15:00:14 #startmeeting manila 15:00:16 Meeting started Thu Oct 16 15:00:14 2014 UTC and is due to finish in 60 minutes. The chair is bswartz. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:17 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:19 The meeting name has been set to 'manila' 15:00:26 hello all 15:00:27 Hi 15:00:28 hi 15:00:29 \o 15:00:30 hi 15:00:32 hi 15:00:33 hi 15:00:37 hi 15:00:54 hi 15:01:01 #agenda https://wiki.openstack.org/wiki/Manila/Meetings 15:01:34 #topic dev status 15:01:41 we'll start with dev status again this week 15:01:45 Dev status for last week: 15:02:01 1) Manila shares list API 15:02:07 hi 15:02:08 Filtering using share list API was greatly improved 15:02:08 server: #link https://github.com/openstack/manila/commit/8f0d0dc6 15:02:08 client: #link https://github.com/openstack/python-manilaclient/commit/c7456c67 15:02:10 status: merged 15:02:19 2) Manila security-services list API 15:02:28 server: #link https://review.openstack.org/125569 15:02:28 client: #link https://review.openstack.org/126240 15:02:30 status: ready for review 15:02:36 3) Manila snapshots list API 15:02:40 server: #link https://review.openstack.org/128740 15:02:41 client: #link https://review.openstack.org/127622 15:02:44 status: work in progress 15:02:55 4) Horizon with Manila 15:02:55 Recently was updated Horizon: #link https://github.com/NetApp/horizon/tree/manila_juno 15:02:55 Now it has "share-servers" tab in admin panel. 15:03:15 that's the main 15:03:34 cool 15:03:48 anything that's been sitting in review too long? 15:03:58 * bswartz has been slow on reviews this week 15:04:13 little commit: https://review.openstack.org/125294 15:04:18 wanted more opinions 15:04:57 okay I'll review 15:05:15 anything else/questions? 15:05:37 okay 15:05:47 #topic design summit 15:05:55 #link https://etherpad.openstack.org/p/kilo-manila-summit-topics 15:06:11 I've been working on the etherpad, mostly down at the bottom 15:06:48 I've tried to separate the topics into topics that would be good for the design summit, and topics that would be good to discuss during the team meetup 15:06:59 does everyone understand the format for the summit this year? 15:07:41 I'll assume yes 15:08:13 I've proposed 6 session so far 15:08:27 some of them are a bit iffy, but I wanted feedback 15:09:28 bswartz, can u brief quickly about the summit format if its changed from last time ? 15:10:07 so in addition to the "conference sessions" there will be "design summit sessions" for ATCs only just like before 15:10:13 bswartz: we only have 3 session slots, right? 15:10:21 however there will be spots for us to meetup on Friday 15:10:41 and discuss things in a smaller group 15:10:52 we have 3 design summit sessions 15:11:10 and those will likely be attended by all kinds of people who just want to learn about what we're doing 15:11:30 the team meetup will be much smaller I'm sure, with no published agenda 15:11:59 bswartz, thanks 15:12:15 bswartz: about topic "Potential Session: Manila Scheduler" 15:12:23 so the 6 sessions I've proposed have topics which may appeal outside our team 15:12:29 what exactly is planned to be disscussed? 15:13:24 vponomaryov: I'm thinking the discussion would be about specific capabilities we should report and general good practice for setting up a multibackend manila installation 15:14:03 i would be interested in attending the team meetup (for the image project) 15:14:23 well if anyone feels that a topic (from the etherpad) is in the wrong place we can move it 15:14:38 we could do a design summit session on the image project 15:14:51 bswartz: so, main idea - extend existing functionality with manila specific stuff and define what exactly should be addded, correct? 15:15:14 bswartz: each session is 40 minutes. So it's unlikely we can fit all 6 sessions in the 3 session slots 15:15:40 yes I wanted to propose more sessions than we have time for do you all can downvote the ones that you don't like 15:15:50 we can still discuss those topics in the meetup 15:16:02 sure 15:16:16 In a few minutes I'll ask about each one and see if we can collect the votes 15:16:22 but first I had some questions 15:16:43 to answer vponomaryov: yes 15:17:14 we can spend some more time sketching out specifics if it's a topic people are interested in -- I have a lot of ideas 15:17:33 so vponomaryov, you put a few things on the etherpad I didn't understand 15:17:45 * share groups as server groups in Nova 15:17:45 * proper calculation of snapshot size? 15:18:13 can you briefly explain those 15:18:13 * share groups as server groups in Nova - create several shares at a time and group them somehow 15:18:56 * proper calculation of snapshot size? - snapshot takes quota of share equal to its size and does not allow to spawn share from it with less size than snapshot has defined in manila 15:19:48 okay 15:19:50 first one is useful when we know that we want create N shares 15:20:02 and we will use them for some specific case 15:20:11 I can see pros/cons for the first one 15:20:16 vponomaryov: Is there a way you have thought about grouping the shares? 