18:04:34 <rkukura> #startmeeting networking_policy 18:04:34 <openstack> Meeting started Thu Aug 31 18:04:34 2017 UTC and is due to finish in 60 minutes. The chair is rkukura. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:04:35 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:04:38 <openstack> The meeting name has been set to 'networking_policy' 18:04:41 <tbachman> rkukura: thx! 18:04:58 <rkukura> hi tbachman, annak[k] 18:05:15 <rkukura> So Sumit is not available today 18:05:55 <rkukura> annakk: Do you want to start with hour horizon question? 18:06:02 <annakk> sure 18:06:34 <annakk> I was trying to check out ui patches in review, and discovered half of the ui is not working for me 18:06:41 <tbachman> :( 18:06:54 <tbachman> annakk: is this stable/ocata? 18:06:55 <annakk> So I wondered whether its ocata to blame, me to blame or something else.. 18:06:58 <annakk> yes 18:07:15 * rkukura wonders if stable/newton still works 18:07:58 <annakk> did anyone try it after newton/ocata sync? 18:08:01 <tbachman> annakk: did it work for you somewhat recently, or is this the first time you’ve tried the UI? 18:08:06 * tbachman has not 18:08:24 <tbachman> annakk: we’re working on creating a stable/ocata setup for testing 18:08:28 <rkukura> I have not tried horizon in a long time 18:08:29 <tbachman> but we’re not there yet 18:08:32 <annakk> I think its the first time. I looked at it a while ago but I don't think I ever tried to create things 18:08:46 <tbachman> I just looked at the stable/ocata merges 18:08:55 <tbachman> ah 18:08:57 <tbachman> wrong project 18:08:58 <tbachman> just a sec... 18:09:46 <tbachman> there are no patches into stable/ocata for group-based-policy-ui 18:09:53 <tbachman> which is interesting 18:10:36 <rkukura> tbachman: are there any ocata sync patches in master? 18:10:51 <tbachman> I must have made an incorrect query 18:11:00 <tbachman> b/c I searched for merges in master and came up with nothing 18:11:48 <annakk> there is this one https://review.openstack.org/#/c/483138/ 18:12:04 <annakk> but its minimal 18:12:18 <tbachman> there is one after it to fix a delete button 18:12:23 * tbachman fixed his query 18:12:44 <tbachman> other than that - march 16th, april 4th, may 17th 18:13:09 <rkukura> the delete button patch needs back-porting I guess 18:13:33 <tbachman> rkukura: ack 18:13:51 <rkukura> wasn’t there something about an upcoming django version dropping some backward compatability we were depending on? Maybe that happended. 18:14:08 <tbachman> rkukura: news to me 18:14:30 <tbachman> rkukura: was that on openstack-dev? 18:14:52 <rkukura> https://review.openstack.org/#/c/493830/ 18:15:22 <annakk> I think some of the patches in review deal with deprecations 18:15:39 <rkukura> If annakk has django 1.10, that patch might fix it 18:16:19 <annakk> OK, will try 18:16:23 <tbachman> annakk: when you say “half of the UI” — can you elaborate? 18:16:42 <rkukura> looks like there are a number of patches waiting review on master 18:17:10 <rkukura> #link https://review.openstack.org/#/q/status:open+project:openstack/group-based-policy-ui,n,z 18:17:51 <rkukura> annakk: Are the recent ones from your team? 18:18:12 <tbachman> ultimum.io 18:18:33 <annakk> tbachman: I wasn't able to create groups, rule sets.. 18:18:43 <annakk> rkukura: you mean the patches? no 18:19:30 <tbachman> are the buttons/selectors not there? Or is that nothing happens when you try to select/use them? 18:20:37 <annakk> tbachman: the buttons are there, but cause errors (i don't think the request reaches our server) 18:21:00 <tbachman> annakk: okay, that helps. Thanks! 