| *** macza has quit IRC | 00:01 | |
| *** cshen has joined #openstack-ansible | 00:09 | |
| *** hwoarang has quit IRC | 00:13 | |
| *** hwoarang has joined #openstack-ansible | 00:13 | |
| *** cshen has quit IRC | 00:15 | |
| *** mathlin has joined #openstack-ansible | 00:16 | |
| *** tosky has quit IRC | 00:21 | |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible-os_keystone master: Cleanup files and templates using smart sources https://review.openstack.org/588960 | 00:30 |
|---|---|---|
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible-os_cinder master: Cleanup files and templates using smart sources https://review.openstack.org/588953 | 00:40 |
| *** cshen has joined #openstack-ansible | 00:45 | |
| *** mathlin has quit IRC | 00:50 | |
| *** cshen has quit IRC | 00:51 | |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible-os_glance master: Cleanup files and templates using smart sources https://review.openstack.org/588959 | 00:56 |
| *** mathlin has joined #openstack-ansible | 01:13 | |
| *** macza has joined #openstack-ansible | 01:38 | |
| *** markvoelker has joined #openstack-ansible | 01:40 | |
| *** markvoelker has quit IRC | 01:45 | |
| *** mathlin has quit IRC | 01:47 | |
| *** gyee has quit IRC | 01:50 | |
| *** macza has quit IRC | 01:50 | |
| *** cshen has joined #openstack-ansible | 02:06 | |
| *** mathlin has joined #openstack-ansible | 02:10 | |
| *** hwoarang has quit IRC | 02:11 | |
| *** cshen has quit IRC | 02:11 | |
| *** hwoarang has joined #openstack-ansible | 02:11 | |
| *** macza has joined #openstack-ansible | 02:17 | |
| *** hwoarang has quit IRC | 02:19 | |
| partlycloudy | A quick question regarding the OSA documentation (Rocky) at https://docs.openstack.org/openstack-ansible/rocky/user/prod/example.html. | 02:20 |
| partlycloudy | The figure shows two dedicated load balancers in use, but the configuration file sets HAProxy onto three infra nodes. Is that a typo? | 02:20 |
| *** macza has quit IRC | 02:22 | |
| *** hwoarang has joined #openstack-ansible | 02:23 | |
| jamesdenton | partlycloudy the image is more representative of hardware load balancers, i think, where there tend to be a active/standby pair | 02:36 |
| cloudnull | sadly looks like the centos percona repo is down ... http://logs.openstack.org/60/588960/13/check/openstack-ansible-functional-distro_install-centos-7/858b5b0/job-output.txt.gz#_2019-01-10_00_59_52_070279 | 02:38 |
| *** broken_one has joined #openstack-ansible | 02:38 | |
| partlycloudy | jamesdenton: thanks you James. So, the figure on the page does not represent the HAProxy settings in the openstack_user_config.yml below, I think? | 02:41 |
| *** cshen has joined #openstack-ansible | 02:41 | |
| jamesdenton | that's correct. For the example file, shove a red load balancer into the infrastructure control plane host box | 02:42 |
| jamesdenton | times 3 | 02:42 |
| jamesdenton | cloudnull you should write a strongly worded letter | 02:42 |
| cloudnull | ? | 02:43 |
| jamesdenton | percona. repo. n/m | 02:43 |
| * cloudnull begins monologue of why i hate centos | 02:43 | |
| jamesdenton | but it's Enterprise, sir. | 02:43 |
| *** mathlin has quit IRC | 02:44 | |
| partlycloudy | jamesdenton: got it. Btw, I have thoroughly enjoyed reading your latest book on Neutron networking (the 3rd edition). Great work! | 02:44 |
| jamesdenton | thanks, partlycloudy! that's a great way to get help around here! lol | 02:45 |
| partlycloudy | oh yeah! haha | 02:46 |
| *** cshen has quit IRC | 02:46 | |
| broken_one | good evening everyone | 02:46 |
| cloudnull | o/ | 02:47 |
| broken_one | i am having problems dialing in my osa install. I cant tell if i am just not configuring something right or if there is some other issue, this does center around networking and neutron though. Is it ok to ask for support in this channel? | 02:49 |
| jamesdenton | we'll do our best. whats up? | 02:50 |
| broken_one | on the upside I can log into horizon and create networks and routers. but I think my vxlan and ha configs may be not quite right | 02:50 |
| *** markvoelker has joined #openstack-ansible | 02:51 | |
| broken_one | so i switch from using linux bridges over to OpenVSwitch | 02:51 |
| jamesdenton | on a new deploy? | 02:51 |
| broken_one | yes, i wiped all the hosts and started with a fresh OS | 02:51 |
| jamesdenton | ok, good. still having issues, though? | 02:52 |
| broken_one | so a deploy node, a cobbler node, 3 controller nodes, and 4 compute nodes | 02:52 |
| broken_one | so i have 4 nics on the systems 2 x 1gb and 2x10Gbe | 02:53 |
| broken_one | they are bonded | 02:53 |
| broken_one | br-vlan is on bond1 | 02:53 |
| broken_one | br-vxlan in on bond1.1003 | 02:54 |
| jamesdenton | k | 02:54 |
| broken_one | br-vlan does not have an ip, and br-vxlan does | 02:54 |
| jamesdenton | right. | 02:54 |
| jamesdenton | what are you seeing now that you're on OVS? | 02:54 |
| broken_one | the ultimate issue right now | 02:55 |
| broken_one | i create an HA router | 02:55 |
| broken_one | but all 3 interfaces are DOWN | 02:55 |
| broken_one | in horizon | 02:55 |
| broken_one | so before with linuxbridges, i only tested a non-HA router | 02:56 |
| jamesdenton | have you been able to successfully test a single VM? Does DHCP work? Can you test a standalone router? | 02:56 |
| broken_one | so the configs could have been broken for HA in linux bridges | 02:56 |
| broken_one | DHCP is working | 02:56 |
| broken_one | when i deploy a test subnet all 3 dhcp agents start up with an IP | 02:57 |
| jamesdenton | ok. Does a spun up VM get its IP from the DHCP server? | 02:57 |
| jamesdenton | when you check console log | 02:57 |
| broken_one | yes the cirros image gets an IP, recently deployed 5 vm's and all 5 got an IP form the test subnet | 02:58 |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible-os_glance master: Cleanup files and templates using smart sources https://review.openstack.org/588959 | 02:58 |
| broken_one | another thing I feel is happening is that all traffic is being sent through the br-mgmt interface and nothing over br-vxlan | 02:58 |
| jamesdenton | well, what is the network type for the network your VM is on? is it vxlan or vlan or flat? | 03:00 |
| broken_one | vxlan | 03:00 |
| jamesdenton | ok. you can confirm the IP that Neutron is using for the VTEP by looking at... the openvswitch_agent.ini file and see what local_ip is set to. It should match the addr on br-vxlan | 03:01 |
| jamesdenton | what version of OSA is this? | 03:01 |
| broken_one | its rocky so 18.1.2 | 03:01 |
| jamesdenton | k | 03:01 |
| broken_one | so i did recently change the local_ip on the controller nodes to match the br-vxlan ip | 03:02 |
| broken_one | and now things are starting to report down | 03:02 |
| jamesdenton | what was local_ip set to before? | 03:02 |
| broken_one | when i did the deploy originally it was the br-mgmt ip and everything was UP | 03:02 |
| broken_one | hence why i feel something could be wrong in my configs | 03:03 |
| broken_one | i will also mention | 03:03 |
| jamesdenton | hmmm. yeah, that wouldn't have been correct to use br-mgmt. Are you able to ping between your infras and computes via their br-vxlan IPs? | 03:03 |
| jamesdenton | and can you share your openstack_user_config.yml via https://pasted.tech/ | 03:04 |
| broken_one | to get the local_ip to change to the br-vxlan i had to heavily modify the dynamic-address-facts.yml playbook because the find_bridge variable was coming up undefined when doing the tunnel_address fact set | 03:04 |
| broken_one | certainly | 03:04 |
| broken_one | and yes we check all pings | 03:05 |
| jamesdenton | yeah, you shouldn't have to do that. maybe we can find something missing | 03:05 |
| broken_one | no issues | 03:05 |
| broken_one | ok give me a minute...need to hop on vpn :D | 03:05 |
| broken_one | so i might disconnect | 03:05 |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible-os_cinder master: Cleanup files and templates using smart sources https://review.openstack.org/588953 | 03:06 |
| *** broken_one has quit IRC | 03:06 | |
| *** mathlin has joined #openstack-ansible | 03:07 | |
| *** broken_one has joined #openstack-ansible | 03:08 | |
| broken_one | sorry about that | 03:08 |
| jamesdenton | no prob | 03:08 |
| broken_one | https://pasted.tech/pastes/67f1ba9703482ddd185e059920f93b715b156400 -- config file | 03:09 |
| jamesdenton | looking | 03:10 |
| broken_one | thankl you | 03:11 |
| jamesdenton | can you also share your openstack_inventory.json? | 03:15 |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible-os_heat master: Cleanup files and templates using smart sources https://review.openstack.org/588954 | 03:15 |
| broken_one | no problem | 03:16 |
| broken_one | https://pasted.tech/pastes/b7d2955a191a976cb5386bacc9bf545759c05264 -- inventory.json | 03:20 |
| jamesdenton | thanks | 03:23 |
| *** broken_one has quit IRC | 03:24 | |
| *** broken_one has joined #openstack-ansible | 03:24 | |
| broken_one | thank you for taking the time | 03:25 |
| cloudnull | nice use of pasted :) | 03:25 |
| jamesdenton | cloudnull do you have an older environment around.. like P or Q? | 03:26 |
