17:00:12 <melwitt> #startmeeting nova_cells
17:00:13 <openstack> Meeting started Wed Oct  7 17:00:12 2015 UTC and is due to finish in 60 minutes.  The chair is melwitt. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:00:14 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:00:16 <openstack> The meeting name has been set to 'nova_cells'
17:00:25 <melwitt> anyone around for cells today?
17:00:31 <belmoreira> o/
17:01:12 <melwitt> ping bauzas
17:01:30 <melwitt> #topic open discussion
17:01:58 <lalitd> o/
17:02:00 <belmoreira> I have some questions
17:02:02 <melwitt> we'll just do open discussion today
17:02:07 <belmoreira> :)
17:02:32 <belmoreira> when can we resubmit the spec again for the flavors?
17:03:03 <melwitt> I think you can do that now, but I thought I saw one already re-approved
17:03:18 * bauzas waves lately
17:03:33 <belmoreira> oh great, maybe lalitd?
17:03:38 <melwitt> I'm looking for it
17:03:58 <melwitt> this one? https://review.openstack.org/#/c/231858/
17:03:59 <lalitd> yes its regarding https://review.openstack.org/#/c/213041/
17:04:31 <belmoreira> great... miss communication between us. sorry
17:04:56 <melwitt> cool, just wanted to make sure there wasn't another flavors one you were talking about
17:04:58 <bauzas> mmm, the spec is approved for mitaka ?
17:05:06 <bauzas> I see it merged
17:05:13 <lalitd> melwitt: the patch is passing all unit and functional tests but still problem in grenade and tempest tests
17:05:52 <belmoreira> bauzas: is there something that you would like to change?
17:06:01 <bauzas> belmoreira: nope just asking
17:06:29 <bauzas> lalitd: ping me around 6.30pm UTC, I'll look at the grenade job
17:07:48 <lalitd> bauzas: thanks but it will be 12:00 AM in india :)
17:08:01 <bauzas> lalitd: then tomorrow afternoon your time
17:09:26 <bauzas> okay, moving on ?
17:09:53 <melwitt> sure. anyone have any other topics?
17:09:56 <bauzas> so, now that we have mitaka open, we should discuss on the features we want to implement
17:10:13 <bauzas> I'm just looking at the specs
17:10:39 <vineetmenon> john just approved that..
17:11:22 <bauzas> I know that alaski_out was thinking about rewriting https://review.openstack.org/#/c/214792/
17:12:42 <bauzas> so we have the persisting spec
17:12:43 <lalitd> bauzas: can we meet tomorrow at same time 1030 IST ...
17:13:03 <lalitd> tomorrow afternoon some of my office meeting is there
17:13:32 <bauzas> lalitd: too early for me (CEST)
17:13:42 <bauzas> lalitd: but I'll look at your change tonight
17:14:26 <bauzas> anyway, so my point is about the mitaka design summit and see whether we need to discuss on something not already accepted
17:14:31 <bauzas> melwitt: thoughts on that?
17:15:46 <melwitt> well, one thing I was wondering recently is, are we going to be migrating everything to the api db and do we need specs for every piece of that?
17:16:15 <bauzas> melwitt: if we use the objects, I don't think so
17:16:25 <bauzas> since it could be a live migration
17:16:36 <lalitd> bauzas: thanks , ok then I will ping you at 6:30pm UTC
17:16:39 <bauzas> I mean an online migration
17:17:05 <bauzas> lalitd: nevermind
17:17:13 <melwitt> bauzas: I was thinking of the schema migrations, creating the tables in the api db
17:17:46 <bauzas> melwitt: then a spec is needed I guess
17:17:58 <bauzas> melwitt: but we could ask johnthetubaguy
17:18:11 <melwitt> I setup devstack the other day and so far we have cell_mappings host_mappings instance_mappings in the api db
17:19:02 <bauzas> melwitt: right, that still leaves the global tables like aggregates to move on
17:19:03 <lalitd> melwitt: then what will happen to old migration scripts which are part of patches
17:19:11 <lalitd> just for info
17:20:49 <melwitt> lalitd: what do you mean by the old migration scripts?
