| opendevreview | Satish Patel proposed openstack/openstack-ansible master: Set FQDN hostname for aio https://review.opendev.org/c/openstack/openstack-ansible/+/801932 | 04:25 |
|---|---|---|
| opendevreview | Andrew Bonney proposed openstack/openstack-ansible-os_keystone stable/victoria: Improvements to federation packaging https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/801940 | 07:28 |
| *** rpittau|afk is now known as rpittau | 07:35 | |
| noonedeadpunk | mornings | 08:47 |
| noonedeadpunk | andrewbonney: if you're around vote on https://review.opendev.org/c/openstack/openstack-ansible-os_panko/+/799806 would be great :) | 08:48 |
| andrewbonney | I'll take a look shortly | 08:48 |
| opendevreview | Merged openstack/openstack-ansible-os_panko master: Deprecate os-panko role https://review.opendev.org/c/openstack/openstack-ansible-os_panko/+/799806 | 08:52 |
| *** mgoddard- is now known as mgoddard | 12:18 | |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-haproxy_server master: Fix service removal condition https://review.opendev.org/c/openstack/openstack-ansible-haproxy_server/+/801910 | 12:25 |
| admin1 | hi guys .. i am trying to use (untagged network ) for the first time .. do I have to use flat or can i use vlan but not use segmentation id ? | 12:57 |
| admin1 | and is the bridge still on br-vlan .. just without a tag ? | 12:57 |
| spatel | i think without tag would be flat network | 13:01 |
| admin1 | do i need to create br-flat? | 13:02 |
| spatel | i believe you just need to map your flat network with phynet interface ( either physical nic or br nic) | 13:04 |
| spatel | something like this physical_interface_mappings = provider:eth1 | 13:05 |
| spatel | flat_networks = provider | 13:06 |
| spatel | This is good example - https://docs.openstack.org/newton/install-guide-rdo/launch-instance-networks-provider.html | 13:06 |
| spatel | noonedeadpunk jrosser are you there? | 13:06 |
| spatel | want to talk about - https://review.opendev.org/c/openstack/openstack-ansible/+/801932 | 13:07 |
| noonedeadpunk | I'm around | 13:09 |
| noonedeadpunk | damn, ceph-ansible still make troubles with depth.... | 13:11 |
| noonedeadpunk | admin1: eventually flat should be different interface from br-vlan | 13:13 |
| noonedeadpunk | It can be tagged vlan itself. But it must be some existing interface other then br-vlan preferably with consistent naming | 13:14 |
| noonedeadpunk | jrosser: do you think instead of depth, we should use --shallow-since or just --single-branch ? | 13:15 |
| noonedeadpunk | so with bump we can update some shallow_date and use it for shallow-since ? | 13:18 |
| spatel | sorry i was in meeting.. | 13:24 |
| spatel | noonedeadpunk i found OVN is failing because of hostname issue specially aio build | 13:24 |
| spatel | we are just setting hostname but not FQDN | 13:25 |
| spatel | looking for why my patch failed with metal build | 13:25 |
| noonedeadpunk | It failed for upgrade and I know why.... | 13:26 |
| noonedeadpunk | spatel: but what to do for ppl who doen't have fqdn? | 13:26 |
| noonedeadpunk | or they've not set it? It's not really obvious for me, why I can't set hostname I want | 13:27 |
| spatel | why don't you have fqdn in production? | 13:27 |
| spatel | that would be bad config right? | 13:27 |
| spatel | when i build aio in lab OVN set hostname aio1.openstack.local but hostname command return aio1 (this is what nova-compute also use) | 13:29 |
| spatel | noonedeadpunk check this out - https://bugs.launchpad.net/openstack-ansible/+bug/1922978 | 13:30 |
| spatel | bottom line whatever hostname this command return we need same in OVN openvswitch - openstack compute service list | 13:31 |
| spatel | when i build my ovn in production i didn't see this issue, so look like aio1 related issue | 13:31 |
| noonedeadpunk | openstack compute service list does return hostname not fqdn | 13:32 |
| noonedeadpunk | hypervisors do return fqdn | 13:32 |
| spatel | in my production its fqdn compute01.foo.com | 13:32 |
| noonedeadpunk | ok, let me clarify it - it takes from python socket.gethostname() | 13:33 |
| noonedeadpunk | which eventually depends on the order of records in hosts file | 13:33 |
| noonedeadpunk | so what I have in there is | 13:34 |