| cloudnull | nope, my home cloud is running the head of master. | 03:27 |
| jamesdenton | k | 03:27 |
| cloudnull | i could make one though | 03:27 |
| cloudnull | need to test something ? | 03:27 |
| jamesdenton | eh, i was wanting to see if ./inventory-manage.py --list-host would return the tunnel_address for respective containers | 03:28 |
| cloudnull | not a lot of changes happen to that script | 03:29 |
| jamesdenton | ya | 03:29 |
| cloudnull | the output should be the same regardless | 03:29 |
| jamesdenton | looking at an old pike inventory, the only thing that had a legit tunnel_address was neutron_agents_container. the baremetal compute node has an entry but it's missing the IP | 03:32 |
| broken_one | if that command is there in rocky i could run it on my cloud | 03:32 |
| jamesdenton | long story short, it looks like ansible_host is being used in the absence of a tunnel address, which would potentially not work. | 03:34 |
| broken_one | yes that is what is happening | 03:34 |
| broken_one | just not sure what is causing it | 03:34 |
| broken_one | the above command lists None in the tunnel_address column on my cloud | 03:35 |
| jamesdenton | yeah, unfortunately that's not a good indicator. but i think i see the same behavior | 03:36 |
| broken_one | would also having the user_variables.yml be helpful? | 03:37 |
| jamesdenton | wouldn't hurt | 03:38 |
| broken_one | ok one moment | 03:38 |
| *** mathlin has quit IRC | 03:41 | |
| broken_one | https://pasted.tech/pastes/14493ca75cebc62891551a21a3ed62166e8a0260 -- user_vars | 03:42 |
| cloudnull | `pip install pasted-client --isolated; pasted /etc/openstack_deploy/user_variables.yml` | 03:42 |
| broken_one | had to remove ldap things but rest is complete | 03:42 |
| * cloudnull shameless plug | 03:42 | |
| jamesdenton | niiiiice | 03:42 |
| broken_one | your code cloudnull? | 03:43 |
| cloudnull | yup | 03:43 |
| cloudnull | pasted is me :) | 03:43 |
| jamesdenton | cloudnull do you know.. if there is or was a mechanism to grab the IP address(es) off a host's bridge (the br-vxlan IP, br-storage IP, etc) in the event it wasn't auto-assigned by ansible? | 03:44 |
| broken_one | very nice...i see we can spin up pasted on our own infra...we do have a need for a pastebin | 03:44 |
| cloudnull | the python client will always return the RAW paste too, so its nice when sharing things or wanting to do things with the pastes, IE: something like a git-diff | 03:44 |
| broken_one | a nice "feature" would be a preview mode...in case something was pasted with credentials or confidential stuffs...before a paste | 03:45 |
| broken_one | maybe even an expiry on some pastes if possible | 03:46 |
| cloudnull | will look into it | 03:46 |
| cloudnull | jamesdenton nothing built in | 03:46 |
| cloudnull | you could ansible -m shell it ? | 03:46 |
| broken_one | jamesdenton is the preferred method to leave the interfaces unassigned and have one of the playbooks do magic? | 03:47 |
| jamesdenton | hmm. Well, I'm just trying to figure out where we'd be pulling the respective tunnel address from for say, a compute node | 03:47 |
| cloudnull | the inventory will have all of the IPs but if you want to check that the IPs on the host are the same in the inventory there's nothing that specifically scans and validates that | 03:47 |
| jamesdenton | since those interfaces are hand created in e/n/i or whereveer | 03:47 |
| jamesdenton | ok, i see it's collected as a fact | 03:49 |
| cloudnull | the tunnel addresses are scanned for and fed into the role | 03:49 |
| cloudnull | in the newton playbook , if im not mistaken | 03:49 |
| cloudnull | broken_one the pastes will auto expire after 90 days of no access | 03:49 |
| cloudnull | though a configurable expire would be nice | 03:49 |
| broken_one | i found some task in dynamic-address-fact.yml in common-tasks/ | 03:50 |
| cloudnull | ^ that one | 03:50 |
| broken_one | that seemed to do some tunnel_address fact set, but it is not working for me | 03:50 |
| broken_one | find_bridge return undefined a lot in the metal_query | 03:51 |
| broken_one | that is the yml i modified to flip to local_ip in the openvswitch.ini file | 03:52 |
| *** markvoelker has quit IRC | 03:53 | |
| *** DanyC has joined #openstack-ansible | 04:01 | |
| *** cshen has joined #openstack-ansible | 04:02 | |
| *** DanyC has quit IRC | 04:06 | |
| *** cshen has quit IRC | 04:13 | |
| jamesdenton | if i'm understanding this correctly, the addrs will never be in the inventory and dynamically determined at runtime playbooks/common-playbooks/neutron.yml based on the facts. | 04:14 |
| jamesdenton | sorry, i'm dense | 04:14 |
| broken_one | that is the behaviour i am witnessing, just not sure if it is correct | 04:15 |
| broken_one | maybe there is something that i need to set in host_vars for each of the nodes? | 04:16 |
| jamesdenton | so, broken_one - i've got to hang this one up for tonight, but given that i'm able to replicate the behavior i'll dig into more tomorrow morning. i have a multi-node install going right now, so that will help confirm | 04:16 |
| broken_one | so they are added to facts | 04:16 |
| broken_one | or inventory | 04:16 |
| jamesdenton | there's nothing you should have to do. it should behave. if it's not, then there's a bug somewhere | 04:17 |
| *** partlycloudy has quit IRC | 04:17 | |
| *** hwoarang has quit IRC | 04:17 | |
| broken_one | sure thing jamesdenton, i was about to spin up a vm to idle here and will check back tomorrow after work, unless i can get around some port blocks at the office :D | 04:17 |
| *** partlycloudy has joined #openstack-ansible | 04:17 | |
| broken_one | anything i can do to help run down the bug let me know | 04:17 |
| jamesdenton | i'm about 30% thru an MNAIO for something else, but it will be good to validate with tomorrow. | 04:18 |
| broken_one | and ill check back in with yall tomorrow as well | 04:18 |
| jamesdenton | thanks | 04:18 |
| broken_one | AIO installs may work just cuz... testing with lots of bare metal nodes seem to bring out the issue | 04:18 |
| broken_one | thank you again for looking into this. | 04:19 |
| jamesdenton | yeah, it's worked on production deploys for a long time. I just don't use much vxlan anymore so it may have been broken along the way and i wouldn't know. or it's using the wrong interface unknowingly. can confirm tomorrow | 04:20 |
| jamesdenton | yep yep. see ya tomorrow | 04:20 |
| broken_one | not using vxlan anymore...will have to talk more about that as well tomorrow. would be interested as to why | 04:21 |
| jamesdenton | legacy reasons mostly, but there are a few things we can discuss. see ya | 04:25 |
| *** mathlin has joined #openstack-ansible | 04:30 | |
| *** broken_one is now known as broken_one_ | 04:30 | |
| *** broken_one_ has quit IRC | 04:37 | |
| *** broken_one has joined #openstack-ansible | 04:44 | |
| *** udesale has joined #openstack-ansible | 04:50 | |
| *** cshen has joined #openstack-ansible | 04:53 | |
| *** markvoelker has joined #openstack-ansible | 04:59 | |
| *** mathlin has quit IRC | 05:04 | |
| *** cshen has quit IRC | 05:07 | |
| *** lbragstad has quit IRC | 05:12 | |
| *** mathlin has joined #openstack-ansible | 05:27 | |
| *** fatdragon has quit IRC | 05:34 | |
| *** fatdragon has joined #openstack-ansible | 05:35 | |
| *** fatdragon has quit IRC | 05:40 | |
| *** shyamb has joined #openstack-ansible | 05:42 | |
| *** radeks_ has joined #openstack-ansible | 05:42 | |
| *** TxGirlGeek has joined #openstack-ansible | 05:45 | |
| *** radeks__ has joined #openstack-ansible | 05:45 | |
| *** radeks_ has quit IRC | 05:47 | |
| *** radeks__ has quit IRC | 05:55 | |
| *** mathlin has quit IRC | 06:00 | |
| *** markvoelker has quit IRC | 06:03 | |
| *** cshen has joined #openstack-ansible | 06:04 | |
| *** markvoelker has joined #openstack-ansible | 06:05 | |
| *** cshen has quit IRC | 06:08 | |
| *** macza has joined #openstack-ansible | 06:10 | |
| *** cshen has joined #openstack-ansible | 06:11 | |
| *** macza has quit IRC | 06:14 | |
| *** cshen has quit IRC | 06:16 | |
| *** mathlin has joined #openstack-ansible | 06:24 | |
| *** vnogin has joined #openstack-ansible | 06:31 | |
| *** radeks__ has joined #openstack-ansible | 06:33 | |
| *** vnogin has quit IRC | 06:35 | |
| *** hamzaachi has joined #openstack-ansible | 06:57 | |
| *** mathlin has quit IRC | 06:58 | |
| *** TxGirlGeek has quit IRC | 07:03 | |
| *** shyamb has quit IRC | 07:09 | |
| *** kopecmartin has joined #openstack-ansible | 07:10 | |
| *** kopecmartin has quit IRC | 07:10 | |
| *** goldenfri has quit IRC | 07:12 | |
| *** kopecmartin has joined #openstack-ansible | 07:14 | |
| *** cshen has joined #openstack-ansible | 07:15 | |
| *** DanyC has joined #openstack-ansible | 07:15 | |
| *** goldenfri has joined #openstack-ansible | 07:17 | |
| *** DanyC has quit IRC | 07:20 | |
| *** mathlin has joined #openstack-ansible | 07:21 | |
| *** macza has joined #openstack-ansible | 07:27 | |
| *** macza has quit IRC | 07:31 | |
| *** goldenfri has quit IRC | 07:38 | |
| *** gkadam has joined #openstack-ansible | 07:43 | |
| *** goldenfri has joined #openstack-ansible | 07:45 | |
