14:00:11 <zhaochao> #startmeeting trove
14:00:12 <openstack> Meeting started Wed Mar 28 14:00:11 2018 UTC and is due to finish in 60 minutes.  The chair is zhaochao. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:14 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:16 <openstack> The meeting name has been set to 'trove'
14:00:29 <zhaochao> #topic roll call
14:00:36 <fanzhang> o/
14:01:36 <zhanggang> o/
14:01:38 <Yao> hi
14:01:51 <zhaochao> hi
14:01:53 <songjian> o/
14:02:00 <zhaochao> hi
14:03:36 <fanzhang> I think kumarmn is here too
14:03:51 <kumarmn> yes
14:04:20 <zhaochao> ok, maciejjozefczyk may not show up, so let's start
14:04:20 <kumarmn> I was traveling last week
14:04:28 <Yao> coo
14:04:29 <Yao> cool
14:04:44 <zhaochao> #topic Rocky goal updates
14:05:26 <zhaochao> as we decided on last weekly meeting, we could do some updates in the recent week
14:05:47 <zhaochao> kumarmn: songjian fyi, http://eavesdrop.openstack.org/meetings/trove/2018/trove.2018-03-21-14.00.html
14:06:54 <zhaochao> we'll use the priority page to track updates, https://etherpad.openstack.org/p/trove-priorities-and-specs-tracking
14:07:26 <zhaochao> and https://docs.google.com/spreadsheets/d/1Jz6TnmRHnhbg6J_tSBXv-SvYIrG4NLh4nWejupxqdeg for self assignment
14:07:26 <Yao> I will commit some patches about osc this weekend, so busy recently, sorry for that.
14:07:45 <fanzhang> zhaochao I just updated my working progress about nova file injection deprecation. Still drafting spec.
14:08:25 <zhaochao> Yao: that's ok, all of us may not have time to focus on the upstream activities
14:08:48 <Yao> yes
14:09:05 <zhaochao> fanzhang: got that, thanks
14:09:34 <zhaochao> I'm still working on mox removal, this should be completed in the trove repo
14:10:21 <zhaochao> and I'm working the first patch for trove-dashboard, it may be done in a few days
14:11:44 <fanzhang> huh, about Octavia way of communicating between API and guestagent, is huntxu around?
14:11:46 <songjian> I think I can provide some help with offline image based on trovestack,like a doc. Of course, this is not the best way to build images.
14:12:02 <zhanggang> About cluster supporting, do you have any idea of improvement direction. Next week I think I have time work around this.
14:13:33 <zhaochao> songjian: thanks, documentation also do help
14:13:49 <fanzhang> zhanggang IMHO, let's find the gaps first. Then you will be clear what needs to be done.
14:14:23 <zhaochao> any other updates?  zhanggang added another topic about cluster supporting, we could move on to it as the next topic
14:14:25 <kumarmn> the reboot discussion that the previous PTL initiated was to make clusters a first-class citizen:
14:14:27 <kumarmn> https://openstack.nimeyo.com/115325/openstack-dev-trove-all-tc-a-proposal-to-rearchitect-trove
14:14:55 <kumarmn> Not sure whether there is anything specific you can get from that discussion.
14:15:22 <zhaochao> ok, let's move the next topic
14:15:42 <zhaochao> #topic Improve cluster supportingļ¼šimprovement direction
14:16:41 <zhanggang> kumarmn: thanks, I will read the info.
14:16:52 <zhaochao> yes, the discussion mentioned by kumarmn is the start, but no progress has been made
14:18:29 <kumarmn> There is nothing specific about how to improve, or what is broken.
14:18:42 <fanzhang> songjian told me that this proposal made by amrith didn't get much support, and then ended up with nothing happened ?
14:19:03 <fanzhang> yep, we may need to find the gaps by ourselves.
14:19:13 <kumarmn> he was thinking that if we start from scratch we can make it like k8s.
14:21:00 <zhanggang> One direction is functional improvement, like add resize_cluster_volume, resize_cluster_flavor etc.
14:21:27 <songjian> fanzhang,I do not know this thing very clearly. . . Just heard of this plan, but the details are not clear
14:21:33 <fanzhang> kumarmn it's kind a huge refactoring things which would take a lot time and development
14:22:18 <fanzhang> maybe put it aside right now? But we can read this mail and try to find something useful.
14:22:33 <kumarmn> right, I was not suggesting that we go that route. I agree that we should tackle the functional like zhanggang suggests.