| noonedeadpunk | | 190 | nova-compute | cc-compute12-dx1 | nova | enabled | up | 2021-07-23T13:34:29.000000 | | 13:35 |
| noonedeadpunk | which kind of means that ovn won't work for me? | 13:35 |
| spatel | In my production, i don't add any compute hostname in /etc/hosts file and DNS. my nova compute pick hostname command and my hostname return compute01.foo.com | 13:35 |
| noonedeadpunk | which is kind of weird don't you think? | 13:35 |
| spatel | tell me one thing.. what hostname command return on your cc-compute12-dx1 ? | 13:36 |
| noonedeadpunk | https://paste.opendev.org/show/807683/ | 13:37 |
| noonedeadpunk | and aio kind of the same? | 13:37 |
| spatel | this is strange.. because in my production i set hostname compute01.foo.com in /etc/hostname file | 13:37 |
| spatel | so in my case hostname command return FQDN | 13:37 |
| spatel | its personal choice to set FQDN or not.. | 13:38 |
| spatel | This is going to be problem who deploying ovn ( if they won't follow rules related hostname then it will be issue which we are having right now) | 13:39 |
| spatel | may be we can put some more check in ansible to find out correct hostname (short or fqdn and set according) | 13:40 |
| noonedeadpunk | https://paste.opendev.org/show/807684/ | 13:41 |
| noonedeadpunk | btw, only hostname -A should retrun fqdn afaik | 13:42 |
| noonedeadpunk | or hostname --fqdn | 13:43 |
| noonedeadpunk | but just hostname is fine to be non-fqdn | 13:43 |
| spatel | without this patch https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/793009/3/tasks/providers/setup_ovs_ovn.yml#b17 | 13:43 |
| spatel | james was using command: "ovs-vsctl set open_vswitch . external-ids:hostname='{{ ansible_facts['hostname'] }}'" | 13:43 |
| spatel | ansible_facts['hostname'] always return short hostname | 13:43 |
| noonedeadpunk | So. ovn hostname should match nova one, right? | 13:44 |
| noonedeadpunk | Or it should be just fqdn and doesn't really matter if it's mathing nova or not? | 13:44 |
| spatel | when i deploy OSA in my production and my compute node hostname command return compute01.foo.com but ansible tell ovs to use compute01 and that break my environment (we need ovs and nova compute hostname similar) | 13:45 |
| noonedeadpunk | yeah I see, that's valid bug | 13:45 |
| noonedeadpunk | Not sure what is valid solution for it though :) | 13:45 |
| spatel | we need to come up with smart way to handle this dual problem where some people set fqdn and some people set fqdn for compute nodes | 13:46 |
| noonedeadpunk | As hostname thing is really weird thing and depends on the record in /etc/hosts | 13:46 |
| noonedeadpunk | yeah, totally agree | 13:46 |
| spatel | what if OSA set hostname on compute node and make it short even someone set fqdn ? | 13:46 |
| noonedeadpunk | we will break things during upgrade | 13:48 |
| spatel | for experiment i changed hostname of my compute node and restart nova service which re-register new compute name with newer name and ovn got confused :) | 13:48 |
| noonedeadpunk | As all computes and net nodes and cinder services and etc would be renamed | 13:48 |
| jrosser | hostname returning short name is correct imho | 13:48 |
| noonedeadpunk | It will break masakari and corosync cluster most likely as well | 13:48 |
| jrosser | it is deployed problem to have proper host / dns setup, not OSA | 13:49 |
| jrosser | *deployer | 13:49 |
| noonedeadpunk | `You can check the FQDN using hostname --fqdn` | 13:49 |
| spatel | ovn has only issue with compute hostname, it doesn't care about other hostname.. | 13:49 |
| noonedeadpunk | man hostname | 13:49 |
| noonedeadpunk | `If a machine has multiple network interfaces/addresses or is used in a mobile environment, then it may either have multiple FQDNs/domain names or none at all.` | 13:50 |
| spatel | agreed its not our problem to fix how people use hostname in their datacenter | 13:50 |
| spatel | we just need to educate or document about this issue to make sure your hostname is short otherwise it will create issue | 13:50 |
| noonedeadpunk | But yes, I agree that it's not osa business how ppl set hostnames | 13:51 |
| noonedeadpunk | but we must deal with these choices... | 13:51 |
| spatel | so lets delete this patch and just document when someone go deploy OSA in production | 13:51 |
| spatel | https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/793009 | 13:51 |
