20:00:08 #startmeeting Octavia 20:00:09 Meeting started Wed Jun 20 20:00:08 2018 UTC and is due to finish in 60 minutes. The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:00:10 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:00:13 The meeting name has been set to 'octavia' 20:00:16 hi 20:00:19 o/ 20:00:19 Hi folks 20:00:44 Another hot and sunny day here 20:00:50 o/ 20:01:01 #topic Announcements 20:01:24 I wanted to highlight that I have setup the Rocky priority patch/review list 20:01:29 #link https://etherpad.openstack.org/p/octavia-priority-reviews 20:01:45 I am trying to update this daily with progress on the patches 20:02:20 If you have some time please try to help review and / or work on patches 20:02:35 Is it okay for others to update this as well? (In case I noticed updates) 20:02:49 It is a decent sized list, but we are making some progress. 20:02:55 nmagnezi Yes please! 20:03:15 ack 20:03:42 We are heading towards feature freeze and MS3 20:03:44 I have a question about those python 3 patches, but let's take it to the open discussion part 20:04:07 MS3 and Feature freeze is the week of July 23!!!! 20:04:13 Ok, sounds good. 20:04:46 The only other announcement I had is about the upcoming PTG 20:04:57 We will have a room for Octavia. I will be attending 20:05:31 If you need travel assistance, the foundation round 1 closes July 1st. 20:05:43 So apply soon 20:05:55 Any other announcements today? 20:06:25 #topic Brief progress reports / bugs needing review 20:06:58 I have been busy trying to squash bugs. There have been a few little ones and tech debt. 20:07:33 I am currently focused on fixing the dual amphora failure bug that causes failover to no be fully successful. 20:07:52 Link? :) 20:08:04 I have a plan, a PoC task and parallel flow. I just need to finish out the rest of the failover flow 20:08:17 #link https://storyboard.openstack.org/#!/story/2001481 20:09:00 Basically if more than one amphora is down/failed, the failover flow will not fully bring up the LB. One amp will be rebuilt, but the LB will be in provisioning ERROR 20:10:05 o/ 20:10:05 Hopefully I can get a patch posted today for it. 20:10:11 * rm_mobile is running late 20:10:20 Put it in the prio list :) 20:10:36 It is, under "Priority Patches (These need work or are WIP)" 20:10:48 Line 45 20:11:25 Any other progress updates from the team? 20:12:56 FYI, for those of you interested in the stable branch backporting, this patch would be good to backport. Someone was asking about it this morning. It will need some minor work to cherrypick though. 20:12:57 #link https://review.openstack.org/#/c/561369/ 20:13:12 not sure it was reported last week: grenade support is merged 20:13:13 So if someone is motivated.... 20:13:27 Yes! Excellent work. 20:13:38 Anybody have comments on making that voting? 20:13:46 if nobody os motivated I can grab that 20:13:57 I think we need to have it voting and an upgrade procedure doc and we can assert our tag(s) 20:14:06 yeah, let’s #vote on voting 20:14:14 xgerman_, go for it. propose the backport 20:14:18 k 20:14:50 Ha, I would only call a formal vote on this if Doug was here. Just to bug him 20:15:18 johnsom, I started the upgrade procedure doc but am out traveling this week 20:15:20 I think I will look at the job history and propose a voting patch 20:15:36 cgoncalves Ok, cool! 20:16:53 Any other updates or should we move on? 20:17:26 #topic Talk about API versioning/microversioning 20:17:40 We opened this topic a few weeks ago. 20:17:53 * nmagnezi looks at rm_work 20:18:49 Any new thoughts on this? Have people had time to look at the reference links? 20:18:52 * rm_work looks at nmagnezi 20:19:01 #link https://specs.openstack.org/openstack/api-wg/guidelines/microversion_specification.html 20:19:07 oh right i was going to look up stuff 20:19:11 #link https://docs.openstack.org/tempest/latest/microversion_testing.html#step4-separate-test-classes-for-each-microversion 20:19:22 * rm_work looks at the floor, which is where that task ended up 20:19:36 johnsom: oh fff no 20:19:47 bleh 20:20:33 As I mentioned last week, I lean towards not doing microversioning, but to increment the API version document with dot versions and only do additions to the API. 20:20:36 I have not. I looked +1 years ago to nova on their microversion implementation. we need some sort of versioning support for rolling upgrades, for sure 20:20:48 +! 20:20:54 Yeah, it's the clients that are becoming the bigger issue 20:21:22 clients don't know if they can ask for features 20:21:48 Ok, so 20:21:52 * johnsom glares at the room 20:22:08 What other projects are doing when the client asks for a feature that is not supported in a specific version? 20:22:10 Let's all take a look at those and have thoughts/ideas/comments ready for next week. 20:22:30 Likely the same 404 we do 20:22:56 have api versioning made to a community goal for rocky? 