| *** tosky has joined #openstack-ansible | 07:49 | |
| *** pcaruana has joined #openstack-ansible | 07:51 | |
| *** mathlin has quit IRC | 07:55 | |
| *** fatdragon has joined #openstack-ansible | 08:01 | |
| *** czunker has joined #openstack-ansible | 08:08 | |
| *** fatdragon has quit IRC | 08:10 | |
| *** mathlin has joined #openstack-ansible | 08:17 | |
| *** hamzaachi_ has joined #openstack-ansible | 08:17 | |
| *** hamzaachi has quit IRC | 08:20 | |
| *** cshen has quit IRC | 08:24 | |
| *** hamzaachi_ has quit IRC | 08:25 | |
| *** hamzaachi_ has joined #openstack-ansible | 08:25 | |
| *** cshen has joined #openstack-ansible | 08:25 | |
| *** radeks_ has joined #openstack-ansible | 08:27 | |
| *** radeks__ has quit IRC | 08:29 | |
| *** cshen has quit IRC | 08:30 | |
| *** czunker has quit IRC | 08:32 | |
| *** cshen has joined #openstack-ansible | 08:32 | |
| *** cshen has quit IRC | 08:37 | |
| *** cshen has joined #openstack-ansible | 08:43 | |
| *** shyamb has joined #openstack-ansible | 08:45 | |
| *** cshen has quit IRC | 08:47 | |
| *** nsmeds has quit IRC | 08:51 | |
| *** mathlin has quit IRC | 08:51 | |
| *** thuydang has joined #openstack-ansible | 08:52 | |
| *** markvoelker has quit IRC | 08:52 | |
| *** thuydang has quit IRC | 08:53 | |
| *** thuydang has joined #openstack-ansible | 08:53 | |
| *** shyam89 has joined #openstack-ansible | 08:54 | |
| *** shyamb has quit IRC | 08:57 | |
| *** thuydang has quit IRC | 09:00 | |
| *** fatdragon has joined #openstack-ansible | 09:03 | |
| *** shyam89 has quit IRC | 09:05 | |
| *** shyam89 has joined #openstack-ansible | 09:05 | |
| *** fatdragon has quit IRC | 09:08 | |
| *** ygk_12345 has joined #openstack-ansible | 09:09 | |
| *** shardy has joined #openstack-ansible | 09:10 | |
| *** shardy has quit IRC | 09:13 | |
| *** shardy has joined #openstack-ansible | 09:13 | |
| *** mathlin has joined #openstack-ansible | 09:13 | |
| *** cshen has joined #openstack-ansible | 09:17 | |
| *** thuydang has joined #openstack-ansible | 09:18 | |
| *** jawad_axd has joined #openstack-ansible | 09:20 | |
| *** logan- has quit IRC | 09:21 | |
| *** cshen has quit IRC | 09:22 | |
| *** thuydang has quit IRC | 09:22 | |
| *** cshen has joined #openstack-ansible | 09:23 | |
| *** logan- has joined #openstack-ansible | 09:28 | |
| *** logan- has quit IRC | 09:28 | |
| *** shyam89 has quit IRC | 09:29 | |
| *** vnogin has joined #openstack-ansible | 09:31 | |
| *** dcdamien has joined #openstack-ansible | 09:31 | |
| *** vnogin has quit IRC | 09:35 | |
| *** dcdamien has quit IRC | 09:35 | |
| *** shyamb has joined #openstack-ansible | 09:38 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-os_cinder master: Ensure public volume type option is evaluated as a boolean https://review.openstack.org/629815 | 09:43 |
| jrosser | odyssey4me: ^ i did wonder about that | 09:46 |
| *** mathlin has quit IRC | 09:48 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/ansible-hardening master: Chrony: add an option to sync the hardware clock https://review.openstack.org/629330 | 09:49 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/ansible-hardening master: Chrony: new NTP server defaults https://review.openstack.org/629684 | 09:49 |
| *** markvoelker has joined #openstack-ansible | 09:49 | |
| *** DanyC has joined #openstack-ansible | 09:49 | |
| *** markvoelker has quit IRC | 09:51 | |
| *** luksky has joined #openstack-ansible | 09:52 | |
| odyssey4me | jamesdenton when you're online, I need to share how you can use MNAIO images to save yourself a lot of time and keep some environments available for testing without having to do a full setup each time | 09:55 |
| odyssey4me | cloudnull perhaps you should approach infra to get pasted installed on paste.openstack.org ;) | 09:57 |
| odyssey4me | jrosser wow, mitogen really makes that metal test go fast! | 10:01 |
| *** shyamb has quit IRC | 10:02 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-os_barbican master: Only implement policy.json if an override is configured https://review.openstack.org/629280 | 10:03 |
| *** shyamb has joined #openstack-ansible | 10:06 | |
| *** cshen has quit IRC | 10:06 | |
| *** priteau has joined #openstack-ansible | 10:08 | |
| *** hamzaachi_ has quit IRC | 10:09 | |
| *** hamzaachi_ has joined #openstack-ansible | 10:09 | |
| *** hamzaachi_ has quit IRC | 10:12 | |
| *** hamzaachi_ has joined #openstack-ansible | 10:12 | |
| *** mathlin has joined #openstack-ansible | 10:19 | |
| *** cshen has joined #openstack-ansible | 10:21 | |
| odyssey4me | jrosser about that heat issue, note that https://review.openstack.org/629544 passed with flying colours | 10:23 |
| odyssey4me | oh, I'm an idiot - I didn't add heat to the services | 10:23 |
| *** hwoarang has joined #openstack-ansible | 10:25 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-os_heat master: Test using integrated build https://review.openstack.org/629544 | 10:30 |
| *** cshen has quit IRC | 10:31 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-ceph_client master: Test using integrated build https://review.openstack.org/623253 | 10:32 |
| *** cshen has joined #openstack-ansible | 10:39 | |
| deployer2 | odyssey4me, jrosser, Hi! guys may I ask 2 short questions:1)should fresh OSA Rocky deployment rsyslog containers listen on syslog port? Investigating why those do not but setup playbooks ran without issue. Cannot find how logs are forwarded to rsyslog containers. | 10:41 |
| deployer2 | 2)can you reccomend any good documentation corner regarding domains? Wishing to set up "domain admins", but documentation regarding domains is quite scarse. | 10:41 |
| odyssey4me | deployer2 from rocky onwards, for all distributions other than suse, systemd journals are used instead of rsyslog for centralising the log files as far as I know | 10:42 |
| odyssey4me | for 2 - not really, perhaps best to ask in #openstack-keystone if you can't find appropriate docs | 10:43 |
| deployer2 | odyssey4me, regarding logs that could be it! Thanks, need to dig in! I was cought on changes queens->rocky as well regarding that dhcp services moved from containers to namespaces and could not find those :) | 10:46 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-galera_server master: Update RPM-GPG-KEY-Percona https://review.openstack.org/629830 | 10:48 |
| *** mathlin has quit IRC | 10:54 | |
| deployer2 | odyssey4me I guess I need to report a bug in doc then here https://docs.openstack.org/openstack-ansible-rsyslog_server/latest/ops-logging.html as that still points to log paths that do not exist. | 10:54 |
| *** vnogin has joined #openstack-ansible | 10:57 | |
| *** strobelight has quit IRC | 11:02 | |
| *** fatdragon has joined #openstack-ansible | 11:04 | |
| *** openstackgerrit has quit IRC | 11:05 | |
| *** fatdragon has quit IRC | 11:10 | |
| *** udesale has quit IRC | 11:14 | |
| *** shardy has quit IRC | 11:14 | |
| *** udesale has joined #openstack-ansible | 11:14 | |
| *** mathlin has joined #openstack-ansible | 11:16 | |
| *** electrofelix has joined #openstack-ansible | 11:28 | |
| *** shyamb has quit IRC | 11:29 | |
| *** cshen has quit IRC | 11:40 | |
| *** shyamb has joined #openstack-ansible | 11:42 | |
| *** shyamb has quit IRC | 11:44 | |
| *** shyamb has joined #openstack-ansible | 11:44 | |
| *** cshen has joined #openstack-ansible | 11:44 | |
| *** shardy has joined #openstack-ansible | 11:46 | |
| *** mathlin has quit IRC | 11:50 | |
| *** dcdamien has joined #openstack-ansible | 11:57 | |
| *** rgogunskiy has joined #openstack-ansible | 12:04 | |
| odyssey4me | deployer2 well, that might exist if that role is used at all - not sure | 12:10 |
| *** shyamb has quit IRC | 12:12 | |
| *** shyamb has joined #openstack-ansible | 12:12 | |
| *** mathlin has joined #openstack-ansible | 12:13 | |
| *** luksky has quit IRC | 12:15 | |
| *** strobelight has joined #openstack-ansible | 12:22 | |
| *** openstackgerrit has joined #openstack-ansible | 12:23 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-galera_server master: Update RPM-GPG-KEY-Percona and add PERCONA-PACKAGING-KEY https://review.openstack.org/629830 | 12:23 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-galera_server master: Add PERCONA-PACKAGING-KEY https://review.openstack.org/629830 | 12:24 |
| *** thuydang has joined #openstack-ansible | 12:26 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-galera_server master: Add PERCONA-PACKAGING-KEY https://review.openstack.org/629830 | 12:27 |
| openstackgerrit | lvxianguo proposed openstack/openstack-ansible-os_nova master: modify the features description https://review.openstack.org/629842 | 12:28 |
| *** cshen has quit IRC | 12:29 | |
| *** thuydang has quit IRC | 12:31 | |
| *** thuydang has joined #openstack-ansible | 12:34 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-galera_server master: Add PERCONA-PACKAGING-KEY https://review.openstack.org/629830 | 12:34 |
| *** cshen has joined #openstack-ansible | 12:37 | |
| *** thuydang has quit IRC | 12:38 | |
| *** cshen has quit IRC | 12:42 | |
| *** cshen has joined #openstack-ansible | 12:46 | |