| jrosser | this is putting just fqdn in /etc/hosts and not a third parameter for hostname? | 13:52 |
| spatel | I think james created this because i opened bug issue when i was deploying in production and hit this same issue related hostname conflict | 13:52 |
| noonedeadpunk | I think as workaround we can retrieve actual hostname with socket.gethostname | 13:52 |
| noonedeadpunk | as that's what used by nova and will give valid result anyway | 13:53 |
| spatel | jrosser does nova compute use /etc/hosts file to read hostname? | 13:53 |
| noonedeadpunk | spatel what `{{ ansible_facts['nodename'] }}` would return for you? | 13:54 |
| spatel | compute01 short name | 13:54 |
| noonedeadpunk | so nodename and hostname are the same? | 13:55 |
| noonedeadpunk | doh ( | 13:55 |
| spatel | ansible localhost -m ansible.builtin.setup | 13:55 |
| spatel | doesn't ansible read that from facts | 13:55 |
| spatel | run this command ansible localhost -m ansible.builtin.setup and see what is hostname facts | 13:55 |
| noonedeadpunk | I just thought that nodename would return real hostname, and not trimmed one | 13:56 |
| spatel | and nodename is always FQDN | 13:57 |
| noonedeadpunk | wait | 13:57 |
| noonedeadpunk | nodename is not fqdn | 13:57 |
| noonedeadpunk | `ansible_facts['nodename']": "cc-compute12-dx1"` | 13:57 |
| spatel | in my case "ansible_nodename": "os-osa.v1v0x.net", | 13:58 |
| spatel | I am talking about this returning all these facts- ansible localhost -m ansible.builtin.setup | 13:59 |
| noonedeadpunk | then we can just replace there hostname with nodename ?:) | 13:59 |
| noonedeadpunk | https://paste.opendev.org/show/807685/ | 13:59 |
| noonedeadpunk | first is nodename, second is hostname and third is fqdn | 14:00 |
| jrosser | why don’t we make the var looked up in the facts itself a variable | 14:00 |
| jrosser | ansible_facts[neutron_ovs_hostname_fact] | 14:01 |
| jrosser | and default it to ‘hostname’ | 14:01 |
| jrosser | then whatever strangeness exists, a deployment can pick the fact that works for them | 14:02 |
| noonedeadpunk | Why not default to nodename? As it seems like nodename == hostname bash command? | 14:02 |
| noonedeadpunk | Or I'm wrong? | 14:02 |
| jrosser | well, default as appropriate:) | 14:02 |
| noonedeadpunk | but yeah, I agree that we can jsut add neutron_ovs_hostname_fact | 14:03 |
| noonedeadpunk | we need to sort out depth issue once and forever as well... https://zuul.opendev.org/t/openstack/build/4b4487db6123489dbb483dbdcc8134c3/log/job-output.txt#3736 | 14:04 |
| spatel | Look at here these three should be similar when it comes to ovn - https://paste.opendev.org/show/807686/ | 14:06 |
| noonedeadpunk | yeah, ok, so let's indeed ad variable which would default to nodename | 14:06 |
| spatel | are you saying use nodename here like this command: "ovs-vsctl set open_vswitch . external-ids:hostname='{{ ansible_facts['nodename'] }}'" | 14:08 |
| spatel | or other work around we can set host = compute01 in /etc/nova.conf (that way no matter what is your OS hostname it will use same name) | 14:10 |
| spatel | noonedeadpunk if you know how to fix this nodename vs hostname could you cut the patch :) so i can verify my lab | 14:11 |
| *** rpittau is now known as rpittau|afk | 14:13 | |
| noonedeadpunk | ok, I got the issue... It's kind of the same one that present for libvirt vs nova way of reporting hostname.... | 14:14 |
| noonedeadpunk | So I think the only solution here is try to set hostname to be persistant for OVN | 14:22 |
| noonedeadpunk | and during server restart ovn retrieves hostname on it's own and not using what has been previoously set for it. | 14:22 |
| noonedeadpunk | I winder what argiments can be passed to the OVN_CTL_OPTS | 14:23 |
| spatel | technically its OVS switch hostname to report to ovn_northd | 14:23 |
| noonedeadpunk | So I guess that OVS_EXTRA can provide hostname. But I see it being used only in network-scripts | 14:24 |
| spatel | did you reboot server to verify what hostname ovs switch picking up | 14:24 |
| noonedeadpunk | I don't have ovn lab so.... | 14:24 |
| noonedeadpunk | but it's what is written in bug:) | 14:25 |
| spatel | i think once you set hostname in ovs switch it will stay same in ovsdb database not going to change after reboot | 14:25 |
