20:00:47 <johnsom> #startmeeting Octavia
20:00:48 <openstack> Meeting started Wed Jun 13 20:00:47 2018 UTC and is due to finish in 60 minutes.  The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:00:49 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
20:00:51 <openstack> The meeting name has been set to 'octavia'
20:00:54 <johnsom> Hi folks!
20:00:58 <cgoncalves> o/
20:01:00 <nmagnezi> o/
20:01:05 <johnsom> Another fine day in Octavia land
20:01:37 <johnsom> #topic Announcements
20:01:48 <johnsom> I don't think I have any announcements this week.
20:01:55 <johnsom> We did release Rocky MS2
20:02:15 <rm_work> o/
20:02:34 <johnsom> We are now on the path towards MS3 the week of July 23rd
20:02:40 <johnsom> #link https://releases.openstack.org/rocky/schedule.html
20:02:56 <johnsom> We need to start prioritizing what we want/need to get into Rocky.
20:03:11 <johnsom> I plan to setup our "priority review" etherpad this week.
20:03:35 <johnsom> Does anyone have any other announcements today?
20:04:15 <johnsom> #topic Brief progress reports / bugs needing review
20:04:41 <johnsom> I have been working on bug fixes, the driver support library, and some internal planning stuffs.
20:05:26 <xgerman_> I got to know OSP 12…
20:05:31 <johnsom> Currently I am working on the amphora driver "updates" bug.  I had a patch and a plan yesterday but testing (+1 for octavia-tempest-plugin) showed that what I had was not a complete fix.
20:05:59 <johnsom> I am now working on a new approach (well, the old one) to the problem.  Hopefully a patch will arrive today.
20:06:37 <johnsom> Then I'm going to tackle the active/standby both amphora down bug. This is critical for Rocky (and backports) IMO
20:06:59 <rm_work> \o/
20:07:02 <nmagnezi> johnsom, can you link that story please?
20:07:08 * rm_work cheers for johnsom
20:07:19 <johnsom> Sure, it might take me a minute to find it.
20:07:21 <rm_work> if you want to discuss that, let me know
20:07:27 <rm_work> https://storyboard.openstack.org/#!/story/2001481
20:07:30 <rm_work> #link https://storyboard.openstack.org/#!/story/2001481
20:07:43 <johnsom> Yep, that one
20:07:45 <cgoncalves> we all can leave the room. johnsom is handling everything so nicely :)
20:07:55 <johnsom> Ha!
20:08:07 <johnsom> Happy to let someone take the dual down bug!!!!!!!
20:08:09 <rm_work> i did have some thoughts, as i've been noodling it (and working in that area of the flows) for a bit
20:08:15 <johnsom> Step right up!  No waiting....
20:08:19 <rm_work> but i don't have time to take it
20:08:23 <rm_work> I think
20:09:18 <johnsom> Yeah, I am not sure if I will "make it good enough" or attack the "better way to do it".  It will just depend on the scope.  Whatever, it needs fixed sooner
20:10:16 <johnsom> That said, I have a number of Octavia patches open that could use some reviews
20:10:20 <johnsom> #link https://review.openstack.org/#/q/owner:%22Michael+Johnson+%253Cjohnsomor%2540gmail.com%253E%22++status:open
20:11:26 <johnsom> Anybody else have updates?
20:12:02 <cgoncalves> grenade patch is ready but awaits parent patch to merge
20:12:44 <cgoncalves> barbical acl patch: I worked a little bit on it and got it to work, including revoke acls when not needed anymore
20:13:13 <johnsom> cgoncalves Do grenade *need* that patch?  I think my zombie patch may replace rm_works
20:13:14 <cgoncalves> I've tested also https://review.openstack.org/#/c/568361/
20:13:31 <cgoncalves> johnsom, I can rebase on top of yours instead, sure
20:13:59 <johnsom> Yeah, wondering if rm_work can review and decide if it's a complete duplicate or not
20:14:11 <johnsom> I think our discussions led us to it was a dup
20:14:43 <johnsom> It would be great to get grenade merged so we can move forward on the updates stuff in Rocky
20:15:07 <johnsom> cgoncalves How did you solve the multi-use of a cert when revoking?
