| noonedeadpunk | votes on ussuri bump would be appreciated https://review.opendev.org/c/openstack/openstack-ansible/+/815589 | 08:16 |
|---|---|---|
| *** akahat|rover is now known as akahat|lunch | 08:44 | |
| jrosser_ | i don't know how we are supposed to handle the comment left on here https://review.opendev.org/c/openstack/ansible-role-python_venv_build/+/817219 | 08:53 |
| jrosser_ | using the zuul template makes it pretty hard to drop jobs | 08:53 |
| *** akahat|lunch is now known as akahat|rover | 09:13 | |
| noonedeadpunk | I don't think we really need to in this specific case | 09:16 |
| noonedeadpunk | but I think we left some nv tempestconf job somehwere... | 09:18 |
| noonedeadpunk | which is worth fixing indeed | 09:18 |
| opendevreview | likui proposed openstack/openstack-ansible-os_placement master: Replace git.openstack.org URLs with opendev.org URLs https://review.opendev.org/c/openstack/openstack-ansible-os_placement/+/817368 | 09:20 |
| opendevreview | likui proposed openstack/openstack-ansible-os_cloudkitty master: Replace git.openstack.org URLs with opendev.org URLs https://review.opendev.org/c/openstack/openstack-ansible-os_cloudkitty/+/817369 | 09:24 |
| opendevreview | likui proposed openstack/openstack-ansible-os_heat master: Replace git.openstack.org URLs with opendev.org URLs https://review.opendev.org/c/openstack/openstack-ansible-os_heat/+/817370 | 09:26 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-tests master: Bump ansible and collection versions https://review.opendev.org/c/openstack/openstack-ansible-tests/+/812684 | 09:54 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Bump ansible and collection versions https://review.opendev.org/c/openstack/openstack-ansible/+/817375 | 09:57 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-rabbitmq_server master: Update rabbitmq version https://review.opendev.org/c/openstack/openstack-ansible-rabbitmq_server/+/817380 | 10:06 |
| *** melwitt is now known as Guest5508 | 10:12 | |
| jonher | hey, i have a wallaby test deployment that's been deployed since about the time it was released. it looks like my gnocchi is not happy though https://paste.openstack.org/show/810902/ it keeps on crashing on controller002 and controller003 where is seems to have what i'm guessing datafiles (gnocchii0000xxxx etc..) while controller001 only has 2 config files | 10:20 |
| jonher | i checked the commits on os_gnocchi but couldn't see anything newish that relates to this, is this a known issue of some sorts? | 10:21 |
| noonedeadpunk | tbh I haven't used gnocchi on Wallaby. And where I do use it - it's integrated with ceph wither by s3 or directly with rbd | 10:22 |
| noonedeadpunk | and yeah - we didn't do any changes lately, but we use gnocchi 4.4 version for quite a while | 10:22 |
| noonedeadpunk | So in my deployments I don't have /var/lib/gnocchi/tmp at all ( | 10:23 |
| noonedeadpunk | eventually I wanted to check gnocchi for quite a while now because I saw other weird things going on in the role when integrated with ceph as well, but never had time for that | 10:25 |
| noonedeadpunk | jonher: oh https://bugs.launchpad.net/openstack-ansible/+bug/1915618 | 10:26 |
| noonedeadpunk | jonher: so probably worth trying to comment out this condition https://opendev.org/openstack/openstack-ansible-os_gnocchi/src/branch/master/tasks/main.yml#L158 | 10:28 |
| noonedeadpunk | but ideally I guess would be to check if storage backend is file and apply this condition only when it's not | 10:28 |
| noonedeadpunk | would be great if you could try to test this out actually | 10:29 |
| jonher | i can try that on my current deployment, i would probably want to go the ceph route myself later | 10:29 |
| noonedeadpunk | for ceph there should be another change made. I believe that we don't need this https://opendev.org/openstack/openstack-ansible-os_gnocchi/src/branch/master/vars/main.yml#L29-L30 | 10:31 |
| noonedeadpunk | since when you add ceph_alternative, setup.cfg of gnocchi tries to install python-librados, which is not available on pip | 10:31 |
| noonedeadpunk | and we anyway workaround this with symlinking ceph libraries inside venv | 10:31 |
| noonedeadpunk | but again - would be great to test this) | 10:32 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-galera_server master: Update mariadb to 10.6.5 https://review.opendev.org/c/openstack/openstack-ansible-galera_server/+/817384 | 10:34 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: [DNM] Check out galera 10.6.5 https://review.opendev.org/c/openstack/openstack-ansible/+/817385 | 10:35 |
