13:00:02 #startmeeting nova api 13:00:03 Meeting started Wed Apr 26 13:00:02 2017 UTC and is due to finish in 60 minutes. The chair is alex_xu. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:04 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:06 The meeting name has been set to 'nova_api' 13:00:10 who is here today? 13:00:30 o/ 13:01:13 emm...let us wait one more minutes 13:01:17 * johnthetubaguy is lurking with intent 13:01:49 super quite today 13:02:03 s/quite/quiet/ 13:02:53 ok, let us start 13:02:59 ha 13:03:36 since OSIC stuff, we have few things need help 13:03:52 #link https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/policy-docs 13:04:29 o/ 13:04:32 o/ 13:04:37 #link http://docs-draft.openstack.org/91/459791/5/check/gate-nova-docs-ubuntu-xenial/c062c69//doc/build/html/sample_policy.html 13:05:45 ^ we fnished most of them, so I already saw people sign up to help, so I think it isn't so hard for us 13:05:50 I will help on those also 13:06:02 #link https://review.openstack.org/459608 13:06:32 and johnthetubaguy already bring up the patch for removing discoverable policy rules 13:06:52 yeah, just tidying that up 13:07:06 johnthetubaguy: thanks for bring those patches up 13:07:13 I was wondering about splitting it, but its not really as bad as it first seems 13:07:32 yea, it isn't hard in one patch 13:07:49 yeah, seemed easier in the end, given all the files are given a similar treatment 13:08:18 I did go through policy-docs 13:08:28 and updated the BP with a list of what is missing 13:08:39 #link https://blueprints.launchpad.net/nova/+spec/policy-docs 13:09:08 finding raising a few of those patches quite... therapeutic given everything thats going on at the moment 13:09:43 cool, it is great to have a list 13:11:34 #link https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bug/1661360 13:12:03 cdent: nova api via apache wsgi ^ 13:12:09 should https://review.openstack.org/#/c/428404/ be abandoned? 13:12:12 dansmith: ^ 13:12:21 yea, I guess that is old one 13:12:38 yeah, the wsgi stuff is coming along nicely 13:12:47 cdent: that is all what we need for this? 13:12:51 there are wsgi apps for both nova-api and metadata, and some pending devstack stuff 13:12:58 alex_xu: probably some docs too? 13:13:07 the major work is on the devstack and grenade side of the world 13:13:12 so plan is get the nova stuff in now ish, and follow up with the devstack bits? 13:13:32 the nova stuff can merge (without being used) whenever 13:13:33 would be nice to see the devstack changes passed on these nova patches 13:13:48 mriedem: they have to some extent, but need to be changed into order to work with grenade 13:13:54 * cdent locates link 13:14:16 #link nova-wsgi devstack https://review.openstack.org/#/c/457715/ 13:14:25 that's all happy except for grenade 13:15:04 need help with the grenade stuff? 13:15:28 reading the steps it looks like just setting a flag in grenade to not use wsgi, 13:15:33 and devstack on master would use wsgi by default 13:15:37 i don't know what "We honestly probably also need to do the systemd cut over before this to simplify the world." means though 13:15:48 mriedem: maybe, sdague had some comments that mostly mean me putting some stuff back into lib/nova 13:16:06 yeah you can't remove the non-wsgi stuff yet 13:16:18 probably in queens 13:16:48 wasn't planning to remove it, just planning to not test it ;) 13:16:53 but yeah, agree 13:16:56 so I have a quick thing 13:16:57 mriedem: there are some edge conditions around uwsgi without system that's weird 13:17:00 the pbr names 13:17:21 I think we have nova-metadata-api and nova-api-metadata 13:17:34 they need to be different 13:17:38 or else they clobber 13:17:45 yeah, agreed need to be different 13:17:49 just seems confusing 13:18:02 nova-metadata-wsgi? 13:18:09 maybe thats worse 13:18:10 yeah, could do that 13:18:12 heh 13:18:16 * cdent naming is hard 13:18:40 * johnthetubaguy nods 13:18:45 yeh, -wsgi is probably better 13:19:04 I can see me screwing that up 13:19:33 well, the theory is that the eventlet path is deprecated in pike and removed in queens right? 13:19:43 the screw up window is small 13:19:44 I thought so 13:20:03 thats true 13:21:02 so I am not the best at decisions, cdent whats your take on the best way forward? 13:21:25 oh, I guess we need a reno in this chain of patches somewhere 13:21:47 I think nova-metadata-wsgi is fine, maps to nova-api-wsgi 13:21:49 cdent: also metion the doc 13:22:01 s/metion/mention/ 13:22:28 yeah, I guess its always wsgi, so no halm in that being there 13:22:31 but i guess that doc isn' in the nova repo 13:23:22 we also need to figure out how not to repeat the placement api documentation gap here 13:23:27 do we have many other projects doing similar PBR names, out of curiosity, I have really kept track of that effort? 