17:20:50 <bauzas> melwitt: anyway, I guess we should still consider what to move, and that's why some specs could be needed
17:21:34 <bauzas> melwitt: I'm also concerned by the online data migration and the contract action
17:21:52 <bauzas> melwitt: so, I think you made a very valid point
17:22:50 <melwitt> bauzas: yeah, me too. I'm not clear on what we need to do here. I thought we're going to have to add the global tables to the api db and not sure how to start that, other than the flavors patches we already have going on
17:24:08 <lalitd> melwitt: in patch https://review.openstack.org/#/c/201606/14 it includes migration script like  https://review.openstack.org/#/c/201606/14/nova/db/sqlalchemy/api_migrations/migrate_repo/versions/004_flavors.py , is it same ?
17:24:10 <bauzas> melwitt: so we need to prepare ourselves to follow the story
17:24:11 <melwitt> and also when the global tables need to be added. that is, is the goal to get a simple nova boot working with a single cell before we get all the global api tables done?
17:24:32 <bauzas> melwitt: that's my thoughts, one single cellsv2 with a single boot
17:24:57 <melwitt> bauzas: right, but is getting all the api db tables done a prerequisite? I'm not clear on that
17:25:35 <melwitt> I was starting to think of all the build steps, "save X in the api db", "pick a cell to schedule", "save Y in the cell db" and those interactions
17:26:12 <bauzas> melwitt: well, for one cell, sure you don't need that
17:26:29 <bauzas> melwitt: but you should still have that done at one moment
17:26:45 <melwitt> bauzas: okay, I think that was my misunderstanding
17:26:49 <bauzas> melwitt: the scheduling bits are the ones that are a bit worrying me for Mitaka
17:27:31 <vineetmenon> bauzas: any updates on scheduling?
17:27:36 <bauzas> melwitt: given our upgrade process, I'd be far in favor of doing the migrations earlier than later
17:27:38 <vineetmenon> is alaski handling those?
17:28:02 <melwitt> lalitd: yeah, I was trying to say we'll need more of them, one for each of the other global tables
17:28:13 <bauzas> melwitt: while the scheduling bits are just an API contraxt
17:28:15 <bauzas> contract
17:28:29 <bauzas> vineetmenon: it's not tracked for the moment
17:28:39 <lalitd> melwitt: okk
17:29:28 <bauzas> vineetmenon: tbc, http://specs.openstack.org/openstack/nova-specs/specs/liberty/approved/cells-scheduling-interaction.html is the only one we merged
17:29:32 <melwitt> bauzas: yeah, true
17:29:54 <vineetmenon> bauzas: so still a long way to go..
17:30:26 <bauzas> vineetmenon: and only a few resources yeah you got it
17:31:26 <bauzas> so, I'd say for mitaka
17:31:36 <bauzas> - db migrations
17:31:45 <bauzas> (incl. flavors)
17:32:02 <bauzas> - scheduling bits given in http://specs.openstack.org/openstack/nova-specs/specs/liberty/approved/cells-scheduling-interaction.html
17:32:08 <vineetmenon> +1
17:32:12 <bauzas> ... and that's basically it
17:32:24 <bauzas> keep it simple
17:32:48 <melwitt> that sounds good to me
17:32:49 <bauzas> because
17:33:02 <bauzas> for db migrations, we need to do as a best effort wayt
17:33:26 <bauzas> for scheduling bits, we need to land the spec object usage + persist it + doing the above spec
17:33:39 <bauzas> just by saying the above, that's already pretty pedantic
17:35:03 <bauzas> in other words, I don't more work than the current planned ones
17:35:08 <bauzas> I don't see*
17:35:31 <bauzas> we just need to make sure that all the existing specs are mitaka-approved
17:35:57 <bauzas> and we should consider some new specs for the global tables we consider needing to be moved to the api db
17:36:11 <melwitt> agreed
17:36:29 <lalitd> +1
17:36:41 <bauzas> okay
17:36:48 <bauzas> I'll need to jump off for a bit
17:36:55 <melwitt> any other topics today?
17:36:58 <bauzas> but I'll be back in a few
17:37:06 <johnthetubaguy> wonders if he can answer a question?
17:37:12 <johnthetubaguy> melwitt: something about a spec?
