| *** watersj has joined #openstack-ansible | 00:19 | |
| fridtjof[m] | noonedeadpunk: ah, thanks. i'll keep that in mind. it's not necessary though if I start off with just one compute host named "compute2", and then add "compute1" later, right? | 00:28 |
|---|---|---|
| *** spatel has joined #openstack-ansible | 00:43 | |
| *** d34dh0r53 has quit IRC | 00:48 | |
| *** spatel has quit IRC | 00:48 | |
| openstackgerrit | Merged openstack/openstack-ansible master: Update global-requirement-pins https://review.opendev.org/717748 | 01:03 |
| *** spatel has joined #openstack-ansible | 01:13 | |
| *** spatel has quit IRC | 01:18 | |
| watersj | looking at deployment guide of osa, does it use openvswitch or linux-bridge? If bridge can it be conf | 02:09 |
| *** spatel has joined #openstack-ansible | 02:27 | |
| *** spatel has quit IRC | 02:29 | |
| *** spatel has joined #openstack-ansible | 02:29 | |
| *** spatel has quit IRC | 02:30 | |
| *** threestrands has joined #openstack-ansible | 02:33 | |
| *** hwoarang has quit IRC | 02:43 | |
| *** hwoarang has joined #openstack-ansible | 02:43 | |
| *** rh-jelabarre has quit IRC | 03:07 | |
| *** _mmethot_ has joined #openstack-ansible | 04:18 | |
| *** mmethot_ has quit IRC | 04:21 | |
| *** udesale has joined #openstack-ansible | 04:22 | |
| *** evrardjp has quit IRC | 04:35 | |
| *** evrardjp has joined #openstack-ansible | 04:35 | |
| *** cshen has joined #openstack-ansible | 05:00 | |
| *** cshen has quit IRC | 05:05 | |
| *** gyee has quit IRC | 05:29 | |
| *** cshen has joined #openstack-ansible | 06:00 | |
| *** this10nly has joined #openstack-ansible | 06:44 | |
| *** KeithMnemonic has quit IRC | 07:12 | |
| jrosser | watersj: you can choose linucbridge or ovs. the defaul setting if you don’t change anything is linuxbridge | 07:18 |
| *** tosky has joined #openstack-ansible | 07:24 | |
| *** yolanda has joined #openstack-ansible | 07:24 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/openstack-ansible master: Imported Translations from Zanata https://review.opendev.org/721064 | 07:27 |
| *** rpittau|afk is now known as rpittau | 07:39 | |
| admin0 | watersj, unless you know that your requirement needs an ovs, you can stick to linuxbridge | 08:00 |
| Adri2000 | hello | 08:22 |
| Adri2000 | I'm running queens (upgrade planned really soon :)) and after deploying the latest OSA version (from tag 17.1.16 to 17.1.17), swift-object-auditor doesn't want to start anymore, failing with ImportError: cannot import name viewkeys | 08:24 |
| kleini | admin0: what can be possible requirements to choose ovs? I am struggling, too, between choosing lb or ovs. I have a "handcrafted" OpenStack beginning with Icehouse release and there the install guides used ovs. Today they all use lb. | 08:25 |
| Adri2000 | I tracked down the issue and found an ugly workaround that involves switching the order of two import lines in /openstack/venvs/swift-17.1.17/lib/python2.7/site-packages/swift/obj/auditor.py, "from eventlet import Timeout" and "from swift.obj import diskfile, replicator" (which in turns does "from six import viewkeys") | 08:25 |
| Adri2000 | can't find what's wrong, can't see a significant diff between the 17.1.16 and 17.1.17 venvs that would explain the issue (eventlet and six versions didn't change) ... not sure how to debug further | 08:26 |
| Adri2000 | summary of my strange problem: https://pastebin.com/6fLcVPsK | 08:29 |
| *** threestrands has quit IRC | 08:33 | |
| *** sshnaidm|afk has quit IRC | 08:55 | |
| *** gshippey has joined #openstack-ansible | 08:56 | |
| noonedeadpunk | Adri2000: I think that it might be already different version of six isntalled | 09:35 |