| noonedeadpunk | would be great to merge https://review.opendev.org/c/openstack/openstack-ansible-tests/+/812684 as well as it will unblock left functional tests | 10:56 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-os_keystone master: Drop Nginx webserver support https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/817390 | 11:54 |
| admin1 | hi all . .haproxy_ssl_self_signed_regen: false and haproxy_user_ssl_cert and _key is set .. but the playbook still wants to generate the cert .. has something changed/updated in the new version ? | 13:17 |
| admin1 | hmm.. it worked again in the 2nd try | 13:20 |
| admin1 | ¯\_(ツ)_/¯ | 13:20 |
| mgariepy | morning everyone | 13:21 |
| damiandabrowski[m] | hey! | 13:25 |
| noonedeadpunk | admin1: I believe we didn't change there anything. but regen is broken in terms that even if you set to to true, it won't really regen | 13:28 |
| mgariepy | if you need some patch that need some attention please let me know. | 13:30 |
| noonedeadpunk | Hm, I wonder why patches does not go to gates... | 13:31 |
| mgariepy | that is indeed weird. | 13:39 |
| noonedeadpunk | feels like zuul has hang now... | 13:39 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-rabbitmq_server master: Update rabbitmq version https://review.opendev.org/c/openstack/openstack-ansible-rabbitmq_server/+/817380 | 13:43 |
| mgariepy | noonedeadpunk, did you poke infra about zuul ? | 14:19 |
| jonher | yes, #opendev | 14:26 |
| mgariepy | channel changes i will probably never get used to that lol. | 14:26 |
| jonher | noonedeadpunk: removing https://opendev.org/openstack/openstack-ansible-os_gnocchi/src/branch/master/tasks/main.yml#L158 did indeed put the gnocchi-config and gnocchi-storage-config on all nodes as expected, making them start and run | 14:45 |
| opendevreview | James Gibson proposed openstack/openstack-ansible-os_nova master: Enable TLS for VNC from novncproxy to compute hosts https://review.opendev.org/c/openstack/openstack-ansible-os_nova/+/817222 | 14:51 |
| jonher | so if i understand you correctly, in the scenario of gnocchi_storage_driver: 'file' it should run the db sync on all nodes, but in a scenario where it's not 'file' you'd want to run it only on "_gnocchi_is_first_play_host" because it would bootsrap on the common area? | 14:53 |
| noonedeadpunk | jonher: if I understood kolla bug correctly then yes it might be the case | 15:03 |
| noonedeadpunk | because for ceph I didn't need to run more then once | 15:04 |
| noonedeadpunk | so some ternary could fit there | 15:04 |
| spatel | noonedeadpunk jrosser_ https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/814825 | 15:09 |
| spatel | i agreed we can improve code for more variables but i think lets go with this at present and in future we will add more methods to do bonding | 15:09 |
| spatel | problem is i don't have lab with multiple nic card where i can test and certify code | 15:10 |
| spatel | This patch will keep things moving at present if someone looking to bond dpdk | 15:10 |
| spatel | in January i am building new cloud in Singapore during that time i will have lots of new hardware to test out more feature of DPDK | 15:11 |
| noonedeadpunk | spatel: just realized it missing at least a release note :) | 15:12 |
| spatel | oh! can you point me to doc about how to do release notes.. its been a while i did that :) | 15:13 |
| noonedeadpunk | https://docs.openstack.org/reno/latest/user/usage.html | 15:14 |
| noonedeadpunk | btw I can hardly get how several interfaces could be made here... | 15:14 |
| noonedeadpunk | well, with extra variable probably... | 15:14 |
| jrosser_ | spatel: the trouble is that you introduce new variables in defaults/main.yml, so this becomes the 'interface' to the os_neutron role | 15:16 |
| jrosser_ | it's probably more that the definition of the bonds needs to be supplied as a list, or something, rather than just one | 15:16 |
| jrosser_ | jamesdenton: ^ | 15:16 |
| noonedeadpunk | I wonder how to loop over it then | 15:17 |
| noonedeadpunk | maybe like with_subelements... | 15:17 |
| noonedeadpunk | nah, it won't work.. | 15:17 |
| noonedeadpunk | the only way I see this working to make dynamic inlcude | 15:18 |
| jrosser_ | like i don't really understand straight away what this already loops over https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/814825/2/tasks/providers/setup_ovs_dpdk.yml | 15:18 |
