| *** errr_ has joined #openstack-ansible | 00:02 | |
| *** sdake has joined #openstack-ansible | 00:03 | |
| *** errr has quit IRC | 00:04 | |
| *** sdake has quit IRC | 00:05 | |
| *** sdake has joined #openstack-ansible | 00:12 | |
| *** markvoelker has joined #openstack-ansible | 00:15 | |
| *** sdake has quit IRC | 00:44 | |
| *** markvoelker has quit IRC | 00:47 | |
| *** kukacz has quit IRC | 01:10 | |
| *** kukacz has joined #openstack-ansible | 01:12 | |
| *** sdake has joined #openstack-ansible | 01:14 | |
| *** DanyC has quit IRC | 01:25 | |
| *** nurdie has joined #openstack-ansible | 01:26 | |
| *** sdake has quit IRC | 01:27 | |
| *** nurdie has quit IRC | 01:31 | |
| *** markvoelker has joined #openstack-ansible | 01:44 | |
| *** nurdie has joined #openstack-ansible | 01:47 | |
| *** nurdie has quit IRC | 01:52 | |
| *** hwoarang has quit IRC | 01:57 | |
| *** hwoarang has joined #openstack-ansible | 02:01 | |
| *** DanyC has joined #openstack-ansible | 02:03 | |
| *** nurdie has joined #openstack-ansible | 02:16 | |
| *** markvoelker has quit IRC | 02:18 | |
| *** nurdie has quit IRC | 02:22 | |
| *** nurdie has joined #openstack-ansible | 02:37 | |
| *** nurdie has quit IRC | 02:42 | |
| *** shyamb has joined #openstack-ansible | 03:06 | |
| *** nurdie has joined #openstack-ansible | 03:07 | |
| *** nurdie has quit IRC | 03:11 | |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible master: [WIP] Use the Mitogen connection method in OSA https://review.openstack.org/591236 | 03:14 |
|---|---|---|
| prometheanfire | cloudnull: power back? | 03:16 |
| cloudnull | yup. | 03:20 |
| cloudnull | its back on | 03:20 |
| cloudnull | however, internet is on a UPS | 03:20 |
| cloudnull | unifi + cyberpower 700VA (https://amzn.to/2DbROpU) | 03:22 |
| cloudnull | works great! | 03:22 |
| prometheanfire | :D | 03:24 |
| *** macza has joined #openstack-ansible | 03:33 | |
| *** macza has quit IRC | 03:37 | |
| *** shyamb has quit IRC | 03:43 | |
| *** gkadam has joined #openstack-ansible | 03:48 | |
| *** udesale has joined #openstack-ansible | 04:07 | |
| *** slaweq has joined #openstack-ansible | 04:37 | |
| *** sohny has joined #openstack-ansible | 04:52 | |
| *** TxGirlGeek has joined #openstack-ansible | 04:53 | |
| *** TxGirlGeek has quit IRC | 05:01 | |
| *** hwoarang has quit IRC | 05:09 | |
| *** ArchiFleKs has quit IRC | 05:10 | |
| *** hwoarang has joined #openstack-ansible | 05:12 | |
| *** ArchiFleKs has joined #openstack-ansible | 05:15 | |
| *** macza has joined #openstack-ansible | 05:29 | |
| openstackgerrit | Chandan Kumar proposed openstack/openstack-ansible-os_tempest master: Ping router once it is created https://review.openstack.org/633883 | 05:37 |
| *** macza has quit IRC | 05:37 | |
| *** slaweq has quit IRC | 05:49 | |
| openstackgerrit | Merged openstack/openstack-ansible-ops master: Omit dahsboard on elk setup by default https://review.openstack.org/633392 | 06:02 |
| *** markvoelker has joined #openstack-ansible | 06:56 | |
| *** slaweq has joined #openstack-ansible | 06:58 | |
| *** aludwar has quit IRC | 07:04 | |
| *** aludwar has joined #openstack-ansible | 07:06 | |
| *** slaweq has quit IRC | 07:10 | |
| *** aludwar has quit IRC | 07:14 | |
| *** mkuf_ has quit IRC | 07:32 | |
| *** aludwar has joined #openstack-ansible | 07:35 | |
| *** pcaruana has joined #openstack-ansible | 07:35 | |
| *** mkuf has joined #openstack-ansible | 07:36 | |
| *** pcaruana has quit IRC | 07:42 | |
| *** pcaruana has joined #openstack-ansible | 07:42 | |
| *** macza has joined #openstack-ansible | 07:46 | |
| *** macza has quit IRC | 07:50 | |
| *** DanyC has quit IRC | 07:54 | |
| *** kopecmartin|off is now known as kopecmartin | 08:02 | |
| chkumar|ruck | jrosser: Hello | 08:07 |
| chkumar|ruck | jrosser: I need some help at this part https://review.openstack.org/#/c/633883/7/tasks/tempest_resources.yml@242 | 08:09 |
| chkumar|ruck | jrosser: please have a loo | 08:09 |
| chkumar|ruck | look thanks! | 08:09 |
| *** macza has joined #openstack-ansible | 08:21 | |
| *** macza has quit IRC | 08:22 | |
| *** macza has joined #openstack-ansible | 08:22 | |
| *** macza has quit IRC | 08:26 | |
| *** tosky has joined #openstack-ansible | 08:42 | |
| jrosser | chkumar|ruck: hi! let me look | 08:43 |
| ioni | hello guys | 08:50 |
| ioni | i have a small question | 08:50 |
| ioni | https://docs.openstack.org/openstack-ansible/queens/admin/scale-environment.html | 08:50 |
| ioni | what does this do? | 08:50 |
| ioni | ansible nova_all -m setup -a 'filter=ansible_local gather_subset="!all"' | 08:50 |
| jrosser | that is an ansible ad-hoc command (rather than playbook) run against the nova_all group, whch runs the setup module (https://docs.ansible.com/ansible/latest/modules/setup_module.html) with the arguments specified after -a. | 08:52 |