| noonedeadpunk | as we upper-constraints might be changed | 09:35 |
| admin0 | kleini, for example, would you/users be doing EPA/NFV tests .. like connecting multiple t-rex traffic generators to multiple ports and trying to performance test an application on how it behaves on say 20gb , 50gb, 100gb traffic ? | 09:36 |
| admin0 | generally speaking, if you are not crossing 10gbit in network traffic, lb works just fine ( simple to setup, work, troubleshoot, no headaches etc) | 09:37 |
| admin0 | ovs is if you know your application will need it or need to do performance testing on the 20gbit+ network traffic, epa or nfv requirements etc | 09:37 |
| kleini | our setup is target for performance tests of dovecot with 20G network connections currently. so ovs is maybe a good choice. | 09:38 |
| admin0 | single dovecot instance is needed to cross 20G network barrier ? | 09:40 |
| kleini | 384 core, 3TB memory cluster setup with scality storage | 09:41 |
| admin0 | that does not answer :) | 09:41 |
| admin0 | a cluster requirement is different from a per-instance requirement | 09:42 |
| kleini | let's say it is a single IMAP server instance but a highly clustered cloudy setup. I am unfortunately not in the details. | 09:43 |
| admin0 | use lb :) | 09:43 |
| admin0 | thats easy to answer because i know dovecot | 09:43 |
| kleini | we develop dovecot :) | 09:43 |
| admin0 | :D | 09:43 |
| admin0 | then you know better in its architecture and how you would set it up in a SaaS or PaaS | 09:44 |
| admin0 | type offering | 09:44 |
| kleini | any concrete reason, why to use lb? | 09:44 |
| admin0 | ease of user, setup, troubleshooting, less headaches | 09:44 |
| kleini | okay, thanks. currently ovs is easier for me than lb, as I am familiar with ovs | 09:45 |
| admin0 | well, it also depends on the other guys in the team .. the guy who gets call at 4am or does shift on holidays etc | 09:46 |
| kleini | I will teach them ;-) | 09:46 |
| admin0 | then its an easy choice right .. stick with ovs | 09:46 |
| *** pcaruana has quit IRC | 09:50 | |
| *** sshnaidm has joined #openstack-ansible | 10:14 | |
| *** rpittau is now known as rpittau|bbl | 10:21 | |
| kleini | so one headache is currently, that OVS bridge with IP can not communication with macvlan type bridge in container. so host <-> container communication is broken there. I am still searching, what the cause could be. | 10:26 |
| *** pcaruana has joined #openstack-ansible | 10:27 | |
| jrosser | kleini: linuxbridge and OVS performance is basically the same - there is not a valid reason to choose between them on the grounds of that | 10:40 |
| jrosser | we don't normally recommend OVS on the control plane nodes | 10:42 |
| jrosser | imho your control plane is being made too complex with nspawn + ovs + macvlan | 10:43 |
| jrosser | it is very much simpler with lxc + bridges | 10:43 |
| kleini | so linux bridge is able to perform same as ovs on br-vxlan and br-vlan, so the tunnel bridge and the public network bridge? those will be the bridges with the most traffic between VMs | 10:51 |
| jrosser | kleini: watch the video here https://www.openstack.org/summit/denver-2019/summit-schedule/events/23525/pushing-packets-how-do-the-ml2-drivers-stack-up | 10:52 |
| kleini | thanks! | 10:53 |
| jrosser | it's totally fine to pick between ovs and linuxbridge on the compute/network hosts for tenant traffic - i just don't think its necessarily a decision to make based on performance | 10:53 |
| jrosser | you make have a preference based on familiarity, and thats totally fine | 10:53 |