18:22:10 <annakk> tbachman: np. I don't understand much in UI sadly 18:23:00 <annakk> I'll see if the patches from ultimum improve it 18:23:02 <tbachman> annakk: sorry we don’t have a better answer for you now 18:23:40 <annakk> tbachman: sure, I just wanted to verify I wasn't missing something 18:24:15 <rkukura> ok, annakk, please let us know if you are using django >= 1.8, and if the recent patches help 18:24:43 <annakk> rkukura: ok 18:24:45 <rkukura> and we should be reviewing those a bit more proactively 18:24:52 <tbachman> rkukura: ack 18:25:01 <rkukura> anything else on horizon? 18:25:09 <annakk> not from me 18:25:22 <rkukura> so lets discuss db migration a bit 18:25:45 <tbachman> #topic DB migrations 18:25:51 <tbachman> ah, I’m not chair 18:25:55 <rkukura> #topic DB migrations 18:25:59 <tbachman> rkukura: thx! 18:26:11 <rkukura> #chairs tbachman, rkukura, annakk 18:26:15 <tbachman> :) 18:26:23 <annakk> :) 18:26:32 <rkukura> not sure if that works 18:26:38 <tbachman> it does 18:26:55 <tbachman> #link https://review.openstack.org/#/c/499240/ <= gerrit where Sumit discusses options for backporting DB migrations 18:27:18 <rkukura> as you probably know, GBP has historically been more aggressive with back-ports than most projects 18:27:39 <rkukura> we’ve been back-porting new features across several stable branches 18:28:05 <rkukura> but have not been back-porting all patches to every branch 18:28:21 <rkukura> so the result is we have different chains of DB migrations for each branch 18:28:39 <rkukura> this works ok for updates within a stable branch 18:29:04 <rkukura> but will result in some migrations being missed when upgrading to a new release 18:30:41 <rkukura> I first noticed this when reviewig a mitaka back-port 18:30:43 <rkukura> #link https://review.openstack.org/#/c/499185/ 18:30:51 <annakk> rkukura: by "within stable branch" do you mean upgrading to newer version of same branch? 18:31:01 <rkukura> annakk: yes 18:31:17 <rkukura> so the chain might make sense within stable/mitaka 18:31:38 <annakk> but is this supposed to happen? 18:31:48 <rkukura> and a new migration might be at the HEAD of both stable/mitaka and stable/newton 18:32:34 <rkukura> but a migration skipped on stable/mitaka will not be run when upgrading to stable/newton, even though it is in the stable/newton chain 18:33:14 <rkukura> I think the OpenStack stable branch policy avoids this kind of issue by forbidding schema changes, along with API changes, etc. 18:33:39 <tbachman> forbidding *backports of* schema changes… 18:33:51 <rkukura> tbachman: can you summarize Sumit’s proposal? 18:33:58 <tbachman> rkukura: I can try :) 18:34:27 <tbachman> I believe SumitNaiksatam’s proposal is to backport all DB migrations in order to keep them consistent across stable branches 18:34:39 <tbachman> This means care needs to be taken when creating DB migrations 18:35:27 <tbachman> SumitNaiksatam also said it’s possible for the DB migration to be a No-Op, if for some reason the actual migration would be determintal as a backport 18:35:32 <rkukura> seems to me that we need to decide how to prevent issues going forward, but also need to fix the issues that already exist 18:35:38 <annakk> tbachman: backport migrations only, without the actual feature? 18:35:44 <tbachman> annakk: ack 18:36:15 <rkukura> we wouldn’t want it to be a no-op, or else we’d need the real version of the migration somewhere else on the chain 18:36:21 <annakk> are we sure all the migration are extend-like? 18:36:34 <tbachman> annakk: they aren’t 18:36:48 <tbachman> for example, one of the NFP patches has some more significant changes 18:36:52 * tbachman pulls up patch 18:36:53 <annakk> tbachman: isn't that a problem? 