| *** Emine has joined #openstack-ansible | 08:53 | |
| ioni | it's updating the facts? | 08:54 |
| jrosser | looks that way, yes | 08:56 |
| *** slaweq has joined #openstack-ansible | 08:56 | |
| *** gillesMo has joined #openstack-ansible | 08:57 | |
| *** slaweq has quit IRC | 09:06 | |
| *** gillesMo has quit IRC | 09:06 | |
| jrosser | chkumar|ruck: look here http://logs.openstack.org/83/633883/7/check/openstack-ansible-functional-ubuntu-bionic/4e0ee25/logs/ara-report/result/edc7f84c-6850-4245-9069-1d74e68a3bb1/ the info about the router is all in a dict called 'router' | 09:08 |
| *** shardy has joined #openstack-ansible | 09:16 | |
| *** DanyC has joined #openstack-ansible | 09:21 | |
| *** DanyC has quit IRC | 09:25 | |
| *** DanyC has joined #openstack-ansible | 09:27 | |
| openstackgerrit | Martin Kopec proposed openstack/openstack-ansible-os_tempest master: Improve overview subpage https://review.openstack.org/633934 | 09:45 |
| *** gkadam_ has joined #openstack-ansible | 09:55 | |
| *** slaweq has joined #openstack-ansible | 09:56 | |
| *** gkadam has quit IRC | 09:58 | |
| *** chkumar|ruck has quit IRC | 09:58 | |
| *** gkadam__ has joined #openstack-ansible | 10:02 | |
| *** gkadam_ has quit IRC | 10:05 | |
| *** slaweq has quit IRC | 10:05 | |
| *** macza has joined #openstack-ansible | 10:18 | |
| *** macza has quit IRC | 10:22 | |
| *** chandan_kumar has joined #openstack-ansible | 10:24 | |
| *** shardy has quit IRC | 10:24 | |
| *** chandan_kumar is now known as chkumar|ruck | 10:24 | |
| *** shardy has joined #openstack-ansible | 10:26 | |
| *** sdake has joined #openstack-ansible | 10:35 | |
| *** gillesMo has joined #openstack-ansible | 10:38 | |
| *** slaweq has joined #openstack-ansible | 10:39 | |
| *** slaweq has quit IRC | 10:44 | |
| *** sohny has quit IRC | 10:48 | |
| *** sohny has joined #openstack-ansible | 10:49 | |
| *** sdake has quit IRC | 10:51 | |
| *** sohny has quit IRC | 10:53 | |
| *** Darcidride has joined #openstack-ansible | 10:56 | |
| *** Darcidride has quit IRC | 10:56 | |
| *** slaweq has joined #openstack-ansible | 11:05 | |
| *** slaweq has quit IRC | 11:12 | |
| *** udesale has quit IRC | 11:14 | |
| *** asettle has joined #openstack-ansible | 11:16 | |
| *** slaweq has joined #openstack-ansible | 12:11 | |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible-os_ceilometer master: Cleanup files and templates using smart sources https://review.openstack.org/634431 | 12:16 |
| *** Emine has quit IRC | 12:20 | |
| *** Emine has joined #openstack-ansible | 12:21 | |
| *** sohny has joined #openstack-ansible | 12:36 | |
| *** kaiokmo has quit IRC | 12:46 | |
| *** slaweq has quit IRC | 12:50 | |
| *** ansmith has quit IRC | 13:10 | |
| *** kaiokmo has joined #openstack-ansible | 13:20 | |
| *** dave-mccowan has joined #openstack-ansible | 13:31 | |
| *** slaweq has joined #openstack-ansible | 13:35 | |
| CeeMac | afternoon all | 13:43 |
| CeeMac | so, i'm making progress in my openstack rollout, I can deploy an instance from an image, a port is created and dhcp-agent gives it an IP address | 13:44 |
| *** macza has joined #openstack-ansible | 13:44 | |
| CeeMac | however, the ip address is not passed through to the VM | 13:44 |
| CeeMac | tried on windows image from cloudbase.it and from the cirros image created during the deployment (presumably from tempest.yml) | 13:45 |
| CeeMac | I can ping between the compute node and network node on br-vxlan | 13:45 |
| CeeMac | and the physical networking has MTU 9000 set on it | 13:46 |
| CeeMac | just incase the was some packet size issues with the encapsulation overhead | 13:46 |
| CeeMac | any thoughts on what could be causing the disconnect? | 13:46 |
| jamesdenton | does the instance console log reflect dhcp is failing? | 13:48 |
| CeeMac | let me check again, don't recall seeing anything linke that | 13:48 |
| *** zul has joined #openstack-ansible | 13:49 | |
| CeeMac | No lease, failing | 13:51 |
| jamesdenton | k | 13:51 |
| CeeMac | after udhcp (v1.20.1) started | 13:52 |
| CeeMac | on cirros console | 13:52 |
| jamesdenton | can you do a 'openstack network show' on the network and grab the segmentation id? | 13:52 |
| *** priteau has joined #openstack-ansible | 13:53 | |
| CeeMac | sure, 1 sec | 13:53 |
| jamesdenton | then, on the compute and network node, run 'ip -d link show vxlan-XXX' where XXX is that ID. Put the output up on pastebin or something, please | 13:53 |
| CeeMac | where would I see the segmentation id | 13:54 |
| CeeMac | got | 13:54 |
| CeeMac | it | 13:54 |
| CeeMac | being blind | 13:54 |
| CeeMac | http://paste.openstack.org/show/744496/ | 13:56 |
| jamesdenton | Thanks. I just wanted to confirm that both hosts were using the appropriate address for the VTEP, which they are | 13:58 |