| jrosser | but i think that control plane nodes should be considered differently | 10:54 |
| kleini | I am not only talking about br-mgmt, br-vxlan, br-vlan but also about the ML2 driver. Regarding the ML2 driver I have the choice only once, when deploying the cloud. I can not change that later. That's why I'm hesitating and asking that much. | 10:58 |
| kleini | will watch the video | 10:58 |
| *** watersj has quit IRC | 11:01 | |
| openstackgerrit | Andreas Jaeger proposed openstack/ansible-role-uwsgi master: Cleanup py27 support https://review.opendev.org/722277 | 11:03 |
| *** thuydang has joined #openstack-ansible | 11:42 | |
| Adri2000 | noonedeadpunk: pip list, in both venvs, report the same version of six :( | 11:51 |
| *** rh-jelabarre has joined #openstack-ansible | 11:52 | |
| *** rpittau|bbl is now known as rpittau | 12:02 | |
| *** udesale_ has joined #openstack-ansible | 12:20 | |
| *** udesale has quit IRC | 12:22 | |
| admin0 | how we deploy is use lb in the controllers and ovs in the computes | 12:35 |
| admin0 | where some of the compute nodes are also network ndoes | 12:35 |
| admin0 | controllers are pure controllers only | 12:35 |
| kleini | and ovs as ml2 driver? | 12:49 |
| jrosser | admin0: "how we deploy is use lb in the controllers" <- you mean you use bridges on the controller, not linuxbridge in the ml2 sense (there is room for confusion here!) | 12:51 |
| openstackgerrit | Merged openstack/openstack-ansible-os_tempest master: Enable placeholder openstack-tox-molecule job https://review.opendev.org/673859 | 12:52 |
| kleini | yes, there are bridges on the target hosts to be prepared and an ML2 driver. My current tests show, that OVS on the controller host for bridges breaks nspawn containers talking to the host through macvlan. | 13:03 |
| openstackgerrit | Merged openstack/openstack-ansible-os_glance master: Update glance_store variables https://review.opendev.org/718513 | 13:06 |
| jrosser | kleini: it may be expected that you cant communicate with the host with macvlan - there is a similar situation for docker https://blog.oddbit.com/post/2018-03-12-using-docker-macvlan-networks/ | 13:11 |
| *** cshen has quit IRC | 13:12 | |
| *** udesale_ has quit IRC | 13:16 | |
| *** udesale has joined #openstack-ansible | 13:16 | |
| *** udesale has quit IRC | 13:18 | |
| *** udesale has joined #openstack-ansible | 13:19 | |
| kleini | thanks for the link jrosser. every day new funny stuff but I learn a lot and get an idea in which directory I should go. thank you very much! | 13:32 |
| jrosser | no worries :) | 13:32 |
| *** _mmethot_ has quit IRC | 13:34 | |
| jrosser | kleini: more stuff for you! https://cloudnull.io/2017/06/nspawning-openstack-ansible/ | 13:35 |
| kleini | I met cloudnull at the Berlin infrastructure summit, talked with him about our current setup according to install guides and that I want to move to some more automation. and he mentioned that nspawn containers are supported but obviously it has still some problems | 13:41 |
| jrosser | yes, kevin did all the work on nspawn but he's now spending his time on tripleo | 13:43 |
| jrosser | thats kind of why we don't have anyone actively developing it | 13:43 |
| *** mmethot has joined #openstack-ansible | 13:44 | |
| *** cshen has joined #openstack-ansible | 13:44 | |
| kleini | that's sad! | 13:48 |
| *** cshen has quit IRC | 13:49 | |
| *** d34dh0r53 has joined #openstack-ansible | 13:54 | |
| *** cshen has joined #openstack-ansible | 13:57 | |
| admin0 | i meant the br-vlan and br-vxlan on the controllers is done using lb .. not ovs .. on the compute nodes, its done using ovs | 13:59 |
| admin0 | an LTS version of os/osa would be good :) | 14:00 |