20:23:03 No 20:23:22 The goals for Rocky are no mox (we don't) and enable mutable configs 20:23:48 Stein is likely going to be py3 and something I forgot. 20:23:53 sometimes i think the extensions thing neutron did seems like a sane way to approach this. and then I flip my table and cry in a corner for a while. 20:24:48 +1 as much as I hate extensions they seem a bit better than this microversioning 20:24:51 Can't say I'm a huge fan of the extensions 20:24:56 Will It make sense to do some sort of version discovery behind the scenes for specific client actions that we know not all versions support? And in that case we can mask the error and output something like "this is not supported by the current API version" 20:25:36 for now i kinda just want to keep doing 20:25:38 the thing 20:25:41 that johnsom said 20:25:45 Oh, The other Stein goal that has traction is cold upgrade 20:25:46 For example action like amp failover that was added in queens IIRC 20:25:47 well, if you returned 404 before you are retro-actively changing API behavior 20:26:02 As I mentioned last week, I lean towards not doing microversioning, but to increment the API version document with dot versions and only do additions to the API. 20:26:09 johnsom, cold upgrade: check! :) 20:26:28 cgoncalves Yep! 20:27:18 So, if I update my version discovery patch and update the version are we good with that? leave the path /2.0 and have the multiple dot versions? 20:27:52 I think I will do that as a proposed patch. Then we can discuss again next week and make a call. 20:27:57 I think so IMO 20:28:03 +1 20:28:20 +1 20:28:43 #link https://review.openstack.org/#/c/559460/ 20:29:04 Yeah, that needs fixed though. I think I don't like how it is now 20:29:52 #topic Open Discussion 20:30:40 nmagnezi I think you had a py3 queston 20:30:44 question 20:30:48 yup 20:30:52 So re: https://review.openstack.org/#/c/573348/ 20:31:05 hSo re: https://review.openstack.org/#/c/573348/ 20:31:13 once that merged 20:31:29 Does that mean we gate on python 3 *only* ? 20:31:36 Just double checking myself here 20:32:06 It means that those jobs will run under py3, but we will still have py27 unit, functional, tempest 20:32:44 So, pep8 will run under py3, cover, docs, releasenotes, and the debug. 20:32:52 But for pep8 and cover it will run only with python3 right? 20:33:01 yeah 20:33:06 Basically they want to make py27 the exception instead of py3 being the exception as it is today 20:33:25 Correct, this change moves those over to run only under py3 20:33:25 yeah A++ 20:34:06 This is lining things up for the proposed Stein goal of all projects running py3 jobs as a minimum 20:34:30 yeah. I just wonder because we still gate on python 2.x on RDO and internally 20:35:06 Yeah, the important jobs: unit tests, functional, and tempest will still run both py27 and py3 to show that our code works under both 20:35:35 For the near future. Eventually OpenStack will drop py2.7 support. 20:35:59 in U at the earliest 20:36:07 Got it. Thanks for the answers :) 20:36:11 #link https://pythonclock.org/ 20:36:44 This is being driven by the fact that python 2.7 itself is going end-of-life in a year and a half 20:37:09 Well, 1 year, 6 months, 11 days, 9 hours, 22 minutes 20:37:41 Haha 20:38:02 I trust them to start Python 4.x just for having two versions 20:38:04 I will just be happy to only have to deal with one version 20:38:12 lol, yeah, probably 20:38:18 yep 20:38:35 Ok, any other discussions for today? 20:38:36 but by then we might have rewritten OpenStack in golang 20:39:08 C#? 20:39:09 xgerman_, dougwig once wanted to port neutron-lbaas to Ruby.. :-) 20:39:32 as an aside, python recommends not doing version detection :P https://docs.python.org/3/howto/pyporting.html#use-feature-detection-instead-of-version-detection 20:39:35 johnsom, fortran 20:39:42 (related to the microversioning discussion) 20:40:01 I recommend we rewrite the amp-agent in Rust 20:40:06 Hey, fortran is not a dead language.... 20:40:12 Pascal? 20:40:37 Oye, pascal, now we are going down hill. Might as well throw out perl or LISP 20:40:40 oooh, we had to learn J in uni... 20:40:54 https://en.wikipedia.org/wiki/J_(programming_language) 20:41:01 APL but typable by normal humans with normal keyboards 20:41:05 Sadly I have coded in all of the above except for rust 20:41:11 at one point or another 20:41:12 Rust is new 20:41:56 Ok, if nothing else I will let you fine folks go work on reviews. Please help with those, MS3 is coming quick. 20:42:09 https://en.wikipedia.org/wiki/System_programming_language#Major_languages 20:42:13 pretty sure http://www.cs.trinity.edu/old-index.cgi is written in J 20:42:14 take your pick 20:43:19 alright. I'm ... temporarily kinda semi-away from octavia for a few weeks, ping me for reviews or if there's patches that I need to update 20:43:25 :( 20:43:55 ok 20:44:18 Ok, thanks folks! 20:44:28 #endmeeting