| spatel | jrosser_ this is what i did, took this list which defined by operator and just tell this code to bond them - https://paste.opendev.org/show/810911/ | 15:19 |
| noonedeadpunk | it loops over mapping defined in provider_networks | 15:19 |
| noonedeadpunk | ie https://opendev.org/openstack/openstack-ansible-os_neutron/src/branch/master/defaults/main.yml#L377 | 15:19 |
| spatel | i agreed if we have more provider network then this code may fail.. that part i didn't check yet | 15:20 |
| spatel | This is extracting provider network bridge - bridge: "{{ bridge_mapping.split(':')[1] }}" | 15:21 |
| noonedeadpunk | spatel: I believe there should be some variable defined like that https://paste.opendev.org/show/810912/ | 15:21 |
| noonedeadpunk | or https://paste.opendev.org/show/810914/ | 15:22 |
| noonedeadpunk | and `bridge` I believe should be another key? | 15:23 |
| spatel | hmm now my brain stopped work.. | 15:25 |
| spatel | noonedeadpunk why do you want to define - ports: "{{ ovs_dpdk_pci_addresses }}" | 15:25 |
| noonedeadpunk | I mean - this way you can define mutiple aggregates? | 15:28 |
| spatel | let me understand, we are trying to make this code more flexible so if someone has multiple provider it should work in that scenario right? example. i have 4 port nic - A, B - bond0 and C,D - bond1 and bond0 is provider-0 and bond1 is provide-1 | 15:28 |
| noonedeadpunk | So you would need to define just structure. and if you iterate over ovs_dpdk_lacp_bond - each element would have all data to create an aggregate | 15:29 |
| noonedeadpunk | spatel: yes, correct | 15:29 |
| spatel | noonedeadpunk i am following you.. | 15:29 |
| jrosser_ | you can test this by just defining one key in the dict or one list element | 15:30 |
| jrosser_ | if you only have one bond | 15:30 |
| noonedeadpunk | I dunno what's better list or dict - not sure it really matters as well | 15:31 |
| noonedeadpunk | I'm not sure yet how to process mapping in automated way though - you might need to define these as keys explicitly... | 15:31 |
| spatel | i believe automated way will make thing complicate.. i would prefer define keys and lets this code suck them in.. | 15:36 |
| spatel | DPDK is not common for all deployment so i don't think anyone care about more automation | 15:36 |
| noonedeadpunk | I agree here | 15:36 |
| noonedeadpunk | I think jsut defining keys is best option here | 15:36 |
| spatel | I am having hard time to deploy.. :( look like i should start looking at XDP | 15:37 |
| * noonedeadpunk no idea what that is | 15:37 | |
| spatel | https://docs.openvswitch.org/en/latest/intro/install/afxdp/ | 15:38 |
| spatel | XDP is alternative way of DPDK. (XDP doesn't need any fancy config like hugepage, CPU isolation etc.. which DPDK required) | 15:39 |
| spatel | i believe soon XDP will replace DPDK | 15:40 |
| noonedeadpunk | well concept looks better then dpdk tbh | 15:40 |
| spatel | it provide comparable performance to DPDK | 15:40 |
| noonedeadpunk | not sure about real performance though... | 15:40 |
| noonedeadpunk | well then it's promising | 15:40 |
| spatel | it does provide equal performance and beauty is it does support all kernel feature like container networking, namespace etc.. | 15:42 |
| spatel | DPDK doesn't support namespace that is why you can't gain performance if you deploy vRouter in namespace :( | 15:43 |
| spatel | https://www.youtube.com/watch?v=1oAsRzrwAAw&ab_channel=OpenInfrastructureFoundation | 15:44 |
| *** akahat|rover is now known as akahat|lunch | 16:03 | |
| *** akahat|lunch is now known as akahat|dinner | 16:03 | |
| spatel | jrosser_ planning to add this snippet in default/main.yml - https://paste.opendev.org/show/810916/ | 16:06 |
| jrosser_ | awesone, thats now extensible | 16:07 |
| spatel | This way we can have two bond with different config if someone looking to do this way.. | 16:07 |
| spatel | let me commit this change | 16:09 |
| mgariepy | wow.. how long does zuul didn't check the +w patches ? | 16:20 |
| mgariepy | https://review.opendev.org/c/openstack/openstack-ansible-os_ceilometer/+/815317 | 16:22 |
| mgariepy | this one didn't merge for some reason since nov. 2nd. | 16:23 |
| noonedeadpunk | I guess there might be somehow restarted then... | 16:23 |
| noonedeadpunk | but yeah we need to take action on tons of stuff... | 16:24 |
| mgariepy | do you have a dashboard with patches that are not merged ? | 16:24 |
| mgariepy | but that should be ? | 16:24 |