| *** priteau has quit IRC | 12:48 | |
| *** mathlin has quit IRC | 12:48 | |
| odyssey4me | other than that, I guess you'd need to figure out what the priority areas are for your env and focus them on efforts in that space in particular | 12:48 |
| odyssey4me | oops, wrong window :p | 12:48 |
| *** cshen has quit IRC | 12:51 | |
| *** shyamb has quit IRC | 12:57 | |
| odyssey4me | ok cores, it looks to me like https://review.openstack.org/629830 resolves the issue with centos builds failing - reviews asap please, because that'll also require porting back | 13:00 |
| *** fatdragon has joined #openstack-ansible | 13:06 | |
| ygk_12345 | jrosser: my installation of rocky is failing | 13:07 |
| ygk_12345 | jrosser: http://paste.openstack.org/show/741573/ | 13:07 |
| ygk_12345 | jrosser: http://paste.openstack.org/show/741574/ | 13:09 |
| *** rgogunskiy has quit IRC | 13:09 | |
| *** mathlin has joined #openstack-ansible | 13:11 | |
| *** fatdragon has quit IRC | 13:11 | |
| *** luksky has joined #openstack-ansible | 13:16 | |
| *** dave-mccowan has joined #openstack-ansible | 13:18 | |
| *** fnpanic has quit IRC | 13:19 | |
| *** ygk_12345 has quit IRC | 13:25 | |
| *** stuartgr has joined #openstack-ansible | 13:28 | |
| *** cshen has joined #openstack-ansible | 13:29 | |
| *** cshen has quit IRC | 13:34 | |
| *** ansmith_ has joined #openstack-ansible | 13:41 | |
| *** priteau has joined #openstack-ansible | 13:42 | |
| *** mathlin has quit IRC | 13:45 | |
| *** dave-mccowan has quit IRC | 13:48 | |
| *** dave-mccowan has joined #openstack-ansible | 13:51 | |
| *** ygk_12345 has joined #openstack-ansible | 13:52 | |
| *** priteau has quit IRC | 13:54 | |
| ygk_12345 | guilhermesp: the playbook failed again due this bug http://paste.openstack.org/show/741579/ | 13:54 |
| *** lbragstad has joined #openstack-ansible | 13:58 | |
| *** priteau has joined #openstack-ansible | 14:04 | |
| odyssey4me | evrardjp can you review https://review.openstack.org/629830 please to unblock master? otherwise hopefully mnaser will be in soon | 14:05 |
| *** mathlin has joined #openstack-ansible | 14:08 | |
| evrardjp | odyssey4me: yes just a minute | 14:10 |
| *** fnpanic has joined #openstack-ansible | 14:10 | |
| fnpanic | hi | 14:10 |
| odyssey4me | fnpanic o/ | 14:12 |
| *** TxGirlGeek has joined #openstack-ansible | 14:23 | |
| *** vnogin has quit IRC | 14:30 | |
| *** electrofelix has quit IRC | 14:31 | |
| openstackgerrit | Merged openstack/openstack-ansible-galera_server master: Add PERCONA-PACKAGING-KEY https://review.openstack.org/629830 | 14:40 |
| *** mathlin has quit IRC | 14:42 | |
| ygk_12345 | odyssey4me: is anyone working on this bug https://bugs.launchpad.net/openstack-ansible/+bug/1807346 | 14:46 |
| openstack | Launchpad bug 1807346 in openstack-ansible "[heat] Installations fails during Update Q->R" [Undecided,New] - Assigned to Guilherme Steinmuller Pimentel (guilhermesp) | 14:46 |
| ygk_12345 | odyssey4me: still the playbook is failing on a fresh rocky install | 14:47 |
| odyssey4me | ygk_12345 this is an open source community - people work on bugs when they have the time to do so... and if you need a fix urgently, then you should either diagnose the issue and provide the requisite details to someone who can prepare a patch, or you should prepare a patch if you can | 14:48 |
| odyssey4me | currently the bug is assigned to guilhermesp - but he hasn't responded to your prompts, so he's possibly on holiday or busy elsewhere | 14:49 |
| ygk_12345 | odyssey4me: ok | 14:50 |
| *** ygk_12345 has quit IRC | 14:51 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-galera_server stable/rocky: Add PERCONA-PACKAGING-KEY https://review.openstack.org/629879 | 14:52 |
| jenglisch | i'm currently using this patch for LXC wiring with ovs https://review.openstack.org/#/c/518230/2/ . i'd like to integrate it into upstream if possible, since the autodetection is quite unreliable, i'd leave it to the user by introducting a new config option | 14:53 |
| jenglisch | what do you think? | 14:53 |
| jenglisch | *the proposed detection mechanism | 14:53 |
| odyssey4me | jenglisch that may be of interest for mnaser and jamesdenton to look through when they come online | 14:53 |
| jamesdenton | aye | 14:54 |
| *** cshen has joined #openstack-ansible | 14:56 | |
| *** shardy_ has joined #openstack-ansible | 14:57 | |
| *** jawad_axd has quit IRC | 14:58 | |
| *** shardy has quit IRC | 15:00 | |
| *** cshen has quit IRC | 15:01 | |
| mnaser | thanks for the unblock catch odyssey4me | 15:03 |
| *** TxGirlGeek has quit IRC | 15:03 | |
| jamesdenton | markos has a good point. there does not appear to be a way to define the bridge type (linuxbridge vs ovs). I would think a config option could be added that would default to 'bridge' for backwards compatibility. Interestingly enough, the docs call out using OVS bridges instead of linux bridges, but i can't see where the wiring would happen to support that. https://docs.openstack.org/openstack-ansible-os_neutron/latest | 15:04 |
| jamesdenton | /app-openvswitch.html | 15:04 |
| *** mathlin has joined #openstack-ansible | 15:05 | |
| *** TxGirlGeek has joined #openstack-ansible | 15:06 | |
| *** fatdragon has joined #openstack-ansible | 15:07 | |
| jenglisch | another thingy on nova: nova_dhcp_domain has been deprecated from nova BUT nova still uses this config option and it is the only way to fix the metadata for nova instances, sind the fqdn is always novalocal (the default from nova) if it's not confiured by OSA | 15:08 |
| jenglisch | https://github.com/openstack/openstack-ansible-os_nova/blob/9d558a5922465b50d7fd924d23a0ef8e78e75bcd/releasenotes/notes/nova-ocata-conf-deprecations-0a99d13c97424c51.yaml | 15:08 |
| jenglisch | https://github.com/openstack/nova/blob/c6218428e9b29a2c52808ec7d27b4b21aadc0299/nova/api/metadata/base.py#L538 | 15:08 |
| jenglisch | since the option is still in use, is reintroducing it into OSA a bad thing? | 15:08 |
| mnaser | i wonder why we deprecated it | 15:10 |
| *** TxGirlGeek has quit IRC | 15:10 | |
| mnaser | jenglisch: any reason why you don't want to use overrides for this? | 15:10 |
| jenglisch | well nova itself deprecated it aswell | 15:10 |
| jenglisch | but they're still using it in their code, so the variable is not dead | 15:10 |
| mnaser | i'm trying to push to minimize the # of values we predefine to avoid moving away from upstream | 15:11 |
| *** deployer2 has quit IRC | 15:11 | |
| mnaser | because what happens if we define it with a value of novalocal. .. if that changes, we'll probably miss it | 15:11 |
| *** fatdragon has quit IRC | 15:12 | |
| jenglisch | indeed, i'm just looking through some patch-sets i made previously (in my lab) - however config overrides are certainly the better option, thanks ! | 15:12 |
| jamesdenton | looks like dhcp_domain deprecation in nova was.. premature.. https://etherpad.openstack.org/p/YVR18-novaneutron-ops-cross-project-session | 15:14 |
| jenglisch | jamesdenton: yes. i don't think it's good to depend on the installed packages, since ovs and bridge-tools are quite common ones, and especially brctl is deprecated in favour of 'bridge' already. if a config option is fine, i'd propose a change for that, defaulting to bridge wiring, and add the possibility to change it to the docs / examples of ovs | 15:14 |
| jamesdenton | *thumbs up* | 15:15 |
| odyssey4me | jenglisch you can put *any* .conf/.ini/.json/.yaml file entry in place with https://docs.openstack.org/openstack-ansible/latest/reference/configuration/using-overrides.html#overriding-openstack-configuration-defaults regardless of whether there is a specific var to override something in a role | 15:15 |
| jenglisch | i could just imaginge to implement a default if only one of them is installed, and fail on runtime if both are detected, but i'm not sure if this is a good solution | 15:15 |
| jamesdenton | i just noticed that patch is like, over a year old :o | 15:17 |
| jamesdenton | i vote for keeping it simple... w/o autodection. | 15:17 |
| jamesdenton | *detection | 15:17 |
| *** mattt has left #openstack-ansible | 15:18 | |
| jenglisch | i'd add it in the 'global_overrides' section https://github.com/openstack/openstack-ansible/blob/708b51d6d37574d73b1cd4eb666986c9707fa2d1/etc/openstack_deploy/openstack_user_config.yml.prod-ceph.example#L23 | 15:22 |
| jenglisch | in theory this variable could be a host_variable, if someone decides to deploy a mixed env of linux bridges and ovs | 15:23 |
| broken_one | adding fallback behaviour doesnt sound right ot me | 15:23 |
| broken_one | i mean if I as a user want OVS then why give me linuxbridge functionality? | 15:23 |
| jenglisch | broken_one: hm? | 15:24 |
| broken_one | jenglisch: something i saw mentioned in the chat to default to bridge | 15:25 |
| broken_one | and good morning everyone unless you are on the otherside of the world then good evening o/ | 15:26 |
| jenglisch | well, this is the current behaviour, ovs is currently broken/unsupported actually | 15:26 |
| jamesdenton | it's related to keeping compatibility with existing behavior, which means that control plane bridges (br-mgmt, br-storage, etc.) default to linux bridges rather than openvswitches. | 15:26 |