| jrosser | it is worth checking that multicast is working properly | 13:59 |
| CeeMac | is there anything required on the underlay to support that jrosser ? | 14:02 |
| jrosser | any BUM (broadcast, unknown, multicast) traffic inside the vxlan tunnel gets turned into multicast traffic on the underlay | 14:04 |
| jamesdenton | yes and no. You can setup a multicast group to do it right, or it may default to broadcast/flood. It's not really spelled out well. | 14:04 |
| *** ansmith has joined #openstack-ansible | 14:04 | |
| jrosser | you can see the group address in your paste 239.1.1.1 | 14:04 |
| jrosser | when i had a very simple setup with just one nexus 9000 it didnt work at all until i configured the igmp querier on the switch | 14:05 |
| CeeMac | i think thats the default from the config? | 14:05 |
| CeeMac | oh man, i've been avoiding multicast on our core for years :( | 14:05 |
| CeeMac | i have the fear | 14:05 |
| jamesdenton | Or you could enable l2pop | 14:05 |
| CeeMac | think l2pop is enabled | 14:05 |
| CeeMac | thats in ml2 config? | 14:06 |
| jamesdenton | it is | 14:06 |
| jamesdenton | there is also an override for it | 14:06 |
| CeeMac | maybe it isn't, 1 sec | 14:07 |
| CeeMac | mechanism_drivers right | 14:08 |
| jamesdenton | neutron_l2_population: bool | 14:08 |
| *** ansmith_ has joined #openstack-ansible | 14:09 | |
| CeeMac | thats the override? | 14:09 |
| jamesdenton | yes | 14:09 |
| CeeMac | in variables file | 14:09 |
| CeeMac | ? | 14:09 |
| jamesdenton | Yes, it defaults to false | 14:09 |
| CeeMac | ok, looks like its not on | 14:09 |
| jamesdenton | You can set it to true, and re-run the playbooks | 14:09 |
| CeeMac | i'd just been reading about it, got confused with real life >_< | 14:10 |
| jamesdenton | it would be easiest to just delete the vxlan networks you have and rebuild them if possible | 14:10 |
| CeeMac | after [playbook run right? | 14:10 |
| CeeMac | oh | 14:11 |
| CeeMac | that reminds me jamesdenton | 14:11 |
| *** ansmith has quit IRC | 14:11 | |
| CeeMac | remember a couple weeks back we were talking about netplan and the veth issue | 14:12 |
| CeeMac | one sece | 14:12 |
| jamesdenton | kinda sorta? | 14:12 |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible-os_ceilometer master: Cleanup files and templates using smart sources https://review.openstack.org/634431 | 14:17 |
| *** gkadam__ has quit IRC | 14:18 | |
| *** dmsimard has quit IRC | 14:18 | |
| *** dmsimard has joined #openstack-ansible | 14:18 | |
| CeeMac | so, the conclusion we came to was to configure the physical interface for tenant networking | 14:21 |
| CeeMac | vlan wasn't being used except on the network node for provider networks so I configured the host_bind_override on the vlan type br-vlan config | 14:22 |
| jamesdenton | ahh ok, yes. Good call | 14:22 |
| CeeMac | the issue with that, is the physical interface on the compute note is different, so linuxbridge-agent wouldn't start | 14:22 |
| CeeMac | i had to manually prune out the config from the compute node | 14:23 |
| CeeMac | running the playbook again will put it back | 14:23 |
| jamesdenton | It is likely possible to set up per-host overrides for those things | 14:23 |
| CeeMac | unless I can come up with a way using netplan to have a "virtual" or specifically named interface in common | 14:23 |
| CeeMac | oh | 14:24 |
| jrosser | jamesdenton: i think this was something i had too - this is now do-able i think | 14:24 |
| CeeMac | is there like a host_vars: element? | 14:24 |
| CeeMac | as opposed to container_vars: | 14:24 |
| CeeMac | for metal roles | 14:24 |
| jamesdenton | it might still be container if it's considered on_metal | 14:25 |
| CeeMac | happy to play guinea pig if it'll get this working | 14:25 |
| jamesdenton | sorry, it may be container_vars with on_metal: true, | 14:25 |
| jamesdenton | https://pasted.tech/pastes/9a4841f34ff2470d25f762277f77d3886f8d3afa | 14:26 |
| jamesdenton | So you're really only concerned about neutron_provider_networks | 14:26 |
| CeeMac | so, the network_mappings? | 14:27 |
| jamesdenton | network_mappings would be something like, "vlan:eth1" in a default deploy | 14:27 |
| jamesdenton | vlan being the name. or physnet1 is better | 14:27 |
| CeeMac | this could work | 14:27 |
| CeeMac | my other thought, was would interface aliasing work on the host level? | 14:27 |
| jamesdenton | you *might* need to also define neutron_ml2_drivers_type, don't know | 14:27 |
| CeeMac | and reference the alias in the host_bind_override | 14:27 |
| jamesdenton | aliasing should be fine, i think | 14:28 |
| *** chandankumar has joined #openstack-ansible | 14:28 | |
| jamesdenton | can you create vlan subinterfaces off an alias? i don't know | 14:28 |
| * CeeMac shrugs | 14:29 | |
