15:01:32 <sjmc7> #startmeeting openstack search
15:01:33 <openstack> Meeting started Thu May  4 15:01:32 2017 UTC and is due to finish in 60 minutes.  The chair is sjmc7. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:34 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:01:37 <openstack> The meeting name has been set to 'openstack_search'
15:02:07 <sjmc7> good morning
15:02:29 <yingjun> o/
15:03:09 <lei-zh> o/
15:04:25 <sjmc7> small crowd today. ok, let’s get started
15:04:48 <sjmc7> #topic nova cells
15:05:05 <sjmc7> there was some feedback on https://review.openstack.org/#/c/453352/ that i’ll put up a patch for after this, mainly around versioning
15:05:27 <sjmc7> and that inner objects are also versioned which is a bit awkward for consumers but doesn’t matter too much for us
15:05:51 <sjmc7> it might make upgrades a bit difficult
15:06:34 <sjmc7> it looks like most of the nova work is done
15:06:56 <lei-zh> yeah, it seems we have to update code logic every time while notification version upgrades
15:07:17 <sjmc7> i think we’ll just have to make sure we do a final update at the end of a release to the current versions
15:07:24 <sjmc7> i don’t imagine they’ll be changing very frequently
15:08:31 <sjmc7> i don’t know how much testing the nova team have done; i haven’t heard much
15:08:48 <sjmc7> i’ll try to check in with kevin
15:09:45 <lei-zh> I guess there are not so many consumers on versioned notifications
15:09:56 <sjmc7> no.. we may be the first
15:10:17 <lei-zh> cool : )
15:10:34 <lei-zh> anyway, it's a good start
15:10:50 <sjmc7> i think the reality is you’ll need to update all the consumers at the same time as the producers; running lots of versions at the same time is a nice idea but impractical
15:11:08 <sjmc7> but at least it gives a consumer the choice
15:12:26 <sjmc7> ok, moving on
15:12:30 <sjmc7> #topic WSGI
15:12:44 <sjmc7> lei-zh: there have been some mailing list threads about this
15:13:11 <sjmc7> which i now can’t find, but the summary is that the Powers That Be have decided to use uWSGI in devstack rather than mod_wsgi
15:13:38 <lei-zh> I haven't read them all, but from what I've seen uwsgi wins
15:14:00 <sjmc7> apache will still run but will proxy requests to uwsgi. the reasoning was that mod_wsgi significantly changes the behavior as far as a user is concerned; logs end up in different places, configuration is different, controlling threading is different
15:14:40 <sjmc7> yeah, uwsgi seems a better fit. your patch https://review.openstack.org/#/c/456627/ will mostly stay the same but i would recommend looking at https://review.openstack.org/#/c/456344/  to see how keystone did it with uwsgi in devstack
15:15:04 <sjmc7> it’s a shame the TC are moving the goalposts but the aim is still to get things running under wsgi in pike, just a different container
15:15:27 <lei-zh> yeah, restarting apache is awkward for developing
15:16:00 <lei-zh> I'll take a look at the patch and switch to uwsgi
15:16:22 <sjmc7> ok, thanks
15:16:54 <sjmc7> it looks like they’re also moving to not run things in screen in devstack any more
15:16:59 <sjmc7> which is a change i don’t agree with
15:17:56 <lei-zh> no more screen provided?
15:18:00 <sjmc7> https://review.openstack.org/#/c/461716/ will switch devstack over to using systemd
15:18:11 <sjmc7> oh, it merged already
15:18:28 <sjmc7> you can still set USE_SCREEN=True for now in local.conf
15:18:40 <sjmc7> but it defaults to false, and it’ll go away in the Q release
15:19:34 <lei-zh> that's fairly a big change
15:19:51 <lei-zh> I see no point keep both mod_wsgi and uwsgi in devstack, perhaps just switching to uwsgi is ok?