18:38:01 <tbachman> #link https://github.com/openstack/group-based-policy/commit/dc25f7154314d4b0d143a3f55d4ea4cc8a3eafe2#diff-9ae7c407a57ed5bb5e916dbbf4530e8b <== NFP patch, where a primary key is added and columns are altered 18:38:06 <tbachman> annakk: that was my thought as well 18:38:16 <tbachman> that said, I know SumitNaiksatam spent some time thinking about this 18:38:23 <tbachman> so maybe he has some thoughts there 18:38:40 <tbachman> The above patch seemed to be problematic as a backport w/o code 18:39:06 <rkukura> I doubt we can in general back port migrations without the correponding code 18:39:23 <tbachman> we may have to shelve this discussion until next week, where SumitNaiksatam can shed some light on the issues 18:39:48 <rkukura> shorter term, I think we need to resolve some of the current issues 18:40:32 <rkukura> For example, there are two migrations skipped on stable/mitaka that are needed on stable/newton, but will be skipped on an upgrade 18:41:30 <annakk> I think someone mentioned that it we re-do a migration that already exist, it will not cause an error? 18:42:43 <rkukura> annakk: I think migrations need to be written to handle that, for example checking if the new table or column already exists before adding it 18:43:35 <annakk> maybe that's a tolerable solution for short-term 18:43:52 <tbachman> #link https://review.openstack.org/#/c/498940/4/gbpservice/neutron/db/migration/alembic_migrations/versions/4c0c1e2c0160_ha_ip_address_to_port_id_association.py@32 <== example of checking for an existing table before applying migration 18:44:40 <rkukura> So one option is to make the skipped migration idempotent and move it to towards the head of the chain in the newer stable branch 18:47:18 <rkukura> I think there are 4 migrations total that are skipped on some stable branch and needed on a newer branch 18:47:43 <tbachman> rkukura: where does that patch go in master then? 18:48:15 <tbachman> it seems like it has to go to the point in the migration in the furthest backport 18:48:21 <tbachman> across all branches 18:48:30 <rkukura> tbachman: if we make a migration idempotent, I guess we should move it towards the head on all newer branches, including master 18:49:02 <tbachman> Didn’t we say that would cause migrations to be skipped on upgrades? 18:49:19 <tbachman> if not all branches have the same migrations? 18:50:15 <annakk> sorry but what is the reason for aggressive backports? 18:50:17 <rkukura> Adding a migration in the middle of chain on a stable branch doesn’t help deployments that are using that stable branch 18:50:29 <rkukura> annakk: good question! 18:50:46 * tbachman thinks we need SumitNaiksatam for this answer 18:51:34 <annakk> maybe we could consider changing the approach going forward.. 18:52:19 <rkukura> annakk: we should consider that, or at least come up with a policy going forward that will prevent these issues 18:52:56 * tbachman notes the time 18:52:59 <rkukura> but we also need to fix things so that current deployments don’t break when upgrading to newer stable versions 18:53:23 <rkukura> doesn’t seem we are going to resolve either the short term or longer term issue here right now 18:53:32 <tbachman> rkukura: ack 18:53:38 <annakk> ack 18:54:11 <tbachman> I think we need SumitNaiksatam here to provide his insight/perspective on this 18:54:21 <tbachman> we’re left guessing at his intent 18:54:27 <tbachman> which doesn’t seem safe 18:54:31 <rkukura> but I think we are likely to try to come up with some sort of fix for the short term problem (existing branches and migrations) over the next few days, and can discuss this on email and/or gerrit 18:54:43 <tbachman> rkukura: sounds good 18:54:52 <rkukura> annakk: are you ok with that? 18:54:56 <annakk> (also with backporting we might hit issues with neutron tables, like FK to non-existing tables..) 18:55:03 <annakk> rkukura: yes, sure 18:55:19 <rkukura> annak: good point about FKs, etc. 18:55:39 <rkukura> ok, anything else to cover before we wrap up? 18:55:50 <tbachman> rkukura: nothing from me 18:56:17 <annakk> neither from me 18:56:25 <rkukura> ok, thanks everyone! 18:56:29 <annakk> thanks! 18:56:33 <rkukura> #endmeeting