20:00:02 #startmeeting Octavia 20:00:03 Meeting started Wed Oct 19 20:00:02 2016 UTC and is due to finish in 60 minutes. The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:00:05 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:00:07 The meeting name has been set to 'octavia' 20:00:10 Hello folks! 20:00:13 Hi there 20:00:33 It could be a light turnout today as some folks have said they can't make it 20:00:43 #topic Announcements 20:00:51 hi all 20:01:03 o/ 20:01:21 We have a design summit session, add any topics comments to the etherpad 20:01:28 #link https://www.openstack.org/summit/barcelona-2016/summit-schedule/global-search?t=Neutron%3A+LBaaS+retrospective+and+next+steps 20:01:44 anyone is here? 20:01:50 Also, we had some gate breakage yesterday 20:02:01 #link https://review.openstack.org/#/c/345818/ 20:02:27 Infra is dropping projects from the "projects" list for devstack. They started with diskimage-builder 20:02:38 He! 20:02:39 This is to save cloning repos I guess. 20:02:40 heh! 20:02:54 johnsom, sorry for that question but what is the meaning of this? the fact that they drop it from the list 20:02:57 Yeah; would have been nice if they checked that we were using it first. 20:03:04 Anyway, I put a quick fix in yesterday for our gates, but it doesn't help folks adding us via the plugin 20:03:22 it makes the underlying image build quicker, and the devstack jobs have to do more cloning from the mirrors. 20:03:46 nmagnezi It means when we do a git clone of diskimage-builder during our devstack setup we get an error saying it's not an OpenStack project. 20:04:08 So, this is the start. I guess over time they are going to pull more out. 20:04:20 got it 20:04:33 I am going to do some tests and probably update the patch for out devstack plugin. 20:04:48 Just an FYI item in case someone was having trouble. 20:04:59 #topic Brief progress reports / bugs needing review 20:05:20 i18n/deb queries are out, waiting for replies. 20:05:32 dougwig I think these need your blessing: https://review.openstack.org/388834 https://review.openstack.org/386245 20:05:39 There are a number of CRs in neutron-lbaas in particular that I put a +2 on a couple weeks ago that haven't seen any reviews. 20:05:59 One is our functional job for Octavia, the other is the merge the jobs into one API test 20:06:22 I also have some strange problem in my change for distributor image creation that I wanted to ask you about 20:06:24 Ok, I will try to take a pass there soon 20:06:28 sbalukoff: ok, will look 20:07:09 Mostly I have been focused on getting the gate stuff straightened out. Our grafana dashboard is up now: 20:07:10 Dustin and I have been working on C10k optimizations for amphorae. We were able to get a test environment to reliably scale to 180k connections to a single amphora. 20:07:25 #link http://grafana.openstack.org/dashboard/db/octavia-failure-rates 20:07:39 And there will be 1 or 2 more patches related to that. (The only one not up for review is a documentation patch I'm working on talking about "high performance Octavia") 20:07:40 sbalukoff That is pretty darn cool. 20:07:49 More than I thought we would handle with one amp 20:08:06 Well, we had to beef up the amphora's RAM. XD 20:08:28 On that note-- I now have a good use case for why we might want amphorae to survive a reboot. 20:08:36 perelman We can address the image questions in open discussion if that is ok 20:08:51 Because resizing an amphora in place ends up looking a lot like a reboot. 20:09:00 sbalukoff Ugh, you are killing me. You just convinced me to give up on that..... 20:09:04 while building distributor image we try to clone ovs (git clone https://github.com/openvswitch/ovs.git /opt/ovs-install) 20:09:22 johnsom: Well, yes-- because nobody could come up with a good use case on that. But... now we have one. :P 20:09:29 and it fails on gerrit tests with - connection timeout to server 20:09:37 Granted, there are other ways to go about it if you want single instances to scale. 20:10:02 But... you know me: I am prone to change my position on something when new information comes to light. :P 20:10:49 * johnsom coughs multi-processes of haproxy 20:10:58 Haha! 20:10:59 sorry, i type too slowly and while typing dont see your requests :( 20:11:17 Well, yes-- I saw the light on that too and changed my position. Now... somebody just needs to do the work. :P 20:12:08 Sigh, yes. So, any thoughts on the cert stuff in light of the reboot need? 20:12:35 Well... a couple. 20:12:47 Is it a graceful reboot ? 20:13:08 It would be great if the amphora could send a signal to the controller (one of its various components) that it needs certs if it doesn't find them. And this would trigger a push to the amphora. 20:13:28 We might be able to piggy-back that on to the existing health check. 