20:15:36 <johnsom> Also, thank you for you work on all three of these.  Do you plan to vote on the DB speedup?
20:16:38 <cgoncalves> johnsom, by querying if there's any other listener pointing to the same cert
20:16:47 <cgoncalves> if there is, do not revoke
20:17:26 <cgoncalves> johnsom, I do plan but first I want to check with L7 rules. also health monitor listing still takes a long time
20:17:38 <johnsom> Hmm, ok, I will have to look at how that locks so we don't hit concurrency issues
20:17:46 <johnsom> Ok
20:18:17 <cgoncalves> I didn't want to discourage the author by asking to also work on the health monitor if other reviewers don't think it's required
20:18:47 <johnsom> Yeah, it is fine to incrementally make improvements.  As long as it doesn't make it worse...
20:19:46 <cgoncalves> so far it's been improving. the author had not touched listeners and I asked if he could otherwise I could submit since I had something drafted
20:19:58 <cgoncalves> anyway, I'll keep on reviewing it
20:20:04 <johnsom> Nice
20:20:46 <johnsom> Any other updates?
20:21:08 <johnsom> #topic Talk about API versioning/microversioning
20:21:43 <johnsom> Now I know Adam is not able to participate in this discussion this week, so we will probably carry it over to next weeks meeting.
20:21:51 <johnsom> However, I did look at this a bit.
20:22:03 <johnsom> I'm not sure if others did any research on it.
20:22:37 <nmagnezi> johnsom, I asked around about micro versioning in Neutron
20:22:51 <xgerman_> well, they use extensions
20:22:52 <johnsom> I think my biggest issue with microversions is the fact that if you don't ask for the newer versions with an HTTP header you get the oldest major
20:23:06 <johnsom> nmagnezi What did you learn?
20:23:09 <nmagnezi> johnsom, I was told to look at Nova... so, looks like Neutron doesn't really implement this. they indeed use api extentions
20:24:14 <johnsom> Yeah, that was my perception as well
20:24:15 <nmagnezi> as for Nova I didn't had the chance to reach out to anybody who work on it, but I started to read https://docs.openstack.org/nova/latest/contributor/microversions.html
20:24:44 <johnsom> #link https://specs.openstack.org/openstack/api-wg/guidelines/microversion_specification.html
20:24:50 <johnsom> #link https://docs.openstack.org/tempest/latest/microversion_testing.html#step4-separate-test-classes-for-each-microversion
20:24:57 <johnsom> Are two references I have been looking at.
20:25:24 <nmagnezi> johnsom, question is: say we decide to use microversioning (or just bump minor versions) is that a thing we can still do in Rocky?
20:25:29 <johnsom> Tempest is basically just skipping tests if the API doesn't support the version
20:26:02 <xgerman_> we can do version discovery
20:26:26 <johnsom> nmagnezi I really think we should have a plan and advertise a version in Rocky.  Now a full on microversion support, maybe not.
20:26:44 <johnsom> xgerman_ Not at the moment, the patch has not merged over the versioning discussion
20:26:49 <xgerman_> +1 unless somebody has a lot of time
20:27:13 <nmagnezi> xgerman_, yeah I was suggested with that too (I presume you mean doing it in the tempest plugin). the question is how terrible will be the maintanance of it
20:27:38 <xgerman_> yeah, that whole microversion stuff feels pre-mature + most clouds have a pretty recent version of Ovctavia
20:28:00 <xgerman_> nova, on the other hand, has icehouse clouds to deal with
20:29:06 <johnsom> Right.  I have some thoughts I need to put together, but at a high level I'm thinking we don't do microversions now. We do dot versions, 2.0, 2.1, 2.3 and only do additions to the API.  We then use a similar tempest setup that puts new features into versioned test classes. Or we cheat and just do skips.