| jenglisch | so the default wouldn't require current installations to set this explicit | 15:27 |
| broken_one | ovs is broken in OSA currently? ( rocky ) | 15:27 |
| jamesdenton | not that i'm aware | 15:27 |
| broken_one | jenglisch: can you elaborate? | 15:28 |
| jenglisch | the last time i checked, the lxc-wiring script doesn't work for ovs | 15:28 |
| jenglisch | since brctl addif likes to operate on linux bridges only | 15:28 |
| jamesdenton | traditionally, OVS has been used only for the neutron provider networks and the control plane (containers) remained connected to linuxbridges | 15:28 |
| *** vnogin has joined #openstack-ansible | 15:30 | |
| broken_one | i dont remember reading any of this in the OSA docs. Did i miss it? | 15:30 |
| *** vnogin has quit IRC | 15:31 | |
| jenglisch | https://docs.openstack.org/openstack-ansible-os_neutron/latest/app-openvswitch.html ; configuring bridges (open vswitch) | 15:31 |
| *** priteau has quit IRC | 15:31 | |
| jenglisch | the lxc container create role creates the lxc container and uses the lxc-wiring script afterwards to attach the lxc container to the required bridges (as defined in global_overrides/network/*) | 15:32 |
| jenglisch | https://github.com/openstack/openstack-ansible-lxc_container_create/blob/927fcae2ac01e5d4594111df197f1d32f42fbbff/tasks/lxc_container_config.yml#L187 | 15:33 |
| jenglisch | https://github.com/openstack/openstack-ansible-lxc_container_create/blob/master/files/lxc-veth-wiring.sh#L55 -> this only works for linux bridges | 15:33 |
| broken_one | there is only 1 mention of an issue with OVS in ubuntu 16 witha work around | 15:34 |
| broken_one | nothing mentions issues with OVS specifically | 15:34 |
| broken_one | i will note | 15:34 |
| broken_one | inside any of the containers...i noticed none of the spcified veth devices are there | 15:34 |
| jenglisch | the fix is quite easy, since you only need to call 'ovs-vsctl add-port' instead of 'brctl addif' | 15:34 |
| broken_one | and the veth pairs in OVS never get created | 15:34 |
| jenglisch | https://review.openstack.org/#/c/518230/2/files/lxc-veth-wiring.sh | 15:34 |
| odyssey4me | jenglisch essentially the OSA control plane should run linuxbridge and OpenStack can use OVS - there's no real reason to make the OSA containers use OVS as far as I can see | 15:35 |
| mnaser | jenglisch, odyssey4me: indeed, i have alway used normal linux bridges for the OSA containers | 15:35 |
| odyssey4me | trying to force everything to use OVS seems somewhat pointless | 15:35 |
| mnaser | but OVS for the actual networking | 15:35 |
| broken_one | makes the entire system homogenous | 15:36 |
| broken_one | i can see arguments both ways | 15:36 |
| mnaser | my concern with that is neutron and lxc using the same ovs instance | 15:36 |
| mnaser | i think neutron usually assumes it owns ovs so that might make weird things happen | 15:36 |
| jenglisch | i decided to setup an environment with OVS only, since i don't wanted to introduce the complexity of two different switching planes | 15:36 |
| mnaser | odyssey4me: https://review.openstack.org/#/c/628718/ -- can we replicate this across all other roles? | 15:37 |
| odyssey4me | jenglisch sure, but then you walk into a situation where you're using something totally untested in all OSA testing | 15:37 |
| broken_one | jenglisch: trying to do that as well, and having a bit of issues | 15:37 |
| mnaser | i can have someone do it, but it pretty much means developer mode seems broken | 15:37 |
| jenglisch | hm, i hadn't any issues in my env with ~40 nodes and ovs only | 15:38 |
| *** vicky84 has joined #openstack-ansible | 15:38 | |
| odyssey4me | mnaser if you'd like, sure | 15:38 |
| jenglisch | but that doesn't mean there aren't any | 15:38 |
| mnaser | jenglisch: that's awesome | 15:38 |
| mnaser | odyssey4me: i just wanted to make sure it was a valid change, as python_venv_build is your domain :) | 15:38 |
| odyssey4me | mnaser ok, I haven't really had time to look at it properly | 15:38 |
| mnaser | before we throw it a zillion patches | 15:38 |
| odyssey4me | lemme do that in a bit - meetings for the next few hours :/ | 15:39 |
| broken_one | jenglisch: currently deploying rocky OVA and having issues with things wiring up to br-vxlan...everything is moving traffic through br-mgmt and L3 HA routers are showing down on all interfaces | 15:39 |
| jrosser | I guess we pay no attention to isolation between control plane and neutron in that kind of converged deployment | 15:39 |
| jrosser | I’d be quite worried about that | 15:39 |
| mnaser | odyssey4me: no worries but pretty much https://github.com/openstack/ansible-role-python_venv_build/blob/master/tasks/python_venv_wheel_build.yml#L19-L20 vs https://review.openstack.org/#/c/628718/1/tasks/nova_install_source.yml | 15:39 |
| *** mathlin has quit IRC | 15:39 | |
| jamesdenton | broken_one i will be looking into that later this morning | 15:39 |
| evrardjp | jrosser: +1 . That for me is a no-go | 15:39 |
| evrardjp | but ppl might be different :) | 15:39 |
| broken_one | jrosser: isolation happens with netowrks either in vlan tagging or the flow tables in OVS | 15:40 |
| mnaser | they mismatch and it tries to do the wrong thing and delegates to something that doesn't exist (possibly) | 15:40 |
| odyssey4me | mnaser there we. go -that's what I wanted to verify | 15:40 |
| odyssey4me | ok, good to go - let's do it | 15:40 |
| evrardjp | it's also possible to have two different ovs at the same time IIRC | 15:40 |
| mnaser | odyssey4me: the other tiny one ill do is that we use `/opt/developer-pip-constraints.txt` everywhere | 15:40 |
| broken_one | jamesdenton: i appreciate that. | 15:40 |
| mnaser | but that leads to roles being non idempotent | 15:40 |
| mnaser | because if you run two roles on the same node, they overwrite each othre | 15:40 |
| odyssey4me | mnaser don't bother changing that - that's going to go away | 15:40 |
| mnaser | ah | 15:40 |
| mnaser | okay | 15:40 |
| mnaser | :) | 15:40 |
| odyssey4me | I've got work part-done to change that up | 15:40 |
| mnaser | awesome | 15:41 |
| mnaser | good sync :) | 15:41 |
| broken_one | jenglisch: would you mind haring your configs and any customization you have done to get your 40 node cloud up and running? | 15:41 |
| odyssey4me | I'll hopefully put time into that tomorrow. | 15:41 |
| broken_one | s/haring/sharing | 15:41 |
| *** priteau has joined #openstack-ansible | 15:46 | |
| jenglisch | broken_one: yep, it's currenly running queens, i just have to leave now, so i'll come back to you later/tomorrow | 15:54 |
| broken_one | jenglisch: would help out. have a good one :) | 15:55 |
| *** vnogin has joined #openstack-ansible | 15:56 | |
| cloudnull | mornings | 15:58 |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible-os_cinder master: Cleanup files and templates using smart sources https://review.openstack.org/588953 | 15:59 |
| broken_one | morning cloudnull o/ | 16:03 |
| *** udesale has quit IRC | 16:06 | |
| evrardjp | mnaser: about releases, I propose I do a first release of OSA stein, then bump master branches of upstream projects, because I didn't bump for a while (my bad) | 16:07 |
| cloudnull | o/ broken_one | 16:08 |
| cloudnull | evrardjp +1 | 16:08 |
| openstackgerrit | Merged openstack/openstack-ansible-galera_server stable/rocky: Add PERCONA-PACKAGING-KEY https://review.openstack.org/629879 | 16:10 |
| *** TxGirlGeek has joined #openstack-ansible | 16:14 | |
| *** hamzaachi_ has quit IRC | 16:17 | |
| mnaser | evrardjp: how far back are right now? also, i think our gate is broken or maybe just recently fixed | 16:19 |
| jrosser | ceph jobs are still sad | 16:26 |
| mnaser | reeee | 16:27 |
| mnaser | okay | 16:27 |
| mnaser | i need to finish up something related to keystone and ill try to look into it | 16:28 |
| jrosser | i have this in progress https://review.openstack.org/#/c/629317/ but i've made a rookie error on the task i added | 16:28 |
| *** mathlin has joined #openstack-ansible | 16:28 | |
| * jrosser not able to push code right now | 16:29 | |
| *** macza has joined #openstack-ansible | 16:37 | |
| *** nsmeds has joined #openstack-ansible | 16:37 | |
| *** macza_ has joined #openstack-ansible | 16:40 | |
| *** macza has quit IRC | 16:42 | |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible master: Call ceph-facts role as required by changes to ceph-ansible https://review.openstack.org/629317 | 16:44 |
| cloudnull | jrosser ^ pushed a change to quote the octal | 16:45 |
| *** strattao has joined #openstack-ansible | 16:50 | |
| *** luksky has quit IRC | 16:56 | |
| *** cshen has joined #openstack-ansible | 16:57 | |
| *** fatdragon has joined #openstack-ansible | 17:00 | |
| *** DanyC has quit IRC | 17:00 | |
| *** cshen has quit IRC | 17:02 | |
| *** mathlin has quit IRC | 17:03 | |
| *** DanyC has joined #openstack-ansible | 17:03 | |
| jamesdenton | broken_one I think I found the issue. The single quotes here, https://github.com/openstack/openstack-ansible/blob/master/playbooks/common-tasks/dynamic-address-fact.yml#L22, result in a path of "'ansible_br_vxlan'.ipv4.address" instead of "ansible_br_vxlan.ipv4.address". Doesn't match the path in the facts, and defaults to ansible_host | 17:06 |
