17:33:53 <dougwig> #startmeeting networking_lib
17:33:54 <openstack> Meeting started Wed May 25 17:33:53 2016 UTC and is due to finish in 60 minutes.  The chair is dougwig. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:33:54 <dougwig> hi all
17:33:56 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:33:58 <openstack> The meeting name has been set to 'networking_lib'
17:34:00 <armax> here
17:34:02 <armax> and not here
17:34:42 <dougwig> haha, hi armax.
17:34:46 <dougwig> #topic Brief status
17:34:52 <dougwig> boden, how goes the agent stuff?
17:35:03 <boden> I have 3 patches out for review
17:35:42 <boden> And a few more I could throw out there as well, but have been waiting for a few others to merge to reduce a (patch) dependency chain
17:35:52 <boden> That’s what I got : )
17:36:01 <dougwig> are they all ready for review, or still wip?
17:36:08 <boden> they are ready
17:36:21 <HenryG> ok, I will start reviewing
17:36:26 <boden> if a WIP I already try to put that in the commit subject
17:36:32 <boden> ok thanks
17:36:33 <dougwig> ok, good.
17:36:48 <dougwig> henry, you were going to rebase your db stuff?
17:37:14 <HenryG> the context & policy patch is rebased
17:37:45 <boden> I think I had some outstanding comments on the your db patch HenryG
17:38:22 <HenryG> the db one I am still working on to avoid OVO implications
17:38:47 <dougwig> alright, let's get to a fun topic
17:38:51 <dougwig> #topic Gate jobs
17:39:03 <dougwig> we had the 0.2.0 release this week, and it busted neutron.
17:39:16 <dougwig> armax and i discussed gating the lib on more of the neutron jobs.
17:39:26 <dougwig> any objections or comments to that?
17:39:36 <HenryG> +1
17:39:45 <boden> agreed +1
17:40:13 <dougwig> i have cycles to look at that next week, unless someone wants to pick up the torch sooner?  but i'm dang well taking memorial day weekend off.  :)
17:40:33 <dougwig> what jobs should we add?  definitely unit.  we already do -full.
17:40:51 <HenryG> not functional, too flakey
17:40:54 <boden> dougwig I don’t mind helping where I can, but maybe a little slow to ramp-up
17:41:24 <kevinbenton> Can we just add them all as periodic or experimental?
17:41:26 <dougwig> boden: thanks, and no worries, you help with the reviews to get to know how those pieces work.
17:41:27 <armax> dougwig: unit would suffice, HenryG and I thought about a couple of strategies
17:41:29 <kevinbenton> We K
17:41:38 <dougwig> armax/HenryG - shoot
17:41:43 <kevinbenton> We only really need to pay attention at release
17:41:59 <dougwig> kevinbenton: i'd prefer that master was always pullable, IMO.
17:42:09 <armax> dougwig: one step is getting unit tests to co-gate
17:42:38 <armax> dougwig: this is inline with what is suggested in the stadium spec too
17:42:40 <dougwig> armax: by co-gate, you mean running lib unit tests for neutron changes?
17:42:55 <armax> dougwig: no, the otehr way around
17:43:03 <dougwig> ok, yes, definitely.
17:43:04 <armax> dougwig: running neutron’s unit tests on neutron-lib changes
17:43:17 <dougwig> arguably we should run all stadium project unit tests.
17:43:24 <armax> dougwig: we could do so on a periodic basis rather than a check basis
17:43:31 <armax> dougwig: but that’s splitting hair
17:43:39 <dougwig> unit tests are cheap/fast compared to the -full that's already in there.
17:43:47 <armax> dougwig: indeed
17:43:54 <dougwig> we can also likely make a job that runs all the projects unit tests in one job.