| *** sshnaidm has quit IRC | 14:01 | |
| *** watersj has joined #openstack-ansible | 14:02 | |
| *** sshnaidm has joined #openstack-ansible | 14:02 | |
| *** this10nly has quit IRC | 14:06 | |
| kleini | that's a combination I am heading for now, too. LB on control plane (infra, controller, all containers) and OVS on data plane (compute & network nodes, no container at all) | 14:34 |
| jrosser | kleini: the default config puts the neutron servers on the network nodes - you should probably move those your controllers | 14:42 |
| kleini | did that already according to your proposal some days ago | 14:46 |
| kleini | thanks again | 14:46 |
| mnaser | noonedeadpunk: can you help out and take care of setting up the ptg slots that can work for everyone? | 14:48 |
| mnaser | i can make myself available anymore, jrosser ? | 14:48 |
| jrosser | hmm? | 14:48 |
| noonedeadpunk | yeah, ok, wil check this out | 14:49 |
| * noonedeadpunk didn't get the last part as well | 14:50 | |
| noonedeadpunk | read through ML thread and looked through ethercalc... Either I don't understand smth or there're too much slots... | 14:59 |
| noonedeadpunk | and what type of team are we? horizontal I guess? | 15:01 |
| noonedeadpunk | jrosser: out of https://ethercalc.openstack.org/126u8ek25noy what do you think the most appropiate time? | 15:03 |
| noonedeadpunk | like I feel we can take slots on tuesday for sure | 15:04 |
| noonedeadpunk | do we want to take monday as well? | 15:04 |
| noonedeadpunk | 13-17 utc sounds ok for me, this will cover mnaser time I guess as well? | 15:05 |
| noonedeadpunk | 21- midnight UTC fits we as well actually | 15:05 |
| noonedeadpunk | the thing is that 13-17 UTC will be the same time with kolla and tripleo | 15:07 |
| jrosser | probably 17:00 UTC is as late as would be good if we want lots of my team to join | 15:07 |
| jrosser | i.e before 17:00 UTC | 15:07 |
| noonedeadpunk | so I was thinking to take from 13 till 17 | 15:08 |
| noonedeadpunk | oh wait | 15:08 |
| noonedeadpunk | jrosser: so what's your suggestion? | 15:08 |
| * noonedeadpunk for second became unsure what UTC is | 15:09 | |
| jrosser | i have to google "utc time" each time just to double check | 15:09 |
| noonedeadpunk | I think utc it's actually your native one :D | 15:10 |
| jrosser | except my wall clock is 1 hour later than that right now | 15:10 |
| noonedeadpunk | ah, summer time, right | 15:10 |
| noonedeadpunk | so then we should take just till 16? | 15:11 |
| jrosser | i'm missing something here - do we need to choose particular days? | 15:12 |
| noonedeadpunk | yeah, from the first part of the schedule... | 15:12 |
| noonedeadpunk | http://lists.openstack.org/pipermail/openstack-discuss/2020-April/014126.html | 15:12 |
| noonedeadpunk | but looking at tripleo we might take even several | 15:12 |
| jrosser | and we are concerned about the overlap with kolla/tripleo? | 15:14 |
| noonedeadpunk | not sure that we are... but maybe some ppl would like to join both | 15:15 |
| jrosser | i'm asking folk here if they have any preference | 15:20 |
| noonedeadpunk | And I'm probably need to look at https://governance.openstack.org/tc/goals/selected/ussuri/project-ptl-and-contrib-docs.html as well.... | 15:21 |
| *** sshnaidm has quit IRC | 15:22 | |
| noonedeadpunk | mgariepy: do you have any preferences regarding ptg? | 15:24 |
| *** pcaruana has quit IRC | 15:27 | |
| mgariepy | i would be ok from 12:00pm utc to 8:00pm utc. | 15:39 |
| mgariepy | but i'm not 100% sure i'll join or have a lot of stuff to contribute. | 15:39 |
| *** pcaruana has joined #openstack-ansible | 15:39 | |
