16:00:12 <johnsom> #startmeeting Octavia 16:00:13 <openstack> Meeting started Wed Aug 14 16:00:12 2019 UTC and is due to finish in 60 minutes. The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:14 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:16 <openstack> The meeting name has been set to 'octavia' 16:00:31 <johnsom> Hello everyone 16:00:36 <cgoncalves> hi 16:00:38 <colin-> greetings 16:01:01 <johnsom> I know some folks are on holiday, so it might be light attendance today 16:01:52 <johnsom> Including our PTL that is on vacation as well. 16:01:57 <johnsom> So, you are stuck with me 16:02:01 <johnsom> #topic Announcements 16:02:33 <johnsom> First up, please note that the Shanghai summit registration price goes up after today. 16:02:47 <johnsom> If you are planning to attend and have not yet registered, now is the time to do it. 16:03:20 <johnsom> I also understand we requested a 40 minute project update this time, but those have not been confirmed yet. 16:03:43 <johnsom> Also of note this week, the zuul UI for job output has changed 16:03:51 <johnsom> #link http://lists.openstack.org/pipermail/openstack-discuss/2019-August/008313.html 16:04:12 <johnsom> It is now one to two additional clicks to get to the job output, but it is all still there. 16:04:39 <johnsom> To find the rendered docs, click through the logs link and then into the html directory. 16:06:32 <johnsom> Just a reminder, we are coming up on feature freeze quickly. Libraries are Sept 2nd, the other repos Sept. 9th. Please plan accordingly to have feature work done well before that for review. 16:06:40 <johnsom> #link https://releases.openstack.org/train/schedule.html 16:06:55 <johnsom> Finally, check your e-mail for the "U" series naming poll. 16:07:24 <johnsom> I think that is all I have, does anyone else have any announcements today? 16:07:59 <cgoncalves> I think you covered it all :) 16:08:19 <johnsom> groovy 16:08:20 <johnsom> #topic Brief progress reports / bugs needing review 16:09:15 <johnsom> What have I been working on? lol I have been trying to fix some gate random failure issues. 16:09:42 <johnsom> Some are fixing jobs to use local mirrors to remove package download failures. 16:09:59 <cgoncalves> nothing from my side. my time has been spent on internal stuff 16:10:02 <johnsom> One was interesting where strptime turns out to not be thread safe on py27 16:10:28 <johnsom> I fixed a interface naming race condition in the amphora 16:10:44 <johnsom> The octavia-lib get code is done with tests. 16:11:10 <colin-> validated this internally and team provided some feedback in the comments https://review.opendev.org/#/c/667484/ seems to accomplish what we wanted so if anyone else wants to take a peek that'd be appreciated 16:11:35 <johnsom> I have now started my overhaul of the failover flows. So far, better VIP plugging and old amphora deletion code. I'm really just getting started however. 16:11:58 <johnsom> #link https://review.opendev.org/#/c/667484/ 16:12:21 <johnsom> colin- Cool. Please remember to review and vote too! 16:13:11 <johnsom> Also, thanks for highlighting this, it got overlooked and is not on the priority bug review list. I will add it 16:15:00 <johnsom> I have also been spending some downstream time. 16:16:17 <johnsom> Also, FYI, stable releases of octavia-dashboard have gone out today. 16:16:26 <johnsom> Any other updates? 16:16:54 <johnsom> I know Ann proposed a fix for taskflow for the history/results column filling with our retry logic. 16:17:05 <johnsom> So, jobboard is still progressing! 16:17:21 <johnsom> #topic Open Discussion 16:17:26 <johnsom> Other topics today? 16:17:58 <cgoncalves> today I proposed a patch that switches taskflow engine from serial to parallel 16:18:00 <cgoncalves> #link https://review.opendev.org/#/c/676379/ 16:18:42 <cgoncalves> I would like to know if someone objects to the idea due to e.g. bad experiences 16:19:13 <cgoncalves> our active-standby CI jobs run in parallel. it helps accelerate LB provisioning 16:19:18 <johnsom> It's a good thing. I know some downstream users are using it and I think we have some jobs with it enabled. 16:19:56 <colin-> any guidance for what exactly to tweak in jinja so that haproxy uses the full number of cores available to the amp? 16:20:01 <johnsom> It will accelerate a few things where we have parallel flows 16:20:15 <johnsom> colin- Ah, this is an interesting topic. 16:20:35 <johnsom> So, first answer, don't enable nbproc (mutli-process mode). 16:20:42 <johnsom> It breaks a bunch of stuff. 16:21:23 <johnsom> If you have a need for more cores, you will want to build an image with 1.9 or newer haproxy, and then enabled the threading support. 16:21:47 <colin-> interesting, and stable/stein is on 1.8 iirc? 16:21:51 <johnsom> Are you seeing workloads that eat enough CPU? Purely curious 16:22:11 <colin-> well, as i deploy stein and introduce flavors i want to be able to show effective utilization of additional cores 16:23:02 <johnsom> Right, this is why I haven't prioritized doing the thread enable work yet. All of the major distros still ship 1.8. HAproxy 1.8 does support threads, but there were some big changes in 1.9/2.0 that made it much more stable and scale better. 16:23:50 <colin-> ok got it. and purely from ignorance how will the update to 1.9 and beyond be handled in the project, the next major release _after_ the amp distro(s) add the packages? 16:24:33 <johnsom> Yeah, we probably have some flavors work to do for that too. I.e. alternate image 16:25:17 <johnsom> It would be nice to do tuning based on the number of cores available in the compute flavor as well. There is some performance boost with enabling the CPU pinning for the NICs and HAproxy threads 16:26:17 <johnsom> We have been debating this. I think at the latest it would be when an "OpenStack" base OS adds support for 1.9/2.0. 16:27:20 <johnsom> At the earliest, we code do some work in diskimage-create to enable pulling in custom packages and start that work earlier. It's just rough on where to reliably get alternate packages that don't blow up. 16:27:45 <johnsom> Also note, HAProxy advises that 1.9 is a short term release and 2.0 is what they are considering a longer term stable release. 16:28:26 <johnsom> Quoting: "An important point to note, this technical release is not suitable for inclusion in distros, as it will only be maintained for approximately one year (till 2.1 is out). Version 2.0 coming in May will be long-lived and more suitable for distros." 16:29:03 <colin-> thanks for the info, will take it into consideration 16:29:05 <johnsom> 2.0 just came out in June and has had some important bug fix releases, so still getting there. 16:29:36 <johnsom> There are a bunch of goodies in 2.0, such as http/2, etc. 16:30:36 <johnsom> If anyone is up for taking on some work, let us know and we can talk about the strategy for rolling it in. 16:30:52 <johnsom> Though it would be a huge stretch to get it in Train at this point. 16:31:17 <johnsom> Given the other features we agreed on for Train are still in flight. 16:33:02 <johnsom> Any other topics for today? 16:33:49 <johnsom> Thank you to all of you that have been doing reviews! We have burned down 67 outstanding reviews since I setup the patch review list. 16:34:11 <cgoncalves> o/ 16:34:12 <johnsom> Remember every review matters and helps the project become better. 16:35:05 <johnsom> We could also use more core reviewers, so if you are interested, the direct path to core is good reviews. 16:35:39 <johnsom> With that, have a great week everyone! 16:35:44 <johnsom> #endmeeting