| CeeMac | i only just started messing with netplan | 14:29 |
| CeeMac | i was using the veth trick and eth12 like in the examples previously | 14:29 |
| jamesdenton | now that's it's all on_metal, host_bind_override would be the way to go. When the machines deviate from one another, things get tricky | 14:31 |
| jrosser | doesnt this now understand group_binds | 14:31 |
| CeeMac | indeed | 14:32 |
| CeeMac | the neutron-server is still running in a container | 14:33 |
| CeeMac | but the agents are metal | 14:33 |
| CeeMac | jamesdenton, does network_mappings: sync with linuxbridge-agent physical_interface_mappings | 14:35 |
| jamesdenton | >> physical_interface_mappings = {{ neutron_provider_networks.network_mappings }} | 14:36 |
| jamesdenton | so, yes. You're effectively overriding the provider network logic | 14:36 |
| CeeMac | right | 14:37 |
| jamesdenton | that would all be built out automagically via openstack_user_config definitions | 14:37 |
| jamesdenton | jrosser what do you mean by group_binds here, sorry | 14:37 |
| jrosser | it looks like review.o.o isnt working for me otherwise i would find your patch | 14:38 |
| CeeMac | so, the linuxbridge-agent mappings use vlan:<name> apparently here | 14:38 |
| jrosser | but i had the same issue of wanting to have different physical interfaces on computes vs. infra nodes, and i'm sure we made a patch for that | 14:38 |
| *** chkumar|ruck has quit IRC | 14:38 | |
| jamesdenton | jrosser AHH yes, now i know what you mean. one sec | 14:38 |
| jrosser | which i though was what the question was, but i'm now confused about the containers vs. metal angle | 14:39 |
| jamesdenton | CeeMac yes, but the name 'vlan' is quasi-arbitrary | 14:39 |
| CeeMac | as in, i don't need to use it, or i should use it? | 14:39 |
| CeeMac | (trying to keep up, sorry it's monday) | 14:40 |
| jamesdenton | well, in openstack_user_config, when you define a network, the name of that mapping is defined as 'net_name' | 14:40 |
| jamesdenton | and it does default to 'vlan', but it could be 'physnet1' or whatever | 14:40 |
| *** sohny has quit IRC | 14:40 | |
| jamesdenton | so when you override, you just need to be consistent | 14:40 |
| *** chandankumar has quit IRC | 14:41 | |
| jrosser | ^ this is a very good idea, changing the name, otherwise it is horribly confusing | 14:41 |
| *** sohny has joined #openstack-ansible | 14:41 | |
| jamesdenton | jrosser https://review.openstack.org/#/c/592661/ | 14:41 |
| CeeMac | jrosser, sorry, come again? | 14:41 |
| CeeMac | you mean consistency is a good idea too? | 14:42 |
| jrosser | you end up with config files with vlan:vlan in them, and you just go nuts figuring which is the type and which is the name | 14:42 |
| CeeMac | oh | 14:42 |
| CeeMac | got you | 14:42 |
| jrosser | make the name something meaningful to you, like physnet_lab or whatever, then its all totally clear | 14:42 |
| jamesdenton | exactly. | 14:43 |
| CeeMac | where you have an actual interface named "vlan" right? | 14:43 |
| jamesdenton | well the 'vlan' name is meant to be consistent across hosts, while the interface it's mapped to may not be | 14:43 |
| CeeMac | so, the left of the : is a variable, and the right is an actual entity | 14:43 |
| jamesdenton | but 'vlan' doesn't mean anything to anyone. | 14:43 |
| jamesdenton | so we like to use something like... 'exnet:eth1', 'privatenet:eth2', etc. which may reflect different switching layers | 14:44 |
| CeeMac | well. thats not entirely true, but I'll take your point in context to the topic :) | 14:44 |
| jamesdenton | :) | 14:45 |
| CeeMac | so, normally this is more important when generating your provide configures above and beyond the 'standard' config that is deployed right? | 14:45 |
| jamesdenton | well, for multiple provider interfaces, vlan, vlan2, vlan3 may not help :D | 14:45 |
| CeeMac | but because I'm looking at overrides, i need to follow the convention accordingly? | 14:45 |
| *** sohny has quit IRC | 14:45 | |
| jamesdenton | well, one sec | 14:45 |
| CeeMac | is there a way to *not* apply br-blan config to the compute node? I don't think it will ever need it | 14:46 |
| CeeMac | north-south traffic will always go via the network node right? | 14:46 |
| jamesdenton | jrosser that patch - i should've documented the use case a little better, but i believe it was in the context of neutron group names more so than other groups. Like, neutron_linuxbridge_agent vs neutron_openvswitch_agent and not some arbitrary groupings | 14:46 |
| CeeMac | I'm using vxlan exclusively for tenant east-west traffic | 14:47 |
| CeeMac | ack, *br-vlan | 14:47 |
| *** jenglisch_ is now known as jenglisch | 14:48 | |
| jamesdenton | Well, you could probably define the overrides on the network nodes vs the compute nodes, and just let them use br-vlan in the mapping (i.e. don't set host_bind_override)_ | 14:48 |