20:13:49 Yeah, I don't really want to open another port, etc. 20:13:52 Hmm 20:14:25 Ok, any other progress to talk about? 20:14:29 Just a flag that says "I have all the certs I need" or not. 20:14:57 #topic Open Discussion 20:15:04 hi 20:15:13 so lets return to my question 20:15:26 perelman Hi. I hope my comments have made sense on the act/act patches. 20:15:37 So this clone is in the element? 20:15:41 in patch for building distributor image #link: https://review.openstack.org/#/c/327137/33/elements/openvswitch-octavia/install.d/77-run_setup_install 20:15:57 we run command git clone https://github.com/openvswitch/ovs.git /opt/ovs-install 20:16:12 and fail with connection to server timeout 20:16:14 Yeah, that probably isn't going to work. 20:16:28 in our local machines it passes fast 20:16:33 It will need to be in here: https://review.openstack.org/#/c/327137/33/elements/distributor-agent/source-repository-distributor-agent 20:17:25 Take a look at the docs here: 20:17:27 #link https://github.com/openstack/diskimage-builder/tree/master/elements/source-repositories 20:17:42 when you finish with perelman's topic i would like to raise something 20:17:57 johnsom, I am trying to go over all the comments but in the last week we had holidays and also some deadline in other project, so only tomorrow i return to A/A with full gas :( 20:18:13 Cool. 20:18:14 thank you very much for the reviews! 20:18:29 Darn, I forgot a topic. Meeting schedule for the next two weeks. 20:18:49 Most of us will be at the summit next week, so I expect to cancel that meeting. 20:18:53 We can do it after open discussion. 20:19:06 Yep, that seems reasonable-- we'll be meeting in person, eh! 20:19:16 The week after I am vacationing in Barcelona, so we can cancel it as well or someone else can host. 20:19:40 ok, I will see the document 20:19:45 I'm also vacationing the following week. 20:20:14 Most folks are... It sounds like we should cancel that one as well. 20:20:20 I will send a note to the mailing list. 20:20:28 Sounds good! 20:20:41 so the next meeting is in 3 weeks? 20:20:57 perelman Give that a try, if not, leave me a message in the IRC channel and I will have a look 20:21:14 nmagnezi Yes, November 9th. 20:21:23 ack 20:21:34 perelman nmagnezi are either of you coming to the summit? 20:21:39 johnson, OK, thank you very much! 20:22:08 nmagnezi I think you had a topic 20:22:16 no, i think i cant.. we have some urgent deadline 20:22:19 johnsom, not, I can't travel this time 20:22:31 johnsom, yes 20:22:40 so, the another topic for the open discussion topic i would like to raise are two patches that wait for a second +2 (sbalukoff mentioned earlier that some patches are pending reviews for a couple of weeks now): https://review.openstack.org/#/c/299998/ https://review.openstack.org/#/c/344658/ 20:22:52 but i will try to connect if there will be possibility to connect to some of your meetings there 20:22:53 #agreed Next meeting will be November 9th 20:22:59 will be glad to do it 20:23:55 I think those are going to be fairly impromptu. 20:24:01 Ok. 20:24:14 Yeah, we don't have talk session this time. 20:25:32 nmagnezi Ok. I think Doug said he was going to have a look. I will also try to review those, but I have a list of reviews I'm still working on. 20:25:51 It occurred to Dustin and I that giving a #vbrownbag on high-performance Octavia would have been great, but apparently we're also too late for that deadline. 20:26:00 So... I guess we can just talk about it in person and share our notes. 20:26:58 johnsom, thanks. sorry just wanted to raise this one more time so it might get merged before the summit break 20:27:03 Yep. I plan to be in the design summit area most of the time, so opportunity for us to chat/work together. 20:27:20 Woot! 20:27:25 I'm looking forward to it! 20:27:47 Yep. Any other topics for today? 20:27:59 sbalukoff, do you eventually stick around? :) 20:28:14 nmagnezi: Er...? Today? 20:28:22 I'm going to be intermittently available today. 20:28:40 sbalukoff, no, in Octavia in general. asking because of your announcement the other day 20:29:09 johnsom, seems like there are no additional topics 20:29:32 If you're talking about me being paid to work on neutron-lbaas / octavia long term... Well, they've not told me to stop working on it yet. And based on more recent developments there's a very, very slim chance I might still be working on Octavia in three weeks. But it's slim enough that I didn't want to raise anyone's hopes (or fears). 20:29:47 So for now? I'll do what I can. Until they tell me to stop. 20:30:08 We appreciate the help! 20:30:20 I care a lot about these projects, because I've been working with y'all on them from their inception. And I think they're actually pretty good. :D 20:31:10 Ok, that sounds like it for today. Thanks! Safe travel 20:31:18 #endmeeting