| mnaser | i can work with any timezone :) | 15:41 |
| mgariepy | mnaser, you are a machine :) haha | 15:45 |
| noonedeadpunk | So I'd say 13UTC-17UTC then, if we dont mind tripleo and kolla being at the same time | 15:51 |
| noonedeadpunk | and jrosser and gshippey would be able to join | 15:52 |
| noonedeadpunk | mnaser: do we want to book for 2 days or just 1 would be enough? | 15:52 |
| guilhermesp | as mnaser i can also work with any timezone :) | 15:52 |
| *** gyee has joined #openstack-ansible | 15:53 | |
| noonedeadpunk | actually the only which I don't like is 4UTC - 8 UTC.... | 15:53 |
| jrosser | 13-17UTC works for us here | 15:54 |
| noonedeadpunk | ok, nice | 15:54 |
| jrosser | but it's cool to have it longer too, don't limit it to that | 15:54 |
| gshippey | anytime works for me really - but less keen on being up through the night | 15:54 |
| *** sshnaidm has joined #openstack-ansible | 15:55 | |
| noonedeadpunk | jrosser: yeah, I'd agree to prolonge it but not sure if we're able to? I can also maybe take 21-00 - 22-00? | 15:56 |
| *** udesale_ has joined #openstack-ansible | 15:57 | |
| noonedeadpunk | Let's probably take tuesday as well? | 15:57 |
| noonedeadpunk | Ok, we can't sign up for more than 4 hours... | 15:59 |
| noonedeadpunk | so, folks, maybe we should signup for 4 hours on tuesday and maybe another 4 hours at wendsday but in different time slot? | 15:59 |
| *** udesale has quit IRC | 16:00 | |
| *** cshen has quit IRC | 16:02 | |
| noonedeadpunk | let's say 13-00 - 17-00 on tuesday and 21-00 - 01-00 on wednesday? | 16:04 |
| gshippey | that sounds good to me :) I'm quite happy to do the late slot as well | 16:05 |
| noonedeadpunk | not sure I want to pickup monday as pretty sure there will be some technical hiccup | 16:08 |
| noonedeadpunk | ok, booked room5 | 16:11 |
| mgariepy | quick one, is anyone seen keystone keystone.token.token_formatters generating a token of 268 char when the limit should 255 ? | 16:11 |
| noonedeadpunk | mnaser: yould you like to fill in survey? https://openstackfoundation.formstack.com/forms/june2020_virtual_ptg_survey | 16:13 |
| noonedeadpunk | or I can do it eventually | 16:14 |
| *** udesale_ has quit IRC | 16:23 | |
| noonedeadpunk | ok, submited | 16:27 |
| *** thuydang has quit IRC | 16:29 | |
| noonedeadpunk | can I have another vote on https://review.opendev.org/#/c/722020/ and https://review.opendev.org/#/c/722021/ ? | 16:29 |
| *** evrardjp has quit IRC | 16:35 | |
| *** cshen has joined #openstack-ansible | 16:35 | |
| *** evrardjp has joined #openstack-ansible | 16:35 | |
| *** cshen has quit IRC | 16:40 | |
| *** rpittau is now known as rpittau|afk | 16:49 | |
| mgariepy | huh, the len of the token is caused by ldap .. it seems. | 16:52 |
| noonedeadpunk | mgariepy: sorry never was facing smth like that | 16:54 |
| mgariepy | no worries, but my issue was something else, it's only a warning.. | 17:08 |
| mgariepy | not an actual error. | 17:08 |
| mgariepy | uuid mapping in ldap stuff, but my issue was with app_credential about 14-18inch from my screen | 17:11 |
| *** cshen has joined #openstack-ansible | 17:15 | |
| openstackgerrit | Merged openstack/openstack-ansible-tests stable/pike: Retire repo_build and pip_install roles https://review.opendev.org/722021 | 17:17 |
| *** cshen has quit IRC | 17:20 | |
| openstackgerrit | Merged openstack/openstack-ansible-os_tempest stable/stein: Pin neutron-tempest-plugin to 0.9.0 for Stein branch. https://review.opendev.org/721320 | 17:27 |