| jamesdenton | the mapping will still be set, but not used | 14:48 |
| CeeMac | wait, i'm confused again | 14:49 |
| CeeMac | so, set the network_mappings on the network nodes, but not the compute nodes | 14:49 |
| CeeMac | and don't put host_bind_override in the br-vlan provider network "vlan" type? | 14:50 |
| *** zul has quit IRC | 14:50 | |
| jamesdenton | right - if you don't set host_bind_override, it will use the the value of 'container_bridge' | 14:51 |
| jamesdenton | which is br-vlan. | 14:51 |
| jamesdenton | given that you only care about the network nodes (running agents on baremetal), you could set those 'container_vars' overrides | 14:52 |
| jamesdenton | and the computes will get whatever is automatically generated based on openstack_user_config | 14:52 |
| CeeMac | but i would still need br-vlan configured on the compute node for the agent to start right? | 14:52 |
| CeeMac | i ran into something like this already, which is why i went down the veth/eth12 route. | 14:53 |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible-os_ceilometer master: Cleanup files and templates using smart sources https://review.openstack.org/634431 | 14:54 |
| jamesdenton | right. trying something, brb | 14:55 |
| CeeMac | k | 14:55 |
| jamesdenton | CeeMac what version are you on? | 14:58 |
| CeeMac | of | 14:58 |
| CeeMac | openstack in general | 14:58 |
| CeeMac | easy way to check? | 15:00 |
| jamesdenton | cat /etc/openstack-release | 15:01 |
| CeeMac | 18.1.3 | 15:01 |
| jamesdenton | ok cool | 15:03 |
| jamesdenton | ok, so take a look at this: https://pasted.tech/pastes/be43604d0e0b0ac2d247122c1f95cb7053649c66 | 15:03 |
| jamesdenton | there's two definitions - one for compute and one for network. everything is the same except the interface for host_bind_override | 15:03 |
| CeeMac | ah, ok | 15:03 |
| jamesdenton | it may be possible to just define the one for network_hosts, and not the other | 15:04 |
| CeeMac | i'm not adverse to doing both | 15:04 |
| jamesdenton | these would replace any existing "vlan" definition in openstack_user_config | 15:04 |
| jamesdenton | but vxlan would remain | 15:04 |
| CeeMac | sounds ideal | 15:04 |
| jamesdenton | we're basically just changing the group the network gets applied to | 15:04 |
| jamesdenton | you could also, theoretically, define a custom group. but these two should provide the separation | 15:05 |
| CeeMac | no. this looks great mate, i'll give it a go just now. If this works I'll be so happy | 15:05 |
| jamesdenton | you might start with only the network_hosts | 15:05 |
| jamesdenton | and then we can try adding compute | 15:05 |
| CeeMac | (until I find the next thing that isn't working) | 15:05 |
| CeeMac | sure | 15:06 |
| jamesdenton | you also will not need the container_vars overrides we talked about | 15:06 |
| jamesdenton | this patch was intended for another use case, but should work here. If so, we'll update the docs to highlight that | 15:06 |
| CeeMac | so, instead of using the cat cat /etc/openstack-release group_binds? | 15:06 |
| CeeMac | oh ffs, my copy/paste is rubbish today | 15:06 |
| CeeMac | neutron_linuxbridge_agent | 15:07 |
| CeeMac | take ^ out and put network_hosts in? | 15:07 |
| jamesdenton | yes, that's correct | 15:07 |
| CeeMac | awesome | 15:07 |
| CeeMac | i'll go do that | 15:07 |
| CeeMac | thanks a lot | 15:08 |
| jamesdenton | and you would set the host_bind_override to whatever is appropriate for those hosts. If it varies between those hosts, then we can go from there | 15:08 |
| jamesdenton | remember - leave the vxlan provider block as linuxbridge_agent | 15:08 |
| CeeMac | got you | 15:08 |
| CeeMac | so i'll leave the host_bind_override as is, it was already set correctly for the interface names on the network nodes | 15:09 |
| CeeMac | and change the group_binds | 15:09 |
| jamesdenton | yep | 15:09 |
| CeeMac | add in the l2pop override in variables | 15:09 |
| jamesdenton | yeah, for now | 15:09 |
| *** cmart has joined #openstack-ansible | 15:10 | |
| CeeMac | and re-run the neutron playbook | 15:10 |
| jamesdenton | right | 15:10 |
| CeeMac | cool | 15:10 |
| *** Emine has quit IRC | 15:11 | |
| *** sdake has joined #openstack-ansible | 15:12 | |
| CeeMac | bbiab | 15:12 |
| CeeMac | oh, and i need to add br-vlan config to netplan on compute node | 15:13 |
| CeeMac | this wont have changed anything that would affect haproxy right | 15:19 |
| CeeMac | i dont need to re-run that | 15:19 |
| *** Emine has joined #openstack-ansible | 15:19 | |
| jamesdenton | you shouldn't need a br-vlan on computes | 15:19 |
| jamesdenton | at least one that's plumbed to anything | 15:19 |
| jamesdenton | no changes to haproxy | 15:20 |