14:23:08 <fanzhang> +1
14:23:12 <zhaochao> zhanggang: these should work and help, we also can do some abstraction about the cluster implements of all the database backends we now supported
14:24:04 <zhaochao> yes, all of us may agree on doing improvements based on the current Trove implements
14:25:56 <zhaochao> currently cluster implements are using the strategies framework, we could also do some improvements on that
14:26:36 <fanzhang> Looks like zhanggang has to take some time to bring up a spec, and then we can discuss more detail things
14:27:40 <zhanggang> And another direction I thought is cluster_create. we may treated some parameters as cluster parameters instead of instance. For example, az, nics, just like configuration
14:27:59 <zhanggang> configuration = body['cluster'].get('configuration')
14:28:55 <zhanggang> and bugs like https://review.openstack.org/#/c/556764/ will be fixed
14:28:58 <fanzhang> that's more like what zhaochao says above, some abstraction works?
14:29:16 <zhanggang> fanzhang: yes, do more abstraction.
14:29:20 <zhaochao> yes, zhanggang, you can adding your ideas(and explanations) in the prioritiy tracking page, under the "Improve cluster supporting" topic
14:30:18 <zhanggang> zhaochao: ok
14:32:05 <zhaochao> zhanggang: and if you want to notify us about the updates on this topic, you can also talk about it in #openstack-trove channel, I'm on the channel most of the working time
14:33:05 <zhaochao> ok, anything more about the cluster improvements?
14:33:38 <songjian> For this issue, I think I need to investigate some cluster features like roll restart etc...
14:34:01 <zhanggang> zhaochao: ok, if I have more thought, I will use that channel of email.
14:34:53 <zhaochao> songjian: that would be great, thanks
14:35:09 <fanzhang> sounds great. you all have real use cases :)
14:36:04 <songjian> nothing,It's interesting
14:37:00 <zhaochao> ok, we don't have any other topics this week, let's move to open discussion
14:37:19 <zhaochao> #topic Open discussion
14:37:26 <fanzhang> about versioned object
14:38:14 <fanzhang> I recently think it's kind a useful feature in trove to handle things nova microversion
14:38:24 <fanzhang> like nova microversion
14:39:35 <fanzhang> If we had some checks based on microversions, we could use more cross projects' features
14:39:59 <zhaochao> fanzhang: yes, I also agree, but most of us think it's priority is not high, and maybe because we didn't do much modification about API before
14:41:01 <zhaochao> however it will be great to add the basic microversion framework to Trove
14:41:01 <fanzhang> I remembered that Samuel told us that we are using nova api 2.10, but the lates nova api microversion is 2.57. Why do we use 2.10, and how to control our api requests' version with nova, cinder etc?
14:42:42 <zhaochao> I also didn't make it clear about why we stick with nova 2.10, but I did some changes about the novaclient and neutronclient usage, things may have changed
14:43:16 <zhaochao> some feature will be remove from novaclient/neutronclient/cinderclient, even we're using an old api microversion
14:43:29 <fanzhang> yeah, I am interested with that. My thought is if we have versioned things, we can be easily aware of api version we should use if we want some features in trove. Like we can specify api version 2.xx to not use nova 2.57 and something more or less.
14:43:43 <fanzhang> zhaochao yep, that's one of my concerns
14:44:26 <fanzhang> but do not have a clue about how to design that right now.
14:44:45 <fanzhang> so what do you guys think? kumarmn songjian zhanggang Yao
14:46:04 <zhaochao> the basic design about microversion is for client usage, i.e. if we changes the API, we should increase the micronversion, then client with old microversion specified won't be broken
14:47:08 <zhaochao> but this is not about Trove using novaclient/cinderclient/neutronclient, micronversions in Trove should be used for the interaction between troveclient(and other client) with Trove API
14:48:16 <fanzhang> fair enough. but how can we control the api version using novaclient/cinderclient/neutronclient, curious about that.
14:48:50 <zhaochao> for the usage of novaclient/cinderclient/neutronclient, we could only follow the decision of these project, and update the way we use these client libraries
14:50:24 <fanzhang> oh right, that could work.
14:52:06 <fanzhang> If i got some time, I would like to think it through
14:52:25 <zhaochao> fanzhang: yes, though the versioned object is not on the top of the priorities list, but I think it would be great if we do some initial work on this
14:53:17 <zhanggang> Is any other  project use microversion besides nova cinder neutron?
14:53:35 <fanzhang> Sounds cool :) no more topics from me
14:54:38 <fanzhang> zhanggang good question šŸ¤£ I have no idea.
14:54:40 <zhaochao> zhanggang: didn't read code of many other projects, some more active evolving projects should also use this method
14:56:08 <zhaochao> ok, I think that's all today. Do you have anything more to talk about?
14:56:30 <zhanggang> No more :)
14:56:40 <fanzhang> no, and have a good day, have a good night :)
14:56:46 <Yao> no more for me
14:57:09 <Yao> bye
14:57:21 <zhaochao> Ok, Let's end here. Thanks, everyone.
14:57:22 <songjian> no more for me
14:57:34 <zhanggang> Bye
14:57:43 <fanzhang> Bye
14:57:59 <zhaochao> Bye
14:58:01 <zhaochao> #endmeeting