| openstackgerrit | Merged openstack/openstack-ansible-tests stable/queens: Retire repo_build and pip_install roles https://review.opendev.org/722020 | 17:42 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-tests stable/ocata: Retire repo_build and pip_install roles https://review.opendev.org/722022 | 17:43 |
| *** sshnaidm is now known as sshnaidm|afk | 17:50 | |
| *** yolanda has quit IRC | 17:55 | |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-tests stable/ocata: Retire repo_build and pip_install roles https://review.opendev.org/722022 | 18:03 |
| *** jbadiapa has quit IRC | 18:04 | |
| *** cshen has joined #openstack-ansible | 18:30 | |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible-haproxy_server master: Add extra controls for frontend redirects https://review.opendev.org/722419 | 18:34 |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible-haproxy_server master: Allow extra parameters to be passed to certbot https://review.opendev.org/722420 | 18:34 |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible-haproxy_server master: Allow multiple methods of installing certbot https://review.opendev.org/722421 | 18:34 |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible-haproxy_server master: Add option to use distro provided certbot package https://review.opendev.org/722422 | 18:34 |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible-haproxy_server master: Install self-signed certificate before certbot has run for the first time https://review.opendev.org/722423 | 18:34 |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible-haproxy_server master: Add documentation for HA HAProxy with LetsEncrypt certificates. https://review.opendev.org/722424 | 18:34 |
| jrosser | mnaser: noonedeadpunk i did a ton of work on the haproxy role ^ | 18:35 |
| jrosser | this lets you have letsencrypt with haproxy + keepalived, i need to test it yet with > 1 node but its looking good so far on a single one | 18:36 |
| noonedeadpunk | oh, cool, will take a look tomorrow | 18:36 |
| jrosser | review welcomed there as it's quite a big set of changes | 18:36 |
| * jrosser wanted to get something out there before my day off tomorrow :) | 18:37 | |
| *** KeithMnemonic has joined #openstack-ansible | 18:46 | |
| *** cshen has quit IRC | 19:06 | |
| *** gshippey has quit IRC | 19:09 | |
| *** cshen has joined #openstack-ansible | 19:33 | |
| *** vesper has joined #openstack-ansible | 19:37 | |
| *** vesper11 has quit IRC | 19:38 | |
| *** cshen has quit IRC | 19:38 | |
| *** mgariepy has quit IRC | 19:47 | |
| *** cshen has joined #openstack-ansible | 20:00 | |
| *** cshen has quit IRC | 20:04 | |
| *** mloza has quit IRC | 20:27 | |
| *** mloza has joined #openstack-ansible | 20:29 | |
| *** aedc has quit IRC | 20:49 | |
| openstackgerrit | Merged openstack/openstack-ansible-lxc_hosts master: Updated from OpenStack Ansible Tests https://review.opendev.org/709471 | 20:50 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-os_tempest stable/rocky: Pin neutron-tempest-plugin to 0.9.0 for Stein branch. https://review.opendev.org/722498 | 21:07 |
| *** watersj has quit IRC | 22:04 | |
| *** ansmith has quit IRC | 22:17 | |
| *** tosky has quit IRC | 22:58 | |
| redkrieg | I'm having a hard time figuring out what I'm doing wrong with my openstack_user_config.yml file. I want to avoid repeating each host definition in my custom group and compute_hosts. I've been trying to do something with & and * as is used in the forthcoming example, but I am not getting the syntax right. Is there documentation somewhere for using these to put the members of multiple groups | 23:46 |
| redkrieg | in to one parent group? http://paste.openstack.org/show/792627/ | 23:46 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!