| CeeMac | ive set up the configuration for it, but left interfaces: [] blank | 15:20 |
| CeeMac | so its there, but not bound | 15:20 |
| CeeMac | so when the linuxbridge-agent looks for it and its not there it doesn't freak out | 15:20 |
| jamesdenton | it won't look for it, if all goes well | 15:20 |
| CeeMac | *fingers crossed* | 15:21 |
| jamesdenton | You should end up with: Interface mappings: {} | 15:21 |
| CeeMac | in the journal after its started? | 15:21 |
| *** cmart has quit IRC | 15:23 | |
| jamesdenton | yes | 15:24 |
| jamesdenton | it's an INFO message | 15:24 |
| CeeMac | yup | 15:25 |
| CeeMac | got it | 15:25 |
| CeeMac | and the config file has no entries for it so looking good | 15:25 |
| CeeMac | so, i need to shut down instances, delete and re-create network, re-attach instances to new network, power on and cross fingers? | 15:26 |
| CeeMac | rebuild router too? | 15:26 |
| jamesdenton | for the vxlan piece? | 15:29 |
| jamesdenton | yeah, that's one approach | 15:30 |
| jamesdenton | or delete the vxlan-68 interfaces on all of the hosts and restart neutron-linuxbridge-agent | 15:30 |
| CeeMac | oh | 15:31 |
| CeeMac | that would be more straight-forward | 15:31 |
| CeeMac | that should only be on the compute and network nodes right? | 15:31 |
| jamesdenton | yes | 15:31 |
| CeeMac | ok | 15:35 |
| CeeMac | done | 15:35 |
| CeeMac | restart instance | 15:35 |
| CeeMac | ? | 15:35 |
| jamesdenton | yes | 15:36 |
| CeeMac | ooooooo | 15:36 |
| CeeMac | cirros has an IP :D | 15:37 |
| jamesdenton | woot | 15:37 |
| CeeMac | let me check win | 15:37 |
| *** nurdie has joined #openstack-ansible | 15:39 | |
| CeeMac | legend | 15:39 |
| * CeeMac high-fives jamesdenton | 15:39 | |
| jamesdenton | woohoo | 15:39 |
| jamesdenton | glad it worked | 15:39 |
| CeeMac | mate, you're a life-saver | 15:39 |
| jamesdenton | hah, thanks | 15:40 |
| CeeMac | need to work on my external routing but this is a huge step forward | 15:40 |
| CeeMac | i was getting a lot of pressure from the big boss, lol | 15:40 |
| jamesdenton | if/when you have time, it would be great if you could fill out a bug for the docs and maybe describe your use-case and provide that override | 15:40 |
| jamesdenton | or rather, provider configuration | 15:40 |
| *** slaweq has quit IRC | 15:40 | |
| CeeMac | sure | 15:41 |
| jamesdenton | maybe we can get some docs updated for the next release to help break down different use-cases, including yours, and make deployments easier | 15:41 |
| CeeMac | that would be great | 15:41 |
| CeeMac | although i understand its difficult to document for every scenario | 15:42 |
| jamesdenton | well, there are some pretty common ones that i think we can tackle. and the group_binds ability makes that a lot easier now | 15:42 |
| jamesdenton | you verifying that for us is great, thanks | 15:42 |
| CeeMac | i think its more specifically an issue relate to the foibles of netplan than anything else | 15:42 |
| CeeMac | no problem | 15:42 |
| jamesdenton | well, yes and no. netplan doesn't make it easier, but having unnecessary bridges and veths isn't cool, either | 15:43 |
| CeeMac | as binding a veth pair to eth12 in ifupdown worked fine for that scenario | 15:43 |
| CeeMac | true | 15:43 |
| CeeMac | and this works in both scenarios and is quite graceful from a config perspective | 15:43 |
| CeeMac | do you want/need me to test what happens if I throw in the compute_host provider config too? | 15:44 |
| jamesdenton | if you'd like to, sure. you should see the respective interface mapping show up there and it would give you the ability to spin up VMs on the vlan | 15:44 |
| jamesdenton | otherwise, you're limited to vxlan | 15:44 |
| cloudnull | mornings | 15:46 |
| jamesdenton | well hello | 15:47 |
| CeeMac | hey cloudnull | 15:52 |
| cloudnull | o/ | 15:52 |
| *** spatel has joined #openstack-ansible | 15:52 | |
| *** sdake has quit IRC | 16:15 | |
| *** cmart has joined #openstack-ansible | 16:15 | |
| jrosser | cloudnull: i had another local go at the nova smart-sources patch | 16:16 |
| spatel | influxdb vs graphite what do you guys prefer ? | 16:16 |
| cloudnull | oh? | 16:16 |
| jrosser | behaves differently again :/ http://paste.openstack.org/show/744508/ | 16:16 |
| jrosser | curse of pip this time i think | 16:16 |
| cloudnull | :'( | 16:16 |
| jrosser | thats installing neutron, TASK [python_venv_build : Create the virtualenv (if it does not exist)] | 16:17 |
| jrosser | i wonder if theres a tiny difference there in the neutron playbook, as the other roles no doubt work the same | 16:18 |
| cloudnull | I've seen that | 16:18 |
| cloudnull | you have to add --no-setuptools then install setuptools | 16:18 |
| jrosser | you mean we've fixed that elsewhere? | 16:18 |
| cloudnull | to the virtaulenv create | 16:18 |