20:30:03 <johnsom> I only propose the skips because of the LB boot overhead the separate classes would bring.
20:30:20 <nmagnezi> johnsom, skips by what? like in rm_work's patch?
20:30:25 <cgoncalves> so every single patch that touches the API should bump the dot version?
20:30:33 <johnsom> Ugh, another thing someone needs to work on is the multi-node gates
20:30:47 <johnsom> cgoncalves right.
20:31:03 <johnsom> nmagnezi I don't know, I haven't looked at his proposed patch in a few weeks
20:31:09 <nmagnezi> cgoncalves, yeah we might end up with version numbers that look like macOS :)
20:31:21 <johnsom> lol
20:31:25 <xgerman_> nothing wrong with that
20:31:40 <johnsom> Better than "all versions are 2.0"
20:31:46 <nmagnezi> johnsom, +1
20:31:46 <cgoncalves> ok, wanted to make sure. it will be an important thing when reviewing
20:32:13 <johnsom> The nice thing with that approach is people will get the latest by default.
20:32:28 <johnsom> Downside is it means we can't remove or break what we have there now.
20:32:38 <johnsom> Not until a major bump
20:33:34 <nmagnezi> johnsom, do you have anything in mind that you intended to refactor in the current API?
20:33:34 <johnsom> Well, let's keep thinking about it and plan to make some decisions next week.
20:33:40 <xgerman_> k
20:33:53 <johnsom> Unless of course cgoncalves is out one yet another vacaction/holiday....  grin
20:34:10 <xgerman_> he can vote by mail
20:34:14 <johnsom> +1
20:34:34 <johnsom> #topic Open Discussion
20:34:40 <johnsom> Other topics for today?
20:34:41 <nmagnezi> or by pidgin..
20:34:58 <xgerman_> cgoncalves: nmagnezi wonder if you guys have a periodic Pike amp build somehwere at Redhat
20:35:07 <cgoncalves> johnsom, I'm in Portugal since yesterday evening :) home-home office :)
20:35:10 <xgerman_> (remember my OSP12 advantures…)
20:35:20 <nmagnezi> xgerman_, no sir. we have that starting Queens
20:35:34 <xgerman_> mmh, where’s that at?
20:36:01 <nmagnezi> xgerman_, we didn't ship Octavia as part of OSP12.
20:36:13 <xgerman_> yes, I know - hence my adventures
20:36:24 <nmagnezi> xgerman_, hmm. when OSP13 gets released.. but we do have something for RDO
20:36:38 <xgerman_> ok, tell me more…
20:36:54 <nmagnezi> #link https://images.rdoproject.org/octavia/master
20:37:11 <nmagnezi> #link https://images.rdoproject.org/octavia/queens
20:37:17 <cgoncalves> xgerman_, we do have periodic builds for master and queens in rdo
20:37:27 <cgoncalves> oh right, those links :)
20:37:40 <xgerman_> sweet — might go with queens then
20:37:59 <johnsom> Are those centos based?
20:38:13 <johnsom> Oh, yeah, I see it in the hostname
20:38:20 <nmagnezi> johnsom, didn't check myself, but they should be :)
20:38:38 <cgoncalves> sure they are centos-based :)
20:39:08 <johnsom> Yeah, just wasn't sure if it was a rhel that needed a key or something
20:39:17 <johnsom> Cool
20:39:56 <xgerman_> nice
20:40:34 <nmagnezi> yup. just remember that the user is centos (as opposed to ubuntu that is a habit for devstack workloads with default settings)
20:40:57 <johnsom> Any other topics today?
20:41:25 <xgerman_> will just switch off ssh
20:43:39 <johnsom> Ok, well, if there isn't anything else, have a good week folks!
20:43:48 <nmagnezi> o/
20:44:11 <johnsom> #endmeeting