15:19:55 <sjmc7> yeah. i guess for new people it’s a bit more user friendly because it’s more standard
15:20:11 <sjmc7> yeah, i don’t think we need mod_wsgi at all if they’re switching to uwsgi
15:20:17 <sjmc7> keystone already used mod_wsgi
15:20:22 <sjmc7> so i think they left that behavior in
15:20:36 <lei-zh> ok
15:21:54 <lei-zh> dropping screen is a trend, it's no longer recommended, as you can see from removing rejoin screen script
15:22:11 <sjmc7> i’m old fashioned and still use screen a lot :)
15:22:53 <lei-zh> yeah, I like it too : )
15:23:07 <sjmc7> it will be nice being able to restart a lot of services at once i guess
15:23:25 <sjmc7> well, we’ll see. the decision’s made so not much point complaining :)
15:24:14 <sjmc7> ok, last thing i had
15:24:19 <sjmc7> #topic Boston summit
15:24:23 <sjmc7> the summit is next week
15:24:36 <sjmc7> unfortunately i won’t be there
15:24:44 <sjmc7> don’t know if anyone from this team will be
15:25:08 <lei-zh> me neither
15:25:34 <TravT> o/
15:25:52 <sjmc7> you’re late. see me after class
15:26:05 <TravT> yes sir
15:26:05 <lei-zh> haha
15:26:48 <sjmc7> i don’t know what attendance will be at the summit under the new model
15:28:00 <sjmc7> oo, the next one’s sydney. that’d be fun!
15:29:14 <lei-zh> not sure about others, but my team doesn't have engineers going
15:29:53 <lei-zh> and next ptg is denver : )
15:30:14 <sjmc7> no, i don’t either
15:30:20 <sjmc7> huh, that’s weird it’s in the US again
15:30:49 <lei-zh> lol
15:31:58 <TravT> lei-zh, are you still working on the notification forwarding?  wondering if you and robcresswell are still gonna have something to demo at the summit
15:33:17 <lei-zh> last week there was a email from people working on this to inquire the status of notification forwarding
15:33:36 <lei-zh> currently the work is stalled
15:34:26 <lei-zh> there's a rbac issue to send notifications as a service tenant to other tenants in zaqar
15:34:56 <sjmc7> ah :(
15:35:08 <lei-zh> I've talked to zaqar people and they've created a blueprint to solve this
15:35:23 <TravT> okay, cool. any idea when they'll get it done?
15:35:36 <sjmc7> heat uses keystone ’trusts’ to impersonate users
15:35:46 <sjmc7> but i guess all you really need is a policy rule allowing it
15:37:09 <robcresswell> TravT: Interns are still working on it. They've got everything wired up with zaqar right, they're just looking at different services to read from. Ceilometer, heat, searchlight etc.
15:37:34 <robcresswell> I can ask them to CC you in further contact if you'd like to be more involved
15:37:35 <lei-zh> the spec is here: https://review.openstack.org/#/c/449433/
15:38:25 <lei-zh> the problem is that currently if searchlight creates a zaqar, only 'service' tenant could access the queue.
15:38:36 <lei-zh> creates a zaqar queue
15:39:09 <sjmc7> yeah, i could see that. it needs to be able to specify a different tenant from its own
15:39:25 <TravT> ok.
15:39:59 <sjmc7> keep us up to date with that, it’s interesting work
15:40:43 <lei-zh> yeah, i'll keep an eye on that
15:40:56 <sjmc7> #topic open discussion
15:41:07 <sjmc7> i didn’t have any other agenda items. anyone else have anything they wanted to talk about?
15:42:01 <sjmc7> if not, we’ll end it here. robcresswell , have a splendid time in Boston
15:42:37 <robcresswell> Cheers :) I don't have anything to add.
15:42:45 <sjmc7> if you find yourself missing cricket, baseball season started a few weeks ago
15:42:57 <TravT> have fun robcresswell
15:43:27 <sjmc7> ok, thanks everyone. good day/night all
15:44:03 <sjmc7> #endmeeting