| noonedeadpunk | nah I think I didn't create that (ie patches that have +w and not merged yet) | 16:25 |
| opendevreview | Jonathan Herlin proposed openstack/openstack-ansible-os_gnocchi master: Fix gnocchi multi-node when storage_driver is file https://review.opendev.org/c/openstack/openstack-ansible-os_gnocchi/+/817457 | 16:46 |
| *** akahat|dinner is now known as akahat|rover | 16:52 | |
| opendevreview | Satish Patel proposed openstack/openstack-ansible-os_neutron master: Add support of dpdk bonding https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/814825 | 16:56 |
| opendevreview | Satish Patel proposed openstack/openstack-ansible-os_neutron master: Add support of dpdk bonding https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/814825 | 17:01 |
| opendevreview | Merged openstack/openstack-ansible-tests master: Bump ansible and collection versions https://review.opendev.org/c/openstack/openstack-ansible-tests/+/812684 | 17:03 |
| spatel | jamesdenton take a look at this patch when you have time - https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/814825 | 17:05 |
| jamesdenton | hi spatel - i will, thank you | 17:05 |
| jamesdenton | i recall talking about this a couple of weeks ago | 17:05 |
| opendevreview | Satish Patel proposed openstack/openstack-ansible-os_neutron master: Add support of dpdk bonding https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/814825 | 17:05 |
| jamesdenton | the problem with using ovs_dpdk_pci_addresses for the bond members, is that ovs_dpdk_pci_addresses is a list of all qualified dpdk interfaces. you may only want a subset of those for a given bond | 17:06 |
| jamesdenton | i also don't think i'd worry about a var to enable bond support, just build the bond if the list > 0 | 17:09 |
| jamesdenton | and to further simplify it, it may be better to have the bonding setup as an extension to the existing provider network logic; that already has a mechanism for plugging a network_interface in to the bridge. There can be something there like.. bond_interface: [eno1, eno2] or something | 17:15 |
| opendevreview | Satish Patel proposed openstack/openstack-ansible-os_neutron master: Add support of dpdk bonding https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/814825 | 17:15 |
| spatel | jamesdenton hmm this is good point related ovs_dpdk_pci_addresses for the bond members.. | 17:18 |
| jamesdenton | https://github.com/openstack/openstack-ansible-os_neutron/blob/ea9dae41ad61195ae25f7673d016bcf2486204c8/tasks/providers/ovs_config.yml#L35 | 17:20 |
| opendevreview | OpenStack Proposal Bot proposed openstack/openstack-ansible-os_murano master: Updated from OpenStack Ansible Tests https://review.opendev.org/c/openstack/openstack-ansible-os_murano/+/745856 | 17:20 |
| jamesdenton | that mechanisn uses the provider library to say "br-provider: eno1" or "br-provider:bond1" - it uses a pre-existing interface | 17:20 |
| jamesdenton | but you could extend that to be a "bond name: members" and keep it all within the provider network definition | 17:21 |
| jamesdenton | https://github.com/openstack/openstack-ansible-os_neutron/blob/ea9dae41ad61195ae25f7673d016bcf2486204c8/library/provider_networks | 17:21 |
| jamesdenton | https://github.com/openstack/openstack-ansible-os_neutron/blob/master/library/provider_networks | 17:21 |
| spatel | reason i am planning to use static list so we have less complexity in code and let operator decide what they want to do.. | 17:25 |
| jamesdenton | well in this case you still would be allowing the operator to define it, but all in one place | 17:26 |
| spatel | could you please make comments in patch so make it easy to understand | 17:27 |
| jamesdenton | for example, defining provider networks is required, but defining network_interface is not. when it's there, the playbooks add the interface to the corresponding ovs bridge. you could say 'dpdk_bond_members' is a list of members, and when defined will create the dpdk bond on a "per provider network" basis | 17:27 |
| jamesdenton | sure | 17:27 |
| opendevreview | Merged openstack/openstack-ansible stable/ussuri: Bump OpenStack-Ansible Ussuri https://review.opendev.org/c/openstack/openstack-ansible/+/815589 | 18:49 |
| opendevreview | Merged openstack/openstack-ansible master: Remove reference to deprecated python-ceilometerclient https://review.opendev.org/c/openstack/openstack-ansible/+/815315 | 18:49 |
| opendevreview | Merged openstack/ansible-role-python_venv_build stable/train: Set centos-