13:23:39 sdague: +1 13:23:41 i can make some of those adjustment later today 13:23:44 johnthetubaguy: I honestly don't know how far anyone else got here 13:24:13 it'd be in the governance repo 13:24:15 for the release goal 13:24:16 sdague: something tells me OSIC folks volunteered for some of that 13:24:17 I was trying to get the uwsgi and systemd plumbing all in before too many folks did 13:24:19 EmilienM might know 13:24:22 johnthetubaguy: yeh 13:24:33 before we deprecate eventlet we have to be running with wsgi on master in CI 13:24:40 mriedem: sure 13:25:00 #link https://governance.openstack.org/tc/goals/pike/deploy-api-in-wsgi.html 13:25:15 seems lots of folks haven't got their responses merged yet 13:26:48 "These projects already deploy devstack with API service via WSGI" is pretty error prone as well, as I don't know which of those are wsgi by default 13:28:45 in case its useful, this is keystone: #link https://github.com/openstack/keystone/blob/master/setup.cfg#L81 13:28:57 keystone-wsgi-admin and keystone-wsgi-public 13:29:02 * johnthetubaguy shrugs 13:29:19 I didn't do it that way because of grenade's grep style for checking if services are running 13:29:48 however, since the grep will be against the uwsgi process's command line, not the name of the wsgi script it may not matter 13:29:59 Will all the projecs switch to uwsgi? or part of them use apache mod_wsgi? 13:30:03 (I wanted to be able to grep either nova-api the eventlet thing, or nova-api the uwsgi thing) 13:31:00 alex_xu: I want to remove the mod_wsgi bits from devstack 13:31:25 sdague: ok, got it 13:31:28 it's actually kind of error prone in our environment because any apache restart ends up taking all the api services offline for a noticable amount of time 13:32:35 whereas when things are in uwsgi state you can do restarts of services seperated from restarts of apache (and apache restarts are much quicker) 13:33:01 yea 13:33:11 cdent: yeh, so once we get across to systemd this cycle, all the grenade stuff can be so much smarter on status tracking 13:34:29 so adding an extra alias later seems rather cheap, sounds like we are getting to the bottom of some of the details though, which is awesome 13:36:20 whats next on the agenda? 13:36:49 that is all for priorities 13:37:10 and I checked the merged specs 13:37:23 and BPs 13:37:33 we have 13 API related BPs approved 13:37:52 all of them have patches up 13:38:07 mriedem: I think that patch still makes sense as a safety measure, fwiw 13:38:54 the novaclient changes for 2.43, 2.44 and 2.45 are ready 13:39:07 mriedem: yea 13:39:31 #link https://review.openstack.org/#/c/459496/ 13:39:48 ^ just left the client patch, then they are finish 13:40:52 that is all I have 13:41:12 alex_xu: per your comments on the shell and using wraps 13:41:18 mriedem: yea 13:41:20 did we get the tags stuff all fixed already? 13:41:22 the shell will still pass through to the api binding which will fail with a 404 13:41:41 alex_xu: but if we cap the shell, it will fail before we emit any deprecation warning 13:42:01 and the deprecation warning says, "Specify --os-compute-api-version less than ' 13:42:01 '2.43 to continue using this command until it is removed" 13:42:04 we will have help text in the `nova help` 13:42:27 it's a 404 either way, but you get the more detailed deprecation warning this way 13:42:28 is my point 13:43:08 johnthetubaguy: what tags stuff? 13:43:25 the device tags 13:43:36 we accidentally dropped it, then had to re-add it 13:43:41 that was 2.42 13:43:45 takashin fixed that in novaclient 13:43:47 very nicely 13:43:51 mriedem: ok, I can test that later. I remember we will hide that command 13:44:03 ah, good, so this is all on top of that 13:44:05 alex_xu: i don't think we can hide it 13:44:08 I hadn't seen that go in 13:44:14 alex_xu: that would be like dropping it w/o a warning 13:44:22 johnthetubaguy: yup 13:44:42 alex_xu: we also said in the spec we wouldn't automatically drop down a microversion in the shell 13:45:09 so users can still use the deprecated shell but they have to opt in with the --os-compute-api-version option 13:45:32 mriedem: whatever the commands will be failed, since the api returned 404 13:46:05 right, unless you specify 2.1 or 2.42, or something in between 13:46:07 mriedem: even we added that decorator, the user also can use that command with --os-compute-api-version option 13:47:06 mriedem: anyway, let me check that again after the meeting 13:47:38 anything more people want to bring up? 13:48:19 ok, let us close the meeting thanks all! 13:48:22 #endmeeting