| noonedeadpunk | So ansible was setting everyting correctly, but after reboot nova-compute service differs with ovn | 14:25 |
| spatel | I think james got confused there | 14:25 |
| noonedeadpunk | `Pre-reboot external_ids : {hostname=compute3`, `Post-reboot external_ids : {hostname=compute3.openstack.local` | 14:26 |
| noonedeadpunk | so that patch is not going to help at all.... | 14:26 |
| noonedeadpunk | well, setting hostname doesn't matter that way as well... | 14:27 |
| spatel | let me give you example, if i set compute node hostname to hostnamectl set-hostname foo.bar.com and restart nova-compute then openstack report foo.bar.com in openstack compute service list and that is what ovn/ovs switch doesn't like | 14:27 |
| noonedeadpunk | so task now find the way how to make ovs respect provided hostname. | 14:27 |
| noonedeadpunk | I think the one way would be to add override to systemd service | 14:27 |
| noonedeadpunk | and set some ENV or add hook after service start to adjust hostname for ovs | 14:28 |
| noonedeadpunk | Um. We're not talking about hostname change do we? | 14:28 |
| noonedeadpunk | So current problem is the same as nova has with libvirt | 14:28 |
| noonedeadpunk | So ovs, like libvirt, are C written. So in C with gethostname you will get fqdn, while in Python gethostname gets hostname | 14:29 |
| spatel | i am just giving idea that ovn only care about what nova hostname show up in openstack, so question is how to make sure nova-compute hostname and ovn/ovn hostname match | 14:29 |
| noonedeadpunk | Eventually this results in different hostnames being reported for openstack hypervisor list and openstack compute service list | 14:30 |
| noonedeadpunk | (it's kind of fact I see atm) | 14:30 |
| noonedeadpunk | With OVS it's really the same | 14:30 |
| spatel | if we tell /etc/nova.conf host = compute01 and same time tell ovs hostname = compute01 then problem can be solved | 14:30 |
| noonedeadpunk | except it breaks deployemtns | 14:30 |
| noonedeadpunk | how to tell ovs? | 14:30 |
| noonedeadpunk | I dunno how to tell ovs, and that's the current problem | 14:31 |
| spatel | command: "ovs-vsctl set open_vswitch . external-ids:hostname='compute01'}'" | 14:31 |
| noonedeadpunk | yes, but it's reseted with reboot | 14:31 |
| spatel | no it won't | 14:31 |
| noonedeadpunk | and after reboot (or service restart even) OVS rollsback hostname to fqdn | 14:31 |
| spatel | it will set on ovsdb database | 14:31 |
| noonedeadpunk | and does not respect what we have set with that command | 14:32 |
| spatel | i can test in few min.. if you won't i don't think it will get change | 14:32 |
| noonedeadpunk | well, ok, have you read bug report then?:) | 14:32 |
| spatel | want* | 14:32 |
| spatel | yes i did read | 14:32 |
| noonedeadpunk | Eventually this bug report is specifically about that | 14:32 |
| noonedeadpunk | as once host is rebooted - ovs change hostname | 14:32 |
| spatel | how about i can verify right now | 14:33 |
| noonedeadpunk | sure, o rush | 14:33 |
| noonedeadpunk | *no | 14:33 |
| spatel | what test you want to run? | 14:33 |
| spatel | change hostname of OS and reboot ? | 14:33 |
| noonedeadpunk | oh... you would need to have different output for hostname and hostname --fqdn first ) | 14:33 |
| noonedeadpunk | so basically AIO way of things | 14:34 |
| spatel | https://paste.opendev.org/show/807688/ | 14:34 |
| noonedeadpunk | you won't reproduce issue | 14:35 |
| noonedeadpunk | so. hostname == os-compute-1 and hostname --fqdn == os-compute-1.v1v0x.net. Then `ovs-vsctl set open_vswitch . external-ids:hostname='os-compute-1'}'` | 14:36 |
| noonedeadpunk | after reboot you will get external-ids:hostname='os-compute-1.v1v0x.net' | 14:36 |
| noonedeadpunk | I think we actually have 2 issues now ) | 14:36 |
| spatel | hmm | 14:37 |
| noonedeadpunk | 1st is that we always set external-ids:hostname='os-compute-1' ? And the second is that OVS does not preserve hostname | 14:37 |
| spatel | hmm | 14:40 |
| spatel | we don't have direct proof ovs doesn't preserve hostname | 14:41 |
| noonedeadpunk | it's claimed in bug report at least... | 14:41 |
| spatel | if that is the case then we can add command in ovs systemd to set hostname | 14:41 |
| noonedeadpunk | I'd need to spawn sandbox to prove that | 14:41 |
