure.
14:45:23 <wangzhh> Sundar, why we need a new discovery spec?
14:46:11 <Coco_gao> not discovery spec?
14:47:04 <Sundar> Sundar: First, it needs to be in cyborg-specs repo, not cyborg. Secondly, the new PTG discussion has to be reflected there, including references to new specs. Also, the relation to new db schema needs to be discussed.
14:47:50 <Sundar> It could be a single spec that covers the driver API, report structure, etc.
14:48:24 <wangzhh> So, the workflow of discovery is not different.
14:49:01 <Sundar> wangzhh: Yes
14:49:31 <wangzhh> OK, got it.
14:49:43 <Coco_gao> Can we abondon the old or not implemented specs? everytime I checked the website, I just don't know which spec should I look.
14:49:45 <wangzhh> Sundar: Thx.
14:50:25 <Coco_gao> abandon
14:50:34 <Sundar> Coco_gao: I understand. Do you all think that we should close the old ones?
14:50:59 <Coco_gao> Yeah
14:51:19 <wangzhh> I think so.
14:51:27 <Coco_gao> so many work you have be done on specs~
14:52:25 <Coco_gao> Thanks
14:53:04 <Coco_gao> so many work you have been done on specs~
14:53:34 <Sundar> Coco_gao: Welcome, np :)
14:53:54 <Coco_gao> Have the set-up docs ready?
14:54:29 <Sundar> Li_Liu: what do you think? Shall we record this as a conclusion?
14:55:12 <Coco_gao> since  the summit is comming, I guess others may ask for the set-up docs or user-guide docs.
14:56:06 <Li_Liu> agree, we should try to keep it clean
14:56:24 <Li_Liu> if a spec is not needed, just deprecate it
14:56:40 <wangzhh> yumen said she had worked on the set-up docs.
14:56:49 <wangzhh> CoCo.
14:56:52 <Li_Liu> in case you wanna keep them for a while as reference, mark them
14:56:55 <Coco_gao> ok, that'll be great
14:57:36 <Sundar> Li_Liu: deprecate == abandon, right?
14:58:28 <Li_Liu> deprecate = about to be abandoned
14:58:49 <Sundar> OK. Thanks.
14:58:50 <Li_Liu> lol, that should be the exact translation
15:00:01 <Sundar> We have the devstack install doc. What other set up doc do we need? Until we get the code in, anyway, it is not usable, right?
15:00:52 <Coco_gao> Yes, devstack install doc is what I mean.
15:00:54 <Sundar> Li_Liu: I asked because Coco said abandon and you agreed but called it 'deprecate'.
15:01:59 <Coco_gao> I don't know whether 'abandon' is proper, sorry.
15:02:06 <Li_Liu> ah. I was just proposing in case we still wanna keep them for a while as reference. if we don't, just abandon the,
15:02:35 <Sundar> NP, just trying to understand.
15:03:31 <Sundar> Sorry, I need to drop off for my next call.
15:03:41 <Coco_gao> anything else to discuss?
15:03:51 <Coco_gao> Bye, Sundar
15:04:14 <Sundar> Bye, Coco and all
15:04:23 <wangzhh> Bye.
15:04:59 <wangzhh> Li_liu: anything else to disscuss?
15:07:00 <Coco_gao> bye
15:07:44 <wangzhh> Byebye.
15:10:12 <Li_Liu> I will need the meeting now
15:10:19 <Li_Liu> shall I ?
15:10:26 <Li_Liu> Sundar, you still there?
15:15:32 <Li_Liu> #endmeeting