| cloudnull | seems to be something to do with python_venv_build process? | 16:19 |
| cloudnull | not really sure why thats messed up | 16:19 |
| *** vollman has joined #openstack-ansible | 16:19 | |
| jrosser | here is the tasks prior to that http://paste.openstack.org/show/744509/ | 16:20 |
| cloudnull | https://github.com/openstack/openstack-ansible-ops/blob/master/skydive/roles/skydive_common/tasks/skydive_keystone.yml#L31-L55 | 16:20 |
| cloudnull | ^ that's how I've addressed that before | 16:20 |
| *** schwicht has joined #openstack-ansible | 16:23 | |
| *** gyee has joined #openstack-ansible | 16:23 | |
| *** sdake has joined #openstack-ansible | 16:25 | |
| *** sdake has quit IRC | 16:38 | |
| *** TxGirlGeek has joined #openstack-ansible | 16:39 | |
| CeeMac | jamesdenton, that has worked as expected, compute node has picked up interface mappings specific to its group binding | 16:42 |
| jamesdenton | beautiful | 16:42 |
| CeeMac | i'll file a bug in the morning as discussed :) | 16:43 |
| jamesdenton | appreciate that | 16:43 |
| CeeMac | the least i can do :) | 16:43 |
| CeeMac | i'm out. evening all | 16:45 |
| jamesdenton | see ya! | 16:45 |
| *** priteau has quit IRC | 16:53 | |
| *** priteau has joined #openstack-ansible | 16:54 | |
| *** priteau has quit IRC | 16:54 | |
| *** sdake has joined #openstack-ansible | 16:57 | |
| *** priteau has joined #openstack-ansible | 17:04 | |
| *** sdake has quit IRC | 17:06 | |
| *** Emine has quit IRC | 17:18 | |
| *** sdake has joined #openstack-ansible | 17:22 | |
| *** Emine has joined #openstack-ansible | 17:30 | |
| *** cmart has quit IRC | 17:30 | |
| *** cmart has joined #openstack-ansible | 17:36 | |
| *** cmart has quit IRC | 17:49 | |
| *** cmart has joined #openstack-ansible | 17:50 | |
| noonedeadpunk | cores, can you please check this patch? https://review.openstack.org/#/c/630747/ | 17:51 |
| *** gillesMo has quit IRC | 17:51 | |
| noonedeadpunk | cloudnull: do you have a minute? | 17:57 |
| cloudnull | sure | 17:57 |
| cloudnull | whats up ? | 17:57 |
| noonedeadpunk | Why do we need this replace? https://github.com/openstack/openstack-ansible-os_cinder/blob/master/tasks/cinder_pre_install.yml#L80 | 17:58 |
| noonedeadpunk | Isn't it better to symlink from root? | 17:58 |
| *** kopecmartin is now known as kopecmartin|off | 17:59 | |
| *** chandankumar has joined #openstack-ansible | 17:59 | |
| noonedeadpunk | And do you mind, that I've taken git message and your notes for ceilometer https://review.openstack.org/#/c/634431/6 ? | 18:00 |
| *** pcaruana has quit IRC | 18:00 | |
| noonedeadpunk | Probably I should add co-authored or smth like that - I'm not sure how to deal with such things correctly | 18:01 |
| noonedeadpunk | Yeah, I see your note, but not fully understand, in what cases we might be facing the problem | 18:03 |
| noonedeadpunk | probably, I'm not fully understand some issues related to connection plugin... | 18:04 |
| openstackgerrit | Merged openstack/openstack-ansible-os_tempest master: Add tripleo-ci-centos-7-standalone-os-tempest job https://review.openstack.org/633931 | 18:08 |
| *** macza has quit IRC | 18:13 | |
| *** DanyC has quit IRC | 18:17 | |
| *** DanyC has joined #openstack-ansible | 18:18 | |
| *** chandankumar has quit IRC | 18:19 | |
| *** DanyC has quit IRC | 18:22 | |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible-os_ceilometer master: Cleanup files and templates using smart sources https://review.openstack.org/634431 | 18:22 |
| noonedeadpunk | cloudnull: anyway I'd like you to take a look at ^ as one of the evangelist of smart-resources | 18:22 |
| cloudnull | noonedeadpunk that symlink it going back to the /openstack/venv/$NAME/etc directory, it would be better if we link from root however from within the container that's not entirely possible . | 18:24 |
| cloudnull | the link goes back to the system root, not the container root | 18:24 |
| noonedeadpunk | and is there any point in leaving .original configs, except of the terms of "backup"? | 18:26 |
| cloudnull | noonedeadpunk sorry , my bouncer was needing to be bounced | 18:28 |
| cloudnull | back | 18:28 |
| cloudnull | as for taking the commit message and doing adding the smart sources bits to other roles. feel free! | 18:29 |
| cloudnull | I greatly appreciate the help | 18:29 |
| cloudnull | re: the original files. They're just backups when found and will be used to retemplate those files as needed. ' | 18:30 |
| noonedeadpunk | oh, I've just understood the point of using cp:) | 18:31 |
| noonedeadpunk | thanks for your answers! | 18:31 |
| cloudnull | anytime | 18:31 |
| noonedeadpunk | and for your time | 18:31 |
| cloudnull | sorry for teh slow reply | 18:31 |
| *** DanyC has joined #openstack-ansible | 18:36 | |
