13:02:22 <alex_xu> #startmeeting nova api
13:02:23 <openstack> Meeting started Wed Aug 31 13:02:22 2016 UTC and is due to finish in 60 minutes.  The chair is alex_xu. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:02:24 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:02:26 <openstack> The meeting name has been set to 'nova_api'
13:02:31 <johnthetubaguy> o/
13:02:38 <cdent> o/
13:02:47 <alex_xu> sorry, I always remember we use openstack-meeting-alt...
13:03:09 <alex_xu> I guess sdague busy on something
13:03:20 <alex_xu> #topic API Priorities
13:03:36 <alex_xu> let us go through last few patches
13:03:46 <alex_xu> #link https://review.openstack.org/#/c/356963/
13:04:21 <alex_xu> the one is about emit warn msg for user_id based rule
13:04:38 <alex_xu> I didn' some improvement and test
13:04:47 * edleafe wanders in with cup of coffee in hand
13:05:08 <alex_xu> it only checked the rules which loaded from policy.json
13:05:19 <sdague> o/
13:05:39 <sdague> alex_xu: so I didn't see examples of that working on devstack
13:05:48 <sdague> do you have log lines pointing to it working?
13:05:50 <alex_xu> when user have full custom rules, it spend 0.37 second with execute 1000 times
13:05:59 <alex_xu> sdague: I fixed it, let me find the line
13:06:48 <sdague> alex_xu: ok, cool, if you fine examples of it running and put it into the review, I'll look later on it
13:07:03 <sdague> starred it to come back
13:07:22 <woodster_> o/
13:07:42 <alex_xu> sdague: ok
13:07:46 <alex_xu> sdague: there is the log http://logs.openstack.org/43/360943/3/check/gate-tempest-dsvm-full-ubuntu-xenial/fdd2590/logs/screen-n-api.txt.gz#_2016-08-29_15_15_52_844
13:08:05 <alex_xu> johnthetubaguy: are you ok with the performance?
13:08:52 <johnthetubaguy> its not free, but it seems minimal, I am just trying to remember how many requests/sec we have going through an API node, and what difference that could make
13:09:31 <sdague> 0.37 seconds for 1000 iterations means that we're really adding only 1ms right?
13:09:37 <sdague> on a normal request
13:09:45 <alex_xu> johnthetubaguy: yea, if user have empty policy.json, it only use 0.009 second
13:09:56 <sdague> alex_xu: per request?
13:10:05 <alex_xu> sdague: no, 1000 request
13:10:30 <johnthetubaguy> that sounds better
13:10:36 <johnthetubaguy> how did you get 0.37?
13:10:36 <sdague> right, so honestly, if the number is < 1ms per request it's a pointless discussion
13:10:42 <sdague> johnthetubaguy: for 1000
13:10:50 <sdague> if all the rules are custom
13:11:09 <alex_xu> johnthetubaguy: I only execute the check code in the script with a loop execute 1000 times
13:11:26 <johnthetubaguy> I was meaning what was in policy.json
13:11:35 <johnthetubaguy> but anyways, seems like its tidy, so lets call that fine and move on
13:11:41 <sdague> cool
13:11:43 <johnthetubaguy> s/tidy/tiny/
13:11:46 <sdague> yeh
13:12:14 <alex_xu> johnthetubaguy: just load the defualt rule
13:12:18 <alex_xu> johnthetubaguy: cool, thanks
13:12:48 <alex_xu> the last one
13:12:51 <alex_xu> #link https://review.openstack.org/322944
13:13:00 <alex_xu> policy discover CLI
13:13:26 <alex_xu> just waiting for more feedback
13:13:54 <edleafe> alex_xu: added to my queue
13:13:55 <sdague> ok, so this might be one of those things where it's low risk, and we move it forward with only light feedback
13:14:00 <alex_xu> edleafe: thanks
13:14:16 <alex_xu> sdague: yea
13:14:22 <sdague> related to policy, when I brought up the policy in code at the ops summit, people were excited
13:14:36 <johnthetubaguy> sdague: cool that echos the feedback from manchester
13:14:41 <sdague> but they also said we really were going to need a policy reference document
13:14:44 <alex_xu> sdague: cool :)
13:14:55 <johnthetubaguy> sdague: yeah, +1, with descriptions, etc
13:14:56 <sdague> because the policy.json file would end up blank
13:15:06 <sdague> and they wouldn't know which things to change
13:15:34 <johnthetubaguy> the cmd that outputs the merged yaml was the initial work around to help with that a little bit
13:15:37 <alex_xu> yes, but if the rule name can easy to map to url, that will better
13:15:40 <sdague> so, either post freeze, or early next cycle, we should make sure that policy reference document is a high priority
13:16:00 <sdague> alex_xu: yeh, we could also fix it with adding descriptions, which honestly might be better
13:16:13 <sdague> that's the way ops think about things, not always like urls
13:16:21 <alex_xu> ok, at least that is the thing we can do now
13:16:22 <johnthetubaguy> I am tempted by both, honestly
13:16:23 <sdague> this is forward looking, but relevant for O goals
13:17:10 <johnthetubaguy> well, we can do more url related things, once we have alias, so upgrading is not impossible, description first is my take, but yeah, this is for O
13:17:25 <sdague> yep
13:17:32 <sdague> ok, moving on?
13:17:35 <johnthetubaguy> I was hoping the cmd to output the merged should do for now, does that seem OK?
13:17:40 <alex_xu> yea
13:17:43 * edleafe reads 'O goals' as 'zero goals'
13:17:44 <sdague> johnthetubaguy: yeh, probably
13:17:58 <sdague> edleafe: you need a better font :)
13:18:08 <edleafe> sdague: :)
13:18:14 <alex_xu> that is all we have before freeze
13:18:18 <johnthetubaguy> there is another one that tells you about duplicates to help with upgrade too, but yeah, we should move on
13:18:19 <alex_xu> anything I missed?
13:18:22 <sdague> \o/
13:18:32 <alex_xu> \o/
13:18:38 <sdague> johnthetubaguy: yeh
13:19:00 <alex_xu> #topic open
13:19:02 <sdague> honestly, my brain is over on the placement-api this week, so unless there is anything else, probably worth getting back over there
13:19:29 <sdague> oh, this should be a quick approve, api-ref related - https://review.openstack.org/#/c/362170/
13:19:52 <sdague> once dev calms down from the freeze, I'll schedule a final api-ref sprint, probably mid sept
13:19:53 <alex_xu> there is bug mentioned by Matt, Kevin_Zheng happy to help on this https://review.openstack.org/359134, but this can be done after freeze
13:20:20 <alex_xu> sdague: cool
13:20:37 <Kevin_Zheng> yes
13:21:00 <alex_xu> Kevin_Zheng: thanks for helping
13:21:13 <Kevin_Zheng> np
13:21:25 <alex_xu> so...if no more question, let us end the meeting, people are busy this week.
13:21:38 <alex_xu> 3...
13:21:41 <alex_xu> 2..
13:21:45 <alex_xu> 1.
13:21:53 <alex_xu> ok, thanks all!
13:21:54 <edleafe> thanks, alex_xu!
13:22:03 <alex_xu> #endmeeting