17:43:55 <armax> dougwig: however, when it’s time for a release
17:44:13 <armax> we may be potentially break other projects too like functional, fullstack, grenade, multi
17:44:44 <armax> dougwig: and it would be nice if we could figure out a way to test the whole lot at once before we pull the trigger during a release
17:45:05 <armax> eventually the level of decoupling will be adequate enough that we no longer need that
17:45:08 <dougwig> sort of a 'check release' type thing?  i'm sure we could abuse experimental for that. or run them all the time.
17:45:17 <armax> dougwig: indeed
17:45:30 <dougwig> will infra add a new pipeline for that, i wonder?
17:45:39 <armax> dougwig: something along those lines, I think I need a little more brainstorming on the actual strategy though
17:45:49 <dougwig> i'm not opposed to running all of the neutron jobs on every change, to be honest.
17:45:55 <armax> dougwig: we could gauge the appetite
17:46:05 <armax> dougwig: they have an experimental-tripleo one
17:46:14 <armax> dougwig: but it might not be necessary
17:46:35 <armax> dougwig: for neutron-lib you mean?
17:46:44 <dougwig> dhellmann: are you around? are there any provisions for doing an extra big set of test jobs prior to a lib release, or are they usually all just put into check/gate and run every time?
17:46:55 <dougwig> armax: yes.
17:47:05 <armax> dougwig: it seems like an overkill
17:47:33 <armax> dougwig: until a new release any change is ineffective
17:47:49 <dougwig> armax: unless you run master.
17:48:05 <armax> dougwig: no-one would run neutron-lib master
17:48:55 <dougwig> i thought that about neutron, too.
17:49:03 <armax> to start off running unit tests on check (or periodic) would suffice, then we can iterate and plug any remaining hole as it shows up
17:49:43 <dougwig> i'd vote for unit tests in check, because they're quick/cheap.
17:49:51 <armax> in the meantime if we can figure out a way to check the effect of a new release on a ad hoc basis
17:50:06 <armax> for all neutron’s jobs that’s a bonus
17:50:28 <armax> dougwig: ok, I was going to put that together, unfornately it’s not that simple than adding a single line to project-config
17:50:53 <HenryG> why not?
17:51:00 <dougwig> armax: no, because you need the repo on disk, and to not pull requirements. i know how to set those up, though.
17:51:10 <armax> HenryG: ^
17:51:12 <armax> dougwig: ack
17:51:19 <HenryG> hmph
17:51:21 <armax> we’d need to do something like the periodic oslo jobs
17:51:27 <armax> dougwig: right?
17:52:28 <dougwig> for devstack, you use LIBS_FROM_GIT.  for unit, i forget what i did.  but i usually use the oslo templates when i can (which is what the master/master job in neutron experimental is using.)
17:52:51 <armax> dougwig: ack
17:53:04 <armax> so maybe you and I can reconvene later offline to agree on who does what
17:53:10 <dougwig> ok, sounds good.
17:53:18 <armax> dougwig: thanks
17:53:28 <dougwig> my only invariant is that i *am* installing solar on the roof of my RV this weekend.
17:53:45 <dougwig> :)
17:53:48 <armax> dougwig: who said I was going to work over the weekend?
17:54:01 <dougwig> that's been my last two months.
17:54:05 <dougwig> #topic Open discussion
17:54:08 <dougwig> any other topics for today?
17:54:24 <boden> nothing from me
17:54:45 <HenryG> Enabling depends-on: a neutron-lib patch in neutron
17:55:12 <HenryG> The whole tox_install.sh thingy
17:55:28 <HenryG> Can it be done?
17:56:07 <dougwig> so you run the units with non-pypi lib, i take it?
17:56:35 <dougwig> it'd be easier to flip the arrow and put the jobs on the lib side, so we have that mess in one spot.
17:59:11 <HenryG> yeah, that makes sense
18:00:37 <HenryG> nothing else from me
18:03:33 <dougwig> ok, let's blast out of here.
18:03:35 <dougwig> #endmeeitng
18:03:39 <dougwig> #endmeeting