| noonedeadpunk | yeah, or maybe it respects some ENV? | 14:41 |
| noonedeadpunk | like OVS_EXTRA ? | 14:42 |
| spatel | yes we can add something in systemd to make it work | 14:42 |
| spatel | let me run some experiments and see.. i think we both on same page :) | 14:42 |
| spatel | lets me go back to my lab and play with it | 14:43 |
| spatel | OVS by default read hostname --fqdn | 14:43 |
| noonedeadpunk | Yeah I gues we are now! Thanks for checking that in! | 14:43 |
| noonedeadpunk | I think it's more about how C reads hostname - it's always fqdn, yes | 14:44 |
| noonedeadpunk | but nova and other services use python way, which would be different anyway | 14:44 |
| spatel | if somehow we can also tell nova to use that then things can get aligned | 14:44 |
| noonedeadpunk | and cinder and neutron and etc etc etc | 14:45 |
| spatel | ovn doesn't care about cinder and neutron hostname | 14:45 |
| noonedeadpunk | So nova has overcomed with that with libvirt actually | 14:45 |
| spatel | ovn only care about what compute nodes reporting | 14:45 |
| noonedeadpunk | Yes, but nova does kind of? | 14:45 |
| spatel | Yes only compute not api services etc | 14:45 |
| noonedeadpunk | so then we would need to change all services to align ovs? | 14:45 |
| spatel | i don't know if all other service care about hostname or not but i have noticed only issue with nova-compute hostname which run openvswitch service and that is where name should be match | 14:47 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: [DNM] Add Vault role support https://review.opendev.org/c/openstack/openstack-ansible/+/800787 | 14:47 |
| spatel | nova-compute report hostname to ovn_southd and that is how neutron knows where to bind port, if that hostname doesn't match then neutron failed bind | 14:48 |
| noonedeadpunk | Yeah, but I mean I'd rather make ovs match rest of openstack then other way | 14:48 |
| noonedeadpunk | it's really same issue kind of as hypervisor names are reported with libvirt, but nova knows how to work with that properly https://paste.opendev.org/show/807689/ | 14:49 |
| spatel | This is it... that is the problem.. | 14:49 |
| spatel | multiple hostname representation | 14:50 |
| noonedeadpunk | It's not a problem until you use ovn:) | 14:50 |
| spatel | I believe this is the issue with AIO only... i haven't seen this issue in production | 14:50 |
| noonedeadpunk | I have all my prods set that way (except really several) | 14:50 |
| spatel | I think for now we can remove this patch and that will pass all build related aio - https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/793009 | 14:51 |
| noonedeadpunk | So it;s really about if you follow RFC or jsut go convenient way... | 14:51 |
| spatel | and we can make doc for production so people will be aware | 14:51 |
| noonedeadpunk | I think we still need to set nodename? Or ansible_facts['hostname'] gives correct result for you? | 14:52 |
| spatel | ansible_facts['hostname'] giving me short name | 14:53 |
| spatel | if we keep ansible_facts['hostname'] then aio will pass all OVN patches | 14:53 |
| noonedeadpunk | so I guess 2 things: 1. replace hostname with nodename 2. make ovs respect hostname that was set (most likely with systemd override) | 14:53 |
| noonedeadpunk | does this make sense? | 14:54 |
| spatel | do you want me to try this with nodename. -> external-ids:hostname='{{ ansible_facts['nodename'] }}' | 14:54 |
| spatel | and re-deploy aio | 14:54 |
| noonedeadpunk | yeah | 14:54 |
| spatel | let me deploy aio with nodename and see if it pass or fail | 14:55 |
| noonedeadpunk | I think it would be fine both for AIO and for scneario where hostname == hostname --fqdn | 14:55 |
| spatel | yes.. | 14:55 |
| spatel | lets me spin up my lab again and verify | 14:55 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/ansible-role-vault master: Initial commit to Vault role https://review.opendev.org/c/openstack/ansible-role-vault/+/800792 | 14:55 |
| anskiy | ansible_facts['hostname'] would always return short name: https://github.com/ansible/ansible/blob/stable-2.11/lib/ansible/module_utils/facts/system/platform.py#L52-L54 | 14:57 |
| noonedeadpunk | great link! | 15:01 |
| noonedeadpunk | yes, we have to use nodename | 15:01 |
| spatel | i am going to try and will post result here... to see how it goes | 15:04 |