15:20:27 it seems worth of discussion -- but do you think it would make a good design summit session? 15:20:49 bswartz: need collect desires of end users 15:21:05 bswartz: this idea pop up from WFA project 15:21:10 okay 15:22:23 and on the second topic, I feel like we discussed it before and reached a conclusion we were okay with 15:23:01 since we can't know whether backends will store snapshots efficiently or not, we charge them the full quota just in case 15:23:25 bswartz: we can allow force usage in cases we know it for sure 15:23:37 asking backends to report how much space a snapshot consumes requires a lot more implementation in the drivers 15:23:55 and worse, it could change over time 15:24:44 I suppose there could be an upper bound we could report that's less than the total share size 15:24:52 is that what you're going for? 15:25:04 bswartz: yes 15:25:20 for example we can have size of 100Gb for share and snapshot will take additional 100Gb 15:25:30 this huge leak of quota 15:25:46 and use some couple of Gb indeed 15:25:55 but some backends may literally require 100GB for the snapshot 15:26:38 maybe we should look at doing this in an extension, or with some way for the administrator to turn it off 15:27:14 I like the idea, but it creates some cracks in the abstraction we're trying to create 15:27:29 so, we can conclude that this topic does not have strict borders, though I added it 15:27:48 yes -- do you think it makes sense for a design summit talk? 15:28:26 Not sure, there are important topic too 15:28:33 okay 15:28:34 s/topic/topics/ 15:28:46 bswartz: is it still possible to add topics to the pad? 15:28:47 we can "live" with it 15:28:55 but can not, for example, without image 15:29:02 csaba: if you have an idea mention it here 15:29:50 so, topic with image should be in top of priority, I think 15:30:05 bswartz: do you need to choose 3 topics and put them down here? http://kilodesignsummit.sched.org/ 15:30:11 manila image project? you think it should be a design summit session rather than a meetup topic? 15:30:18 xyang1: yes 15:30:24 bswartz: any 15:30:34 xyang1: although we can put 2 topics in one session if we wish 15:30:47 bswartz: I have some gripes with how execution works in manila, but I can't articulate them properly ATM, I need a bit time to think it over 15:30:49 ok 15:31:49 csaba: were you thinking of having a session to review your gripes? or did you just want to bring them to the team meetup? 15:31:49 csaba: if it would not provide gripes we would not have topics =) 15:32:19 we could just have a whole session where everyone stands up and gripes about what they don't like -- it might be cathartic 15:32:24 vponomaryov: the point is to soothe the gripes ;_ 15:33:18 csaba: I'd like to get the design summit schedule nailed down in the next few days, so if you want to propose a session then we need to know urgently 15:33:28 ATM I would not propose this for a whole session 15:33:29 other topics, for the meetup, can wait 15:33:48 bswartz: I can let you know by tomorrow 15:34:06 okay so now I'd like to just go through the list and hear from you +1 or -1 on design summit worthiness 15:34:20 the first one is: Manila Networking 15:34:30 this one gets a +1 from me 15:34:38 +1 15:34:39 +1 15:34:59 +1 15:35:25 sorry the details in the etherpad are sparse -- there's more detail above 15:35:26 okay 15:35:31 next topic: Integration with other projects 15:36:03 Personally I feel like -1 on this, even though it may have broad interest 15:36:13 This topic seems open ended 15:36:29 we would discuss specific integration with * Tempest* Horizon* Devstack* Heat 15:36:42 the discussion could be short though 15:36:58 we could handle this one during the meetup 15:37:12 expose snapshot usage to other project? 15:37:13 but then we won't get members of those other projects 15:37:14 bswartz: I am Ok to leave for meetup 15:37:23 +1/-1? 15:37:23 /leave/leave it/ 15:37:26 (accounting/chargeback purposes) 15:37:26 -1 15:37:50 +1 push it to meetup 15:38:21 jasonsb: that would just be a new feature for our API, assuming we implement a way to query backends for snapshot actual space consumption 15:38:23 good topic, but won't be productive for a session 15:38:23 meetup sounds good 15:38:34 integration refers to when we actually add code to other projects to make them work better with manila 15:38:58 okay we can move on 15:38:59 ah, ok 15:39:07 Next topic: Mount Automation 15:39:50 +1 from me 15:40:04 if you already have concrete ideas, then this should be a good one 15:40:09 bswartz: We definitely need get nailed what exactly do we want to get with this feature 15:40:15 we've never discussed this topic if deep detail, and people keep asking for it 15:40:25 now feels like the time to finally do it 15:40:25 so +1 15:40:47 I do have concrete ideas 15:41:01 and I can easily see this one going the full 40 minutes 15:41:03 +1 15:41:31 ok 15:41:35 +1 15:41:43 next topic: Review features to steal from Cinder 15:42:10 so the exact list of features we cover in this session could be revised, but the idea is to have a multi-topic session on various cinder features 15:42:22 and talk about the specifics of adapting them to manila 15:42:40 stuff like private volume_type/share_type should be pretty trivial I think 15:42:50 whereas manage/unmanage might be quite complicated given the differences 15:43:20 I'm on the fence about this one 15:43:22 +1 15:43:59 -1 15:44:10 rushil what would you prefer? 