16:00:33 #startmeeting Octavia 16:00:34 Meeting started Wed May 20 16:00:33 2020 UTC and is due to finish in 60 minutes. The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:36 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:38 The meeting name has been set to 'octavia' 16:00:47 Hi everyone! 16:00:56 Hi! 16:01:28 hi 16:01:29 hi 16:01:31 Welcome to the second week of Victoria development 16:01:59 #topic Announcements 16:02:01 hi 16:02:14 My regular items for the PTG: 16:02:20 PTG registartion is open 16:02:26 #link https://www.openstack.org/ptg/ 16:02:33 Our PTG etherpad: 16:02:41 #link https://etherpad.opendev.org/p/octavia-virtual-V-ptg 16:03:12 Next week I will try to put together a tentative schedule based on the topics we have in the etherpad. 16:03:34 It almost never goes to that plan, but it's better having a rough idea than nothing. 16:04:15 So, please add any topics you would like to discuss with the community this week or early next week so I can get a reasonable schedule put together. 16:04:42 Any other announcements this week? 16:05:25 #topic Brief progress reports / bugs needing review 16:05:40 I was off doing downstream stuff most of last week. 16:06:00 This week I have been trying to focus on code reviews and helping with rebasing 16:08:40 We have a number of things we need to catch up on that has me nervous. One is porting the failover patch to the v2 code path. Then some follow on patches (amphora delete for example). We also need to go back and do the "multi-process" code path for the TLS work. Double check it is all v2 ok, then finish the tempest and dashboard work on the TLS patches. I also have more tempest refactor work to finish, which 16:08:40 may require multi-node working. 16:09:25 Probably better for open discussion, but grenade job start failing on default ciphers change https://review.opendev.org/#/c/685337/ 16:09:39 Doh 16:09:39 not sure what is causing it... 16:11:15 the failover patch is still open and needs to be reviewed and merged asap, yeah. we've had customers hitting amphora failover issues due to broken clusters. 16:12:17 ataraday_ "unknown keyword 'ssl-default-bind-ciphersuites' in 'global' section\n" is the issue with ciphers 16:12:19 I want to give it a good review, but coming from Michael and been reviewed by Adam I'm sure it's in pretty good shape :) 16:12:42 Not sure why the tls test didn't also fail though. Plus, really points to the need to get the tempest tests done 16:13:33 yeah, I also noticed grenade job failing on the default ciphers. I did not have time to look deep but looked like a legit issue 16:14:00 johnsom, yeah but all other jobs don't fail in that way - https://review.opendev.org/#/c/685337/27/octavia/amphorae/backends/agent/api_server/haproxy_compatibility.py@62 process_cfg_for_version_compat should remove it if it is not supported... 16:14:33 on a second thought, I think I saw in in relation to tls version, not cipher 16:15:01 https://review.opendev.org/#/c/726741/ 16:15:02 I fixed tls version in tempest - so barbican job passed 16:15:10 __init__() got an unexpected keyword argument 'tls_ciphers' (HTTP 500) 16:15:21 I put figuring out the TLS 1.3 ciphersuite settings as a PTG topic as I think we have a choice to make there about the API 16:17:25 FYI, I got reviews on some of the TLS patches this week. I think Dawson is interested in seeing them through, so I will give him some time to address the comments before taking those over. 16:17:59 I'm not sure if the other students will do any additional updates to their patches 16:19:23 Any other updates today? 16:20:39 One more thing I could mention here or in open discussion. I'm poking at the configuration of our gate jobs some. This is specifically related to the tempest refactor and the pending need for faster testing. 16:21:27 I have posted a test job that moves the libvirt machine type over to q35 from i440fx. This is a long overdue change that the nova team is considering as well. 16:21:54 FYI, i440fx platform was the hot thing in 1997.... It is PCI based 16:22:09 Pentium Pro 16:22:21 q35 is a more modern platform and getting more attention from the kvm, etc. folks. It is PCIe based. 16:22:46 I don't expect a lot of performance benefit from this change, but I'm hoping for a bit more stability. 16:22:47 thank you for looking into it 16:23:02 question: why not wait for Nova to make it default? 16:23:13 So, just an FYI really. 16:23:34 Well, that has been kicking around nova for over a year. 16:23:34 is it to fix the KVM crashes in vexxhost? 16:24:05 I think it might help us and I want to get out in front of it so 8if* it breaks us in some way we know before it's the default in nova. 16:24:56 one more thing: tempest tests for UDP LBs are failing on CentOS because of the behavior of nmap-netcat (used by checkers), I proposed a fix at https://review.opendev.org/#/c/729230/ 16:24:57 ok. just wanted to understand the motivation behind it 16:25:03 It *might* help with the KVM stability issues. It also might help us with performance with the virtio drivers since it gets more attention than the i440 16:25:39 Basically I am kicking the tires and giving it a try to see what we get. 16:26:35 #topic RFE/Specifications 16:26:50 I also wanted to highlight we have a couple of specs up for review 16:26:56 Notifications 16:27:03 #link https://review.opendev.org/727915 16:27:13 Active/Active 16:27:22 #link https://review.opendev.org/723864 16:27:58 I would really like us to be responsive on these. It would be great if everyone could review, comment, and help out with wording. 16:28:38 I thought there was a third, but I didn't see it in my quick pre-meeting search. Let me know if you see it. 16:28:57 I expect both of these to need a few cycles before they are good to go. 16:29:12 Notifications are also on the PTG discussion list. 16:29:42 A reminder, we are not looking for perfection on these, but we want to make sure: 16:29:46 The use case is clear 16:29:58 We have designed any problem areas 16:30:06 Call out API changes required 16:30:18 Call out database schema changes/designs 16:30:56 Has a reasonably clear description of the change to the code base (i.e. where hooks for notifications would be placed, etc.) 16:31:33 We want to discuss anything that once merged we would be stuck with type design choices... 16:32:14 Any questions or comments on the specs? 16:34:02 #topic Open Discussion 16:34:02 I agree with your suggestions for the RFEs 16:34:45 Any other topics today? 16:35:17 ataraday_ Did we cover the grenade job issue or is there more we should discuss there? 16:37:10 did something changed in this job recently? 16:37:22 as it was successful before.. 16:38:07 I think the switch to zuul v3 based grenade merged recently (or it's still an open review) 16:38:47 https://review.opendev.org/#/c/726741/ 16:39:29 but it's failing too 16:40:00 __init__() got an unexpected keyword argument 'tls_ciphers' 16:40:07 #link https://review.opendev.org/#/c/725098/ 16:41:00 #link https://zuul.opendev.org/t/openstack/build/8c36ad31ce234aa7a2251566ddbf6e02/log/controller/logs/screen-o-api.txt#1787 16:42:06 my link was to the stable/ussuri one, which is failing in the same way 16:42:43 I see /opt/stack/old/octavia-lib/ in the trace. wondering if octavia-lib is not being upgraded, it should have been at that point 16:42:45 There might have been a bug here: 16:42:47 #link https://review.opendev.org/#/c/717154/ 16:43:23 Oh, that is interesting, yeah, if the transition doesn't update that module it would have issues 16:45:15 I can take point in this 16:45:38 may be a legit issue in our grenade job actually, so my bad 16:45:54 cgoncalves Thank you! 16:46:09 Any other topics today? 16:47:59 Ok, thank you for another great meeting. Good discussions. Have a great week! 16:48:04 #endmeeting