| anskiy | platform and socket are both standart python modules; And according to straces from these: https://paste.opendev.org/show/807690/ it looks like socket.getfqdn() is, like, real networking stuff opposed to platform.node(), which looks like just parsing output from uname syscall | 15:06 |
| opendevreview | Merged openstack/openstack-ansible-os_keystone master: Fix oidc scope misspelling in newer releases https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/801604 | 15:10 |
| opendevreview | Merged openstack/openstack-ansible-os_adjutant master: Run handlers only against present services https://review.opendev.org/c/openstack/openstack-ansible-os_adjutant/+/801188 | 15:13 |
| opendevreview | Merged openstack/openstack-ansible-os_octavia master: Fix self-signed certs distribution https://review.opendev.org/c/openstack/openstack-ansible-os_octavia/+/801505 | 15:16 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-os_octavia stable/wallaby: Fix self-signed certs distribution https://review.opendev.org/c/openstack/openstack-ansible-os_octavia/+/801889 | 15:37 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/ansible-role-vault master: Initial commit to Vault role https://review.opendev.org/c/openstack/ansible-role-vault/+/800792 | 15:49 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/ansible-role-vault master: Initial commit to Vault role https://review.opendev.org/c/openstack/ansible-role-vault/+/800792 | 15:52 |
| spatel | noonedeadpunk around ? | 20:01 |
| admin1 | my flat network adding creates a bridge .. but that bridge does not conenct to any physical network | 20:37 |
| admin1 | i added the network like this: https://pastebin.com/wjkJqwqZ | 20:38 |
| admin1 | so br-vlan has eth1 which is where the public ips are accessible ( directly connected ot the switch ) | 20:38 |
| admin1 | this is a linuxbridge setup | 20:39 |
| admin1 | does not explain flat .. https://docs.openstack.org/openstack-ansible-os_neutron/latest/app-openvswitch.html#configuring-bridges-linux-bridge | 20:40 |
| admin1 | i have been using only br-vlan for the last 5-6 years of using osa .. .first time doing flat | 20:41 |
| jrosser | admin1: every time you build an AIO it has a flat provider network https://github.com/openstack/openstack-ansible/blob/master/etc/openstack_deploy/openstack_user_config.yml.aio.j2#L161 | 21:05 |
| jrosser | the “configuring bridges” stuff really is about infrastructure, not so much about neutron | 21:07 |
| jrosser | look at how eth12 works in the AIO | 21:09 |
| jrosser | it is for the deployer to arrange the flat interface (in this case eth12) to be connected to the right place | 21:09 |
| jrosser | for the AIO the bootstrap scripts wire eth12 to br-vlan with a veth, but that’s totally just for CI convenience | 21:10 |
| jrosser | so set it up how you need it | 21:11 |
| admin1 | i have eth1 which has access to the public IPs .. and i can put any bridge on top of it | 21:29 |
| admin1 | isn't that how it works ? | 21:29 |
| admin1 | i tried br-flat and just mentoning eth1.. but it also did not worked | 21:29 |
| admin1 | so i am not sure how flat actually works | 21:30 |
| admin1 | i am going to build one aio now and check | 21:30 |
| admin1 | thank you jrosser.. i will check it out | 21:45 |
| jrosser | admin1: neutron wants an interface, too much worrying about bridges here - it’s not that complicated | 21:58 |
| admin1 | jrosser -- here is the old one and new one https://pastebin.com/EJLU4bdz -- both did not worked | 22:07 |
| admin1 | br-vlan is on top of eth1 | 22:07 |
| opendevreview | Satish Patel proposed openstack/openstack-ansible-os_neutron master: Set ovn hostname using nodename facts https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/802134 | 22:41 |
| opendevreview | Satish Patel proposed openstack/openstack-ansible-os_neutron master: Change OVN metadata protocol to https https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/802135 | 22:54 |
| opendevreview | Satish Patel proposed openstack/openstack-ansible-os_neutron master: Set ovn hostname using nodename facts https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/802134 | 22:57 |
| spatel | hope these patches solved all OVN tempest issue | 22:58 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!