| broken_one | jamesdenton: i can agree with that somewhat | 17:06 |
| broken_one | the next part would be | 17:07 |
| jamesdenton | Ahh, it was this: https://github.com/openstack/openstack-ansible/commit/9f6bf94d1a922553f7bc9b3a614b857b666822db#diff-255f0b55f1833e385fc3b3edbe740135 | 17:07 |
| broken_one | none of the containers are wired to those networks | 17:07 |
| *** DanyC has quit IRC | 17:07 | |
| *** vnogin has quit IRC | 17:07 | |
| jamesdenton | well, it's not really related to the containers as much as it's the method in which the playbook determines the address to use for tunnel address (local_ip) | 17:08 |
| openstackgerrit | Merged openstack/openstack-ansible-os_nova master: venv: use inventory_hostname instead of ansible_hostname https://review.openstack.org/628718 | 17:08 |
| *** vnogin has joined #openstack-ansible | 17:08 | |
| broken_one | ok so that will fix the local_ip issue | 17:08 |
| jamesdenton | yes | 17:08 |
| *** jpward1981 has joined #openstack-ansible | 17:08 | |
| broken_one | i am certain of this | 17:09 |
| broken_one | so i did it as well | 17:09 |
| jamesdenton | but, i'll have to talk to noonedeadpunk and see how to go about it | 17:09 |
| broken_one | of course :) | 17:09 |
| broken_one | so as a further test | 17:09 |
| broken_one | even if you manually flip the IP | 17:09 |
| broken_one | then things really start breaking | 17:09 |
| broken_one | the dhcp agents | 17:10 |
| broken_one | i made the manual ip change | 17:10 |
| noonedeadpunk | Am I missed smth? | 17:10 |
| broken_one | rebooted a controller node | 17:10 |
| broken_one | and the dhcp agent from my test subnet on that controller | 17:10 |
| broken_one | is now down | 17:10 |
| broken_one | so what i think the issue is now | 17:11 |
| broken_one | is what jenglisch proposed | 17:11 |
| broken_one | https://review.openstack.org/#/c/518230/2/files/lxc-veth-wiring.sh | 17:11 |
| broken_one | since i am not doing linux bridges anywhere | 17:11 |
| broken_one | everything is OVS | 17:11 |
| jamesdenton | hey noonedeadpunk. That patch seems to have a side effect of causing the dynamic address determination fail and default to ansible_host | 17:11 |
| noonedeadpunk | jamesdenton once you pass "'ansible_br_vxlan'.ipv4.address" to json_query(), quotes should be interpreted correctly | 17:11 |
| broken_one | noonedeadpunk: the find_bridge fails in my setup | 17:12 |
| broken_one | returns undefined for br-vxlan and others | 17:12 |
| jamesdenton | it appears to just silently fail and fallback to ansible_host | 17:12 |
| *** vnogin has quit IRC | 17:13 | |
| noonedeadpunk | hm, that's strange... | 17:13 |
| jamesdenton | broken_one are you using OVS bridges for everything, including br-mgmt? | 17:14 |
| broken_one | i am | 17:14 |
| jamesdenton | and do you have the local_ip configured on an interface? Which one? | 17:14 |
| broken_one | the ip i want is on br-vxlan | 17:14 |
| noonedeadpunk | oh, I haven't tested OVS, only linuxbridges... | 17:14 |
| jamesdenton | noonedeadpunk disregard ovs vs lb. | 17:14 |
| broken_one | let me show you how i overcame the issue in the playbook | 17:14 |
| broken_one | one moment | 17:15 |
| noonedeadpunk | jamesdenton so my point of the patch was to be able to use aliases, like br0.0, as in this case json_query() interprets this dot in the interface name as an extra element in the dictionary | 17:16 |
| broken_one | http://paste.openstack.org/show/741594/ -- dynamic-address-fact.yml | 17:17 |
| broken_one | so i allow the wrong IP to be facted,then the following tests "fix" the ip issue | 17:18 |
| openstackgerrit | Merged openstack/openstack-ansible-os_neutron master: vars: stop placing neutron_developer_constraints in vars https://review.openstack.org/628747 | 17:18 |
| noonedeadpunk | I think, that ansible_host may be set either in case when json_query(find_bridge) or replace('-','_') fails | 17:21 |
| *** gkadam has quit IRC | 17:21 | |
| jamesdenton | http://paste.openstack.org/show/741595/ | 17:21 |
| jamesdenton | first example wrapped in quotes, second without | 17:22 |
| *** gyee has joined #openstack-ansible | 17:24 | |
| *** mathlin has joined #openstack-ansible | 17:25 | |
| evrardjp | mnaser: far back, beginning of cycle | 17:32 |
| evrardjp | I want to have a working master before bumping it and see the fallout | 17:32 |
| evrardjp | I need to focus on automating this with a proposal bot | 17:33 |
| *** priteau has quit IRC | 17:33 | |
| noonedeadpunk | jamesdenton it seems, you're right.... | 17:34 |
| noonedeadpunk | I'm trying to do smth with it, or, probably, will just revert this commit | 17:42 |
| jamesdenton | i'm trying a few permutations ,but no luck so far | 17:43 |
| noonedeadpunk | jamesdenton: so the problem I've tried to fix (debug is just printing metal_query): http://paste.openstack.org/show/741597/ | 17:44 |
| *** thuydang has joined #openstack-ansible | 17:45 | |
| broken_one | noonedeadpunk: does it also fail on vlan tagged subinterfaces? | 17:46 |
| broken_one | i.e. bond0.256 | 17:46 |
| broken_one | ? | 17:46 |
| noonedeadpunk | broken_one: this dot inside interface in interpreted incorrectly by json_query(), so yep, it should | 17:47 |
| noonedeadpunk | but now I'm already not 100% sure in anything... | 17:47 |
| broken_one | not sure how to rejoin the tokenized string back together with ansible | 17:48 |
| broken_one | easy in python | 17:48 |
| broken_one | ansible not so much | 17:48 |
| *** vnogin has joined #openstack-ansible | 17:48 | |
| openstackgerrit | Michael Vollman proposed openstack/openstack-ansible-os_manila master: Basic working os_manila role https://review.openstack.org/611930 | 17:50 |
| *** vnogin has quit IRC | 17:50 | |
| *** dcdamien has quit IRC | 17:52 | |
| broken_one | noonedeadpunk: i messed with it for about 3 hours yesterday and could only keep functionality by hacking together my paste | 17:52 |
| broken_one | i still ahve more conditions to test i think but it "works" | 17:53 |
| *** shardy_ has quit IRC | 17:53 | |
| noonedeadpunk | I think, I'll revert patch as for know and will think about solution after. | 17:55 |
| *** DanyC has joined #openstack-ansible | 17:58 | |
| *** DanyC has quit IRC | 17:59 | |
| broken_one | ok i will test your patch after the merge for sure | 17:59 |
| broken_one | the next issue is the veth wiring | 17:59 |
| *** DanyC has joined #openstack-ansible | 17:59 | |
| broken_one | i will also put the patch from jenglisch for the script file | 18:00 |
| *** mathlin has quit IRC | 18:00 | |
| openstackgerrit | Guilherme Steinmuller Pimentel proposed openstack/openstack-ansible-os_placement master: [WIP] Create base files to install placement https://review.openstack.org/618820 | 18:01 |
| *** TxGirlGeek has quit IRC | 18:02 | |
| *** TxGirlGeek has joined #openstack-ansible | 18:03 | |
| openstackgerrit | Mohammed Naser proposed openstack/openstack-ansible-os_keystone master: Allow messaging to be disabled https://review.openstack.org/622551 | 18:03 |
| *** thuydang has left #openstack-ansible | 18:05 | |
| noonedeadpunk | have some idea, but need to test it. | 18:05 |
| broken_one | if you need a 2nd tester for your idea then paste the code and I will be happy to | 18:07 |
| *** cshen has joined #openstack-ansible | 18:12 | |
| *** kopecmartin is now known as kopecmartin|off | 18:13 | |
| *** electrofelix has joined #openstack-ansible | 18:15 | |
| *** electrofelix has quit IRC | 18:20 | |
| noonedeadpunk | broken_one: for sure I will. I'm not the fastest one, and have to participate in a meeting now.... | 18:20 |
| *** mathlin has joined #openstack-ansible | 18:23 | |
| *** luksky has joined #openstack-ansible | 18:24 | |
| openstackgerrit | Jean-Philippe Evrard proposed openstack/openstack-ansible master: Make OVN track master branch https://review.openstack.org/629914 | 18:25 |
| broken_one | noonedeadpunk: sure thing, just let me know when :) | 18:25 |
| broken_one | jamesdenton: would it be a large effort to have OSA assume all things are OVS instead of a mixture of linuxbridges and ovs? | 18:26 |
| jamesdenton | at this point? yes, i think so. | 18:26 |
| jamesdenton | there's getting it to work (probably straightforward) and having it be supported (not likely to happen any time soon) | 18:27 |
| broken_one | what would be the criteria to get it supported? | 18:27 |
| jamesdenton | testing | 18:27 |
| jamesdenton | and adoption | 18:27 |
| broken_one | <--- first candidate :) | 18:28 |
| jamesdenton | and development | 18:28 |
| jamesdenton | heh | 18:28 |
| broken_one | i have a couple of people interested | 18:28 |
| jrosser | and confidence that architecturally we're happy | 18:28 |
| jamesdenton | ^^ | 18:28 |
| jrosser | given that the current controlplane linuxbridge + compute/neutron OVS works just fine | 18:29 |
| broken_one | so we want to use OSA instead of triple0 in production | 18:29 |
| evrardjp | I don't think it would be that bad to implement | 18:29 |
| evrardjp | it's just that nobody really cared that much | 18:29 |