17:37:37 <bauzas> johnthetubaguy: it was about some db migrations we need to land
17:37:56 <bauzas> since that's db migrations, that requires a spec AFAIU
17:38:02 <melwitt> johnthetubaguy: oh, we were just talking about the cells api db global tables. if we'll need a spec per global table migration, I assumed so
17:38:14 <johnthetubaguy> we usually require a spec for those, mostly just so operators have a chance to tell us we are crazy
17:38:16 <bauzas> like we did for the flavor bits
17:38:22 <melwitt> we were thinking about what are the specs we still need t owrite
17:38:31 <johnthetubaguy> cool
17:38:41 <bauzas> that's why I feel a spec by table isn't overkill
17:38:48 <johnthetubaguy> so feel free to group them, if thats more efficient
17:38:52 <bauzas> becuase we could argue per table
17:39:02 <bauzas> but we could ask for specific fast-approval
17:39:39 <bauzas> again, my main worries are not about the expand bits
17:39:46 <johnthetubaguy> so its easier for me if there is one big one that gets done quickly-ish, but I am open to ideas
17:39:52 <bauzas> rather the online data migration and the contract bits
17:40:09 <bauzas> I feel we can discuss that somewhat later with ala
17:40:11 <bauzas> alaski_out:
17:40:13 <bauzas> heh
17:40:26 <melwitt> yeah. at least so far, with flavors we have the expand and the online migrations in the same spec series
17:40:27 <johnthetubaguy> I would focus on the regular migrations, and make sure they are purely addative
17:40:30 <bauzas> that's not super urgent
17:40:43 <johnthetubaguy> oh, wait, online data migrations right?
17:40:47 <melwitt> the contract, I don't think we're covering at all at this point
17:41:01 <bauzas> johnthetubaguy: sure, but that would still leave an open question about the existing data
17:41:23 <bauzas> johnthetubaguy: eg. if we move the aggregates table
17:41:31 <johnthetubaguy> any online migration needs a nova-manage command to do the migration, with a throttling param
17:41:40 <bauzas> yeah that
17:42:02 <bauzas> I need to drop off now
17:42:03 <bauzas> ttyl
17:42:13 <melwitt> seeya bauzas
17:42:26 <melwitt> johnthetubaguy: that's a good reminder
17:42:35 <lalitd> by bauzas
17:42:38 <johnthetubaguy> yeah, its is an interesting one, so we have a servers table in the API cell now I guess?
17:43:42 <johnthetubaguy> we are assuming folks add the API database I guess, I wonder if we should add the child cell database, that could well be stupid
17:44:11 <melwitt> johnthetubaguy: no, instances will be in the cell db. and in the api db will contain mappings
17:44:51 <melwitt> johnthetubaguy: the current state of the api db has only tables: cell_mappings, host_mappings, instance_mappings
17:45:09 <johnthetubaguy> so I have a feeling alaski was having another idea for the initial create
17:45:27 <johnthetubaguy> having a full instance record, then doing a move to the cell, and deleting it from the API at that point
17:45:52 <johnthetubaguy> so a swap for a full record to a mapping, so its still only ever in one place at once
17:45:59 <melwitt> the request spec is supposed to act that way
17:46:23 <johnthetubaguy> yeah, he may have changed his mind about changing his mind
17:46:34 <johnthetubaguy> sorry, just thinking out loud really
17:47:01 <melwitt> I might be wrong :) I need to consult with alaski again
17:47:19 <vineetmenon> ha
17:47:32 <johnthetubaguy> me too
17:47:33 <lalitd> johnthetubaguy: in https://review.openstack.org/#/c/213041/ patch for flavor migration a new option is added in file https://review.openstack.org/#/c/213041/18/nova/cmd/manage.py
17:48:17 <johnthetubaguy> lalitd: yeah, that looks good
17:48:34 <johnthetubaguy> anyways, didn't mean to extend the meeting, honest!
17:48:45 <lalitd> cool
17:49:08 <melwitt> johnthetubaguy: it's cool, glad you joined :)
17:49:21 <johnthetubaguy> back to the other bit, feel free to submit the specs in whatever way is easiest, let me know what works for you all
17:49:54 <melwitt> johnthetubaguy: will do, thanks
17:49:54 <vineetmenon> cool.. thanks for joining johnthetubaguy
17:50:00 <johnthetubaguy> it sounds like a spec that just has a list of tables that will be copied into the APi database would do the trick :)
17:50:15 <johnthetubaguy> no problems
17:50:24 <melwitt> cool
17:50:47 <melwitt> anything else today?
17:50:47 <johnthetubaguy> so I should really go an cook my dinner, before I have to head off to my tuba/brass band rehearsal!
17:51:06 <johnthetubaguy> thanks for keeping cells stuff moving forward all, great stuff!
17:51:16 <melwitt> johnthetubaguy: nice, have fun
17:51:47 <melwitt> okay, guess that's it. thanks all!
17:51:55 <melwitt> #endmeeting