15:44:21 This seems to be more than we can cover, not focused 15:44:30 bswartz: Don't think this needs to go for a full session when this has already been covered in cinder 15:44:38 xyang: well we could focus it 15:44:42 bswartz, ability to manage existing share would be good to have, as that would help manila takeover a existing implementation 15:44:54 we could pick 2 or 3 high priority ones and make the session about those 15:45:02 bswartz: We could cover it, but don't need to dedicate a whole session to it 15:45:21 bswartz, vponomaryov I see you said share_type is available, does that mean Manila supports the multi-backend as Cinder does (1 m-shr process per backend) ? 15:45:26 rushil1: actually we could take a single one of those and turn it into a whole session if we wanted 15:45:34 stuff like backup and replication are HUGE topics 15:45:51 deepakcs: yes 15:46:11 deepakcs: although currently we use the term "volume_type" not share_type 15:46:14 bswartz: I get that. But I feel the audience would be cinder aware and should be aware of the concepts being covered. 15:46:34 yes, those topics are huge, each one will be one session. we definitely need to nail down 15:46:50 okay so this one needs reworking if we're going to include it 15:46:51 rushil1: the goal to get know what we can adapt and why we should do it 15:47:57 we have to do the activity of reviewing all these and scoping/prioritizing them 15:48:07 but maybe that's not appropriate to do in paris 15:48:28 maybe once we have concrete proposals for each one we should consider presenting them 15:49:00 I'll propose that we do that reviewing/scoping/prioritizing in a future meeting 15:49:24 okay since we're running out of time let's move on 15:49:29 next topic: Access Groups 15:49:31 i would be interested in hearing about thoughts on DR at the pod 15:49:31 straight forward ones don't need a session 15:49:40 if anybody is interested 15:49:53 jasonsb: sure 15:50:22 access groups is something that's very concrete and could merge during kilo-1 15:50:33 I think it deserves a session 15:50:39 bswartz: but this is related to v2 15:50:43 is it interesting enough though? 15:51:13 vponomaryov: it could be done in v1 -- all the APIs are new and don't break anything 15:51:21 +1, I think access groups deserves a session. 15:51:39 bswartz: it contradicts to current share access approach, is not it? 15:51:48 it extends the current approach 15:52:09 just makes it possible to apply access rules to multiple shares at a time 15:52:27 bswartz: ok, I just did not see implementation you are talking about 15:52:55 and makes it possible for access rules to change in response to events rather than explicit API calls 15:53:32 the driver interface would have to change but there would be no API breakage 15:53:51 if this becomes a session we will post the details in advance 15:54:18 any last opinions? 15:54:19 is there issues for it? 15:54:20 we have 1 more 15:54:27 +1 for access groups 15:54:41 vponomaryov: hopefully no, but the session would allow people to raise them if they had issues 15:55:13 Ok, I don't mind 15:55:16 cknight worked on this during juno and I think he got most of the design decisions right 15:55:28 ameade too if he's here 15:55:35 okay last one 15:55:41 topic: Manila Scheduler 15:56:06 we sort of discussed this one already 15:56:14 I'm on the fence about this one 15:56:15 low priority 15:56:23 you say -1? 15:56:56 -1 because there are should be more prioritized thing, I guess 15:56:57 anyone else thing we *should* make a session about scheduler enhancements? 15:57:08 s/thing/think/ 15:57:12 I think if we have time we can go over it, otherwise it is fine to skip 15:57:16 This can be discussed, but not a whole session. -1 15:57:29 or discuss about it later in meetup 15:57:32 okay I think that one is clear 15:57:55 so just to review, the 3 highest vote getters (by my count) are: 15:57:56 actually the pool one is a scheduler change as well 15:58:01 Manila Networking 15:58:06 Mount Automation 15:58:12 Access Groups 15:58:23 bswartz: +1 15:58:32 i'm fine with that 15:58:36 sounds good 15:58:51 integration and scheduler we will discuss privately, and the cinder feature list needs more work before we'd have anything to present 15:59:25 okay so I'll watch the ML and IRC for any amazing new ideas 15:59:52 but baring something new that everyone wants those 3 will be the proposal 16:00:02 sorry we ran out of time for open discussion 16:00:15 thanks all 16:00:19 thanks 16:00:22 thanks 16:00:23 thanks 16:00:26 thank all 16:00:28 bye 16:00:30 #endmeeting