| broken_one | trust me | 18:29 |
| broken_one | i/we care | 18:29 |
| evrardjp | Then I guess we'll see the code :) | 18:29 |
| evrardjp | for OSA you probably need to have a different scenario that deploys OSA with different networks using OVS. Shouldn't be that hard, but that's a first step. the bootstrap host role need change there to adapt this. | 18:30 |
| broken_one | ok will meet with the team in about an hour and we will determine the amount of effort required | 18:30 |
| jrosser | interesting to understand properly why its a big deal | 18:31 |
| evrardjp | Then with that in mind, you'll probably need to wire things properly. Assuming you want minimum effort, and avoid containers, you could be ready to go very fast, just need a few overrides in openstack_hosts I'd say | 18:31 |
| jrosser | partiuclarly as you'd be in a small club with that setup, when it comes to getting help / comparing issues | 18:31 |
| evrardjp | but like jrosser said -- There are many fallouts you'll have in prod. with one OVS if you are not careful, and you'll be alone | 18:32 |
| broken_one | i havent seen a big issue with going all OVS | 18:32 |
| evrardjp | I guess it depends on how noisy your neighbours are :D | 18:33 |
| broken_one | i dont see neutron messing with the sdn bridges | 18:33 |
| evrardjp | yeah that thing has been solved in 14.04 IIRC | 18:33 |
| jamesdenton | it's absolutely doable. For OSA, though, it's a bit of an effort to go that direction, that's all. Like you said, TripleO/OSP and some of the others do it. | 18:33 |
| evrardjp | yeah | 18:33 |
| broken_one | yeah | 18:33 |
| broken_one | we just like OSA modularity | 18:34 |
| broken_one | so beautifully designed imo | 18:34 |
| evrardjp | thanks | 18:34 |
| evrardjp | that's always nice to hear | 18:34 |
| jamesdenton | i have yet to justify the effort to personally make it happen and work out all of the tests, AIO/MNAIO scripts, etc. to make it go. | 18:34 |
| broken_one | well if all goes sideways, we do have linuxbridges working...i think without neutron router HA | 18:35 |
| broken_one | so we have a fallback | 18:35 |
| *** DanyC_ has joined #openstack-ansible | 18:35 | |
| broken_one | and can stick with OSA | 18:35 |
| broken_one | but we want OVS and HA routers | 18:35 |
| jrosser | broken_one: can i just check we are all talking about the same thing | 18:36 |
| broken_one | sure | 18:36 |
| jamesdenton | well, it should be pointed out that there is a pretty good separation of control and data plane networks with OSA. You *can* have linuxbridge for control plane (only thing we do) and OVS for data plane (VM networks). They can co-exist quite well | 18:36 |
| jrosser | you can have neutron OVS today and the control plane containers using linuxbridge | 18:36 |
| jrosser | thats all completely OK afaik | 18:36 |
| broken_one | i agree with that | 18:36 |
| jrosser | the only sticky bit for OSA is bringing the whole control plane onto OVS | 18:36 |
| jrosser | if you are prepared to compromise on that you follow a tested supported path here | 18:37 |
| evrardjp | broken_one: you can have ha routers without OVS | 18:37 |
| jamesdenton | your vxlan issues may be a result of the bug we found earlier with IP detection and something else undetermined. Rest assured, we manage many working production setups in pure LB and hybrid LB/OVS scenarios | 18:37 |
| jrosser | jamesdenton: i was interested in comparing notes on that VXLAN stuff with you | 18:38 |
| jrosser | i'm doing VXLAN and HA with linucbridge so it all seems OK here | 18:38 |
| *** DanyC has quit IRC | 18:38 | |
| broken_one | ok then my team will look at both scenarios | 18:39 |
| broken_one | the tested/supported way with LBR and OVS | 18:39 |
| broken_one | and then going full OVS | 18:39 |
| *** DanyC_ has quit IRC | 18:39 | |
| jamesdenton | jrosser the issues we had may be legacy issues related to lack of multicast support (on our end), l2population issues (w/ allowed-address-pairs), MTUs, etc. I think a lot of that has been addressed in recent (post Newton) versions, but we pretty much standardized on vlan at that point | 18:39 |
| jrosser | jamesdenton: ah ok - i'm doing full multicast so might not see the same things | 18:40 |
| broken_one | ah yes jamesdenton, was wondering why you decided on vlan and not somuch vxlan | 18:40 |
| jamesdenton | plus, tenant networking requires neutron routers and the HA story there is sometimes not great. router failover can be delayed due to keepalived issues we found at the time. | 18:41 |
| jamesdenton | kinda one of those.. burn me once, shame on me type things. | 18:41 |
| broken_one | understandable | 18:42 |
| jamesdenton | i messed that quote up, but you get the point. | 18:42 |
| jamesdenton | hehe | 18:42 |
| broken_one | has it not been revisited? | 18:42 |
| broken_one | i know it was horrid prior to liberty | 18:42 |
| broken_one | m and n were suppose to make it more reliable | 18:42 |
| broken_one | with that said I do still see some issues in a pike tripleO | 18:43 |
| broken_one | but hadnt had to touch that cloud for almost a year before the issue cropped up | 18:43 |
| jamesdenton | yeah, we were having the failover issues thru mitaka and it was resolved in newer keepalived version that is/was not yet backported. and we don't like to carry changes/packages that aren't mainstream | 18:43 |
| jamesdenton | we have a range of environments from kilo thru queens, so a lot to manage and keep track of. | 18:44 |
| broken_one | so now yall are just on the , if aint broke...dont fix it trail? | 18:44 |
| noonedeadpunk | broken_one just concept: http://paste.openstack.org/show/741607/ | 18:44 |
| jamesdenton | broken_one We subscribe to that, sure. Always keeping one eye on upcoming changes/tech but needing to maintain stable environments at the same time | 18:45 |
| broken_one | noonedeadpunk: will test after this rancher 2.0 demo | 18:45 |
| jamesdenton | noonedeadpunk did that work for you? i was on the same path trying to call out the full dict | 18:45 |
| jamesdenton | "tunnel_address": "10.0.240.120 || container_networks.tunnel_address.address || ansible_host" | 18:46 |
| noonedeadpunk | jamesdenton: Error in metal_query... Correct one is metal_query: "{{ hostvars[inventory_hostname]['ansible_' ~ (hostvars[inventory_hostname] | json_query(find_bridge) | replace('-','_'))]['ipv4']['address'] }}" | 18:50 |
| noonedeadpunk | not correct, but which works) | 18:50 |
| *** vnogin has joined #openstack-ansible | 18:51 | |
| jamesdenton | gotcha. | 18:51 |
| noonedeadpunk | not sure right now how to correctly combine the old logic now. probably, via default() | 18:52 |
| noonedeadpunk | and if this doesn't suite no reason to search | 18:53 |
| jamesdenton | so default(ansible_host)? | 18:55 |
| *** vnogin has quit IRC | 18:56 | |
| jamesdenton | does the query have to change? Seems to work ok with the new metal_query (in my limited test) | 18:56 |
| *** rholloway has joined #openstack-ansible | 18:58 | |
| noonedeadpunk | jamesdenton: so you just pass this dict structure to json_query()? | 18:58 |
| *** strattao has quit IRC | 19:01 | |
| noonedeadpunk | jamesdenton I was meaning smth like this http://paste.openstack.org/show/741614/ | 19:01 |
| *** electrofelix has joined #openstack-ansible | 19:02 | |
| *** electrofelix has quit IRC | 19:04 | |
| *** strattao has joined #openstack-ansible | 19:07 | |
| noonedeadpunk | broken_one jamesdenton check this one pls http://paste.openstack.org/show/741616/ as +- final option | 19:09 |
| jamesdenton | works for is_metal, not for container | 19:12 |
| jamesdenton | fails against neutron_server container, which doesn't have a tunnel_address and should pick up ansible_host | 19:13 |
| *** rholloway has quit IRC | 19:13 | |
| noonedeadpunk | yeah... I see | 19:13 |
| openstackgerrit | Merged openstack/ansible-hardening master: Switch to using import_tasks for static inclusion https://review.openstack.org/614329 | 19:14 |
| openstackgerrit | Merged openstack/ansible-hardening master: Chrony: make ntp server options configurable https://review.openstack.org/629318 | 19:15 |
| openstackgerrit | Merged openstack/ansible-hardening master: Chrony: add an option to sync the hardware clock https://review.openstack.org/629330 | 19:15 |
| openstackgerrit | Merged openstack/ansible-hardening master: Chrony: new NTP server defaults https://review.openstack.org/629684 | 19:15 |
| noonedeadpunk | it seems, it goes to metal_query despite the fact is_metal is false... | 19:17 |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible master: Call ceph-facts role as required by changes to ceph-ansible https://review.openstack.org/629317 | 19:17 |
| jrosser | noonedeadpunk: wonder if you need a | bool in there | 19:18 |
| noonedeadpunk | I guess, that's the reason, why json_query was used - probably, as it's set in vars, ansible tries to get both variables on ternary? | 19:20 |
| noonedeadpunk | jrosser: doesn't help here... | 19:20 |
| jrosser | It would be worth looking back at how this code used to be before json_query | 19:27 |
| jrosser | There was a huge pile of inline jinja iirc | 19:27 |