| logan- | prometheanfire / jamesdenton: merged the etcd template fix and tagged 2.1 on the role. thanks for the feedback on it | 18:41 |
| *** DanyC has quit IRC | 18:41 | |
| prometheanfire | logan-: cool :D | 18:41 |
| *** sreejithp has joined #openstack-ansible | 18:49 | |
| *** sreejithp has quit IRC | 18:49 | |
| jamesdenton | sweet | 18:57 |
| prometheanfire | logan-: is the etcd stuff considered 'stable' and consumable? (if so, for what versions of osa) | 19:00 |
| prometheanfire | and should the sha be updated in osa for stable branches to your update | 19:00 |
| *** shardy has quit IRC | 19:01 | |
| logan- | prometheanfire: i have been using the 1.x tags to deploy etcd2 for years to use with calico. the role's 2.x tags are a rework to start deploying etcd v3. afaik it is stable now based on jamesdenton's feedback last week and the person who commented in the PR saying it works. I haven't used the 2.x tags in prod yet because we have not yet moved to a calico version that uses etcd3 yet. | 19:04 |
| prometheanfire | ok | 19:04 |
| jamesdenton | haven't hit it hard, but it seems to work so far | 19:06 |
| jrosser | I wonder how we should handle the etcd2/3 thing, because you’d want to switch up which you got depending on calico/vpp | 19:08 |
| jrosser | Perhaps making explicitly etcd3 container group would be smart, then you don’t break things | 19:08 |
| logan- | jrosser: yeah. I have the calico upgrade that uses etcd3 almost gating here | 19:09 |
| logan- | i have not had luck using the older etcd2 calico with queens+ | 19:09 |
| prometheanfire | could run both and have a v2 and v3 endpoint | 19:09 |
| prometheanfire | rely on apps to migrate (or do so when we feel like it) | 19:09 |
| jrosser | We should check what version is needed for a coordinator service, that’s needed for gnocchi and designate | 19:09 |
| logan- | yeah I think v3 is supposed to offer a v2 compatible endpoint also | 19:09 |
| prometheanfire | that'd be good | 19:10 |
| logan- | don't quote me on that | 19:10 |
| prometheanfire | 13:11 < logan- > yeah I think v3 is supposed to offer a v2 compatible endpoint also | 19:10 |
| prometheanfire | gotit | 19:10 |
| logan- | damn it | 19:10 |
| logan- | lol | 19:10 |
| *** Emine has quit IRC | 19:22 | |
| *** macza has joined #openstack-ansible | 19:24 | |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible-os_ceilometer master: Cleanup files and templates using smart sources https://review.openstack.org/634431 | 19:26 |
| *** TxGirlGeek has quit IRC | 20:01 | |
| *** cmart has quit IRC | 20:03 | |
| *** TxGirlGeek has joined #openstack-ansible | 20:08 | |
| openstackgerrit | Michael Vollman proposed openstack/openstack-ansible-os_manila master: Basic working os_manila role https://review.openstack.org/611930 | 20:17 |
| *** cmart has joined #openstack-ansible | 20:20 | |
| *** hamzaachi has joined #openstack-ansible | 20:27 | |
| *** DanyC has joined #openstack-ansible | 20:32 | |
| *** cmart has quit IRC | 20:39 | |
| *** cmart has joined #openstack-ansible | 20:40 | |
| *** sohny has joined #openstack-ansible | 20:42 | |
| *** sohny has quit IRC | 20:46 | |
| *** cmart has quit IRC | 20:50 | |
| *** slaweq has joined #openstack-ansible | 21:00 | |
| *** hamzaachi has quit IRC | 21:01 | |
| *** hamzaachi has joined #openstack-ansible | 21:01 | |
| *** aedc has quit IRC | 21:11 | |
| *** cmart has joined #openstack-ansible | 21:34 | |
| *** macza has quit IRC | 21:44 | |
| *** macza has joined #openstack-ansible | 21:44 | |
| *** ansmith_ has quit IRC | 22:06 | |
| *** spatel has quit IRC | 22:19 | |
| *** priteau has quit IRC | 22:20 | |
| *** DanyC has quit IRC | 22:27 | |
| *** slaweq has quit IRC | 22:36 | |
| *** nurdie has quit IRC | 22:36 | |
| *** nurdie has joined #openstack-ansible | 22:37 | |
| *** nurdie has quit IRC | 22:41 | |
| *** archklikk has joined #openstack-ansible | 22:44 | |
| *** archklikk has quit IRC | 22:45 | |
| *** archklikk has joined #openstack-ansible | 22:47 | |
| *** archklikk_ has joined #openstack-ansible | 22:50 | |
| *** archklikk has quit IRC | 22:53 | |
| *** archklikk_ has quit IRC | 22:55 | |
| *** schwicht has quit IRC | 22:56 | |
| *** hwoarang has quit IRC | 23:07 | |
| *** hwoarang has joined #openstack-ansible | 23:10 | |
| cloudnull | afternoons | 23:15 |
| *** DanyC has joined #openstack-ansible | 23:17 | |
| *** DanyC has quit IRC | 23:21 | |
| *** schwicht has joined #openstack-ansible | 23:23 | |
| *** sdake has quit IRC | 23:28 | |
| *** sdake has joined #openstack-ansible | 23:29 | |
| *** markvoelker has quit IRC | 23:31 | |
| *** hwoarang has quit IRC | 23:43 | |
| *** hwoarang has joined #openstack-ansible | 23:43 | |
| *** nurdie has joined #openstack-ansible | 23:45 | |
| *** schwicht has quit IRC | 23:49 | |
| *** nurdie has quit IRC | 23:49 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!