17:06:08 <nikhil> #startmeeting quotas-wg
17:06:10 <openstack> Meeting started Tue May 31 17:06:08 2016 UTC and is due to finish in 60 minutes.  The chair is nikhil. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:06:11 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:06:14 <openstack> The meeting name has been set to 'quotas_wg'
17:06:17 <nikhil> #topic agenda
17:06:24 <nikhil> There's no formal agenda for today
17:06:43 <nikhil> I will just jot down the plan for 3 weeks proposed by Vilobh earlier.
17:06:50 <nikhil> and we can have the open discussion
17:06:58 <nikhil> #topic plan for next 2-3 weeks
17:07:00 * DuncanT waves
17:07:08 <nikhil> ohai DuncanT
17:07:23 <nikhil> #1. Mysql driver (vilobh working on a prototype)
17:07:46 <nikhil> #2. Tests (unit, functional) (need volunteers)
17:08:21 <nikhil> #3.1 Docs (info from spec will go here when approved
17:08:43 <nikhil> #3.2 generation-id documentation - how consuming projects should use/consume it; what changes the consuming project need to do to use delimiter
17:09:36 <nikhil> #4 Design (objects to be added to the lib, like resrouce and usage objects) (those that are missing)
17:10:16 <nikhil> itisha may have some cycles and I want here to help with testing. But the rest of the work is open floor if people wanted to pick and contribute.
17:10:22 <DuncanT> 3.2 would be useful ASAP - without that it's really tricky to get an idea of impact
17:10:30 <nikhil> ++
17:10:40 <sudipto> nikhil, I have an initial doc up on generation id here: https://review.openstack.org/323415 even though it's completely a WIP - that i added today and would have some more material tomorrow.
17:10:43 <nikhil> Thanks to sudipto, it's now a WIP.
17:11:06 <nikhil> #info gen_id documentation work in progress: https://review.openstack.org/323415
17:11:17 <DuncanT> I've not got any free cycles really, though I've got an experimental tree that cuts reservations out of the current cinder code as prep
17:11:27 <DuncanT> sudipto: Awesome, thanks!
17:12:34 <nikhil> DuncanT: that's interesting, any link for us to see?
17:12:34 <sudipto> DuncanT, nikhil i plan to write 2-3 docs this week - 1. Gen id stuff. 2. about delimiter's data model 3. How other projects could consume it. Would love to hear about comments/reviews - to ensure I have the right idea about it :)
17:12:55 <DuncanT> nikhil: Not yet, but I'll github it this week and post a link
17:13:06 <nikhil> ah nice, thanks DuncanT
17:13:13 <nikhil> any other comments/questions before we move to open discussion?
17:13:23 <nikhil> sudipto: that's pretty awesome!
17:13:24 <DuncanT> Time is not my friend at the moment :-(
17:13:29 <nikhil> :)
17:13:37 <sajeesh> sudipto:thanks...
17:14:10 <nikhil> #topic open discussion
17:14:29 <nikhil> sudipto: do you mind giving us a tl;dr; of what's there for docs?
17:14:56 <nikhil> (just to get some discussions started)
17:15:39 <sudipto> you mean the review above or in general?
17:15:46 <sudipto> nikhil, ^
17:15:52 <nikhil> sudipto: I meant in general
17:16:04 <nikhil> (based on your research so far)
17:16:54 <sudipto> nikhil, ok - basically - i think the docs should have those 3 citations i mentioned above. and most likely a project specific implementation detail...
17:17:06 <sudipto> (ofcourse left to open discussions)
17:17:17 <DuncanT> We need to add some scale tests to the plan fairly early on I think... what does the performance look like with 1k, 5k, 10k, 100k, 1M entities being tracked for quota
17:18:02 <DuncanT> Need to find out if the cache-free approach actually scales... the cache was added for scaling reasons though I can't see any benchmarks done at the time
17:18:22 <sudipto> It's also important for people to know - what the impact of this project will be...
17:18:36 <sudipto> +1 DuncanT
17:19:26 <nikhil> DuncanT: good point... I guess we need some dev testing env for this. I think we could so a bit of testing for glance but 1M entries might be a stretch ..
17:19:42 <DuncanT> Can we write a fake service for testing? Just use delimiter to track widgets
17:20:11 <DuncanT> I do cinder tests with 1M fake db entries now, it's trivial to just insert things into the DB
17:20:34 <nikhil> DuncanT: I think a simplistic patch to glance #images in queued status (only DB record) would do  exactly that.
17:20:50 <nikhil> sudipto: ^ what do you think?
17:20:58 <sudipto> nikhil, yeah that could be achieved.
17:21:23 <nikhil> (also, we can test it on devstack)
17:21:54 <sudipto> nikhil, i would not necessarily commit on doing that this week, since i would like to spend sometime on the docs, but definitely add it to the to-do list.
17:21:55 <DuncanT> It's just to make the count queries hit some big record counts - we should be testing at least 10x the biggest record count we expect to see in the wild
17:23:00 <sudipto> nikhil, DuncanT based on Jay's mail - is that how we would eventually expect delimiter to be consumed?
17:23:18 <sudipto> referring to this NOVA example.
17:24:46 <DuncanT> sudipto: That's the current plan, yes. I'd like some confidence that it fly ASAP, so that we know if we need to think again (and go have a long debate with Jay no doubt)
17:24:47 <sudipto> referring to http://lists.openstack.org/pipermail/openstack-dev/2016-May/095035.html
17:25:21 <nikhil> DuncanT: right on
17:25:57 <sudipto> DuncanT, sounds good.
17:25:58 <nikhil> sudipto: no worries, I think the lib mysql support may take this week. but this can be put in our glance bucket and work with itisha too.
17:26:49 <DuncanT> Having figures and demo code is a good way to deal with any nay-sayers, and to give us confidence before investing too much
17:27:12 * sudipto smells DuncanT is referring to a PoC
17:27:15 <sajeesh> DuncanT +1
17:27:39 <DuncanT> sudipto: A PoC plus an honest attempt to break it / benchmark it
17:27:54 <DuncanT> sudipto: A friendly PoC is a different thing entirely
17:28:03 <sudipto> DuncanT, heh gotcha!
17:28:13 <sudipto> any volunteers for that?
17:28:36 <DuncanT> I might have cinder in some shape to do it, maybe. Got a lot else on at the moment though :-(
17:28:51 <sudipto> DuncanT, I will work with you - will that be alright?
17:29:02 <DuncanT> certainly
17:29:09 <sudipto> I too have some other commitments, but i will make sure that i keep coming back to you about it.
17:29:39 <DuncanT> I'll get a github tree up this week with what I've got - I find them easier than gerrit reviews for messing with
17:30:04 <sudipto> DuncanT, great! That could be our starting point I guess.
17:30:16 <sudipto> DuncanT, I will try to prototype Nova.
17:31:49 <DuncanT> Sounds good
17:31:57 <nikhil> any other topic for open discussion?
17:32:20 <nikhil> going 1
17:32:22 <nikhil> going 2
17:32:29 <nikhil> Thanks all!
17:32:37 <nikhil> Let's reconvene next week.
17:32:46 <nikhil> #endmeeting