| noonedeadpunk | yeah, I saw it, and it was nesty... | 19:28 |
| noonedeadpunk | it move to json_query during R cycle.. | 19:28 |
| *** cshen has quit IRC | 19:54 | |
| *** strattao has quit IRC | 19:59 | |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible master: Revert "Fixed ability of usage interfaces/bridges with dots (aliases)" due to found problem in it https://review.openstack.org/629924 | 20:11 |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible stable/rocky: Revert "Fixed ability of usage interfaces/bridges with dots (aliases)" due to found problem in it https://review.openstack.org/629925 | 20:12 |
| noonedeadpunk | jamesdenton: no ideas left, just reverting ^ | 20:29 |
| noonedeadpunk | oh, come on..... | 20:34 |
| noonedeadpunk | please try this out http://paste.openstack.org/show/741619/ | 20:34 |
| noonedeadpunk | jamesdenton broken_one ^ | 20:36 |
| *** cshen has joined #openstack-ansible | 20:37 | |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible master: Fixes empty metal_query https://review.openstack.org/629930 | 20:40 |
| *** cshen has quit IRC | 20:42 | |
| broken_one | ok checking out 741619 | 20:46 |
| jamesdenton | noonedeadpunk i just tore down that environment, but i have another. will be a few min | 20:46 |
| *** openstackgerrit has quit IRC | 20:50 | |
| jamesdenton | noonedeadpunk +1 | 20:51 |
| jamesdenton | works as advertised | 20:52 |
| noonedeadpunk | jamesdenton glad to hear. Just so stupid mistake and spent hours for it's debug... | 20:53 |
| jamesdenton | lol i know. escaped double quotes. who knew... | 20:53 |
| broken_one | looks to be working on my end as well | 20:53 |
| * jrosser despairs at centos, what does this mean? http://logs.openstack.org/17/629317/7/check/openstack-ansible-deploy-aio_lxc-centos-7/0c862e9/job-output.txt.gz#_2019-01-10_20_14_13_368127 | 20:53 | |
| cloudnull | there was an issue with the gpg key, i think ? | 20:54 |
| broken_one | noonedeadpunk: thank you for that fix | 20:54 |
| jrosser | this fix for that merged earlier i think, which is what i'm a bit wtf about that | 20:54 |
| jrosser | yes earlier in that log it pulls in the new key http://logs.openstack.org/17/629317/7/check/openstack-ansible-deploy-aio_lxc-centos-7/0c862e9/job-output.txt.gz#_2019-01-10_20_10_39_425468 | 20:55 |
| noonedeadpunk | broken_one: I'd broken, I've fixed:) So it just needs to be merged.... | 20:55 |
| broken_one | :D | 20:56 |
| broken_one | having the meeting with team in 15 mins about OVS in OSA :) | 20:59 |
| noonedeadpunk | oh, so it seemed, i was just in time:) | 21:00 |
| noonedeadpunk | but now have to go - it's already 11pm on my clock... | 21:01 |
| jrosser | more centos stuff - do we think this looks right https://github.com/openstack/openstack-ansible-galera_server/blob/master/tasks/galera_install_yum.yml#L117-L128 | 21:01 |
| jrosser | clear the repo metadata after adding the galera repo but *before* adding the percona repo | 21:02 |
| jamesdenton | thanks for knocking that out noonedeadpunk | 21:05 |
| *** SimAloo has joined #openstack-ansible | 21:06 | |
| *** radeks_ has quit IRC | 21:12 | |
| *** openstackgerrit has joined #openstack-ansible | 21:14 | |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible-galera_server master: Tidy yum repository setup https://review.openstack.org/629934 | 21:14 |
| *** radeks has joined #openstack-ansible | 21:15 | |
| *** cshen has joined #openstack-ansible | 21:18 | |
| *** jpward1981 has quit IRC | 21:21 | |
| *** cshen has quit IRC | 21:22 | |
| *** priteau has joined #openstack-ansible | 21:26 | |
| *** priteau has quit IRC | 21:35 | |
| *** ansmith_ has quit IRC | 21:40 | |
| broken_one | so the team has a few questions regarding linuxbridge and ovs for OSA first | 21:41 |
| broken_one | how many and which interfaces need to be created for linuxbridges | 21:41 |
| *** radeks has quit IRC | 21:42 | |
| broken_one | and then which interfaces will live in OVS | 21:42 |
| broken_one | if the deploy is the supported linux bridge + OVS combination | 21:43 |
| broken_one | there is some contention among what the team considers data vs control | 21:43 |
| *** cshen has joined #openstack-ansible | 21:44 | |
| jamesdenton | the terminology may be abused, but we define control as management/api/storage (i.e. non-neutron) and data as neutron traffic (vm) | 21:45 |
| *** pcaruana has quit IRC | 21:46 | |
| jamesdenton | so in a standard two-bond deploy, bond0 and related subinterfaces would be connected to linuxbridges, while bond1 would be connected to ovs provider bridge | 21:46 |
| jamesdenton | just an example | 21:46 |
| jamesdenton | now that neutron agents are installed on metal and not a container, the br-vxlan bridge isn't really bridging anything, but it does have the IP used for VTEP. | 21:47 |
| broken_one | that makes sense | 21:49 |
| broken_one | so then we have br-mgmt and br-storage on say bond0 ( linux bridges ) and we put br-vlan and br-vxlan on bond1 ( ovs ) | 21:50 |
| broken_one | that is the standard deploy OSA expects yes? | 21:50 |
| jamesdenton | more or less | 21:51 |
| *** cshen has quit IRC | 21:51 | |
| broken_one | yeah i mean not getting into vlan subinterfaces etc etc | 21:52 |
| jamesdenton | yeah, in newer releases the provider bridge gets created automatically and you can define the interface (bond1) that gets plugged in | 21:52 |
| jamesdenton | and by newer i mean master, currently. | 21:53 |
| broken_one | ok thsi will be something in the stein release of OSA? | 21:53 |
| jamesdenton | You can do OVS now w/ Newton+, there's just a little more automation there now for some things | 21:54 |
| broken_one | so we decided that | 21:54 |
| broken_one | we will do a deploy of OSA rocky in the standard supported way first | 21:55 |
| broken_one | using lb and ovs | 21:55 |
| broken_one | then we will do a deploy using pure OVS for all | 21:55 |
| jamesdenton | ++ | 21:55 |
| broken_one | and see what is the best bang for the buck | 21:55 |
| jrosser | broken_one: one of the neat things is you can integrate or separate all of the control plane functions across as many hosts as you like | 21:57 |
| jrosser | so if you had dedicated network nodes for example, there would be a very clearly defined OVS world on the computes/network nodes | 21:58 |
| jrosser | and all your infra stuff would look like linuxbridge | 21:58 |
| jrosser | so it would still be nice and tidy | 21:58 |
| broken_one | jrosser: yup that is what we plan to do | 21:58 |
| broken_one | and have hardware already racked for production with that model in mind | 21:58 |
| jrosser | my control plane is on 16 nodes - i've taken a many/small approach rather than few/large | 21:59 |
| broken_one | the control plane design is kind of over kill right now | 21:59 |
| broken_one | but its built to scale to about 10k computes | 21:59 |
| broken_one | that is about the size of our prod control plane | 22:00 |
| mnaser | broken_one: 10,000 vms or 10,000 compute nodes | 22:02 |
| broken_one | 10k computes | 22:02 |
| broken_one | 10k vm's is nothing | 22:02 |
| mnaser | what sort of environment are you building out if you can share? | 22:03 |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible-ops master: remove dynamic ns.enable generators https://review.openstack.org/629939 | 22:04 |
| broken_one | building an IaaS that will run out PaaS and also get our org out of AWS, GCE, and Azure | 22:05 |
| broken_one | s/out/our | 22:05 |
| broken_one | ultimately we are trying to remove vmware from our environment :) | 22:06 |
| mnaser | broken_one: nice | 22:08 |
| mnaser | on top of openstack? | 22:08 |
| broken_one | yeah we plan to run Rancher 2.x on top of OpenStack | 22:09 |
| jrosser | mnaser: one for you here https://review.openstack.org/#/c/629934/1 | 22:11 |
| *** priteau has joined #openstack-ansible | 22:14 | |
| *** priteau has quit IRC | 22:21 | |
| mnaser | jrosser: looking | 22:22 |
| mnaser | i see - > + and i like | 22:22 |
| mnaser | jrosser: lgtm | 22:23 |
| *** strobelight has quit IRC | 22:32 | |
| *** strattao has joined #openstack-ansible | 22:45 | |
| *** dcdamien has joined #openstack-ansible | 22:45 | |
| guilhermesp | huum looks like this continues to be relevant https://bugs.launchpad.net/openstack-ansible/+bug/1780733 | 23:03 |
| openstack | Launchpad bug 1780733 in openstack-ansible "Neutron rootwrap error" [High,Triaged] - Assigned to David Wilde (david-wilde-rackspace) | 23:03 |
| *** SimAloo has quit IRC | 23:05 | |
| *** jbadiapa has quit IRC | 23:10 | |
| *** ansmith_ has joined #openstack-ansible | 23:11 | |
| *** gyee has quit IRC | 23:11 | |
| *** strattao has quit IRC | 23:17 | |
| *** priteau has joined #openstack-ansible | 23:27 | |
| *** dcdamien has quit IRC | 23:27 | |
| *** hwoarang has quit IRC | 23:30 | |
| *** hwoarang has joined #openstack-ansible | 23:31 | |
| *** priteau has quit IRC | 23:32 | |
| *** luksky has quit IRC | 23:36 | |
| *** vicky84 has quit IRC | 23:36 | |
| *** cshen has joined #openstack-ansible | 23:47 | |
| *** cshen has quit IRC | 23:52 | |
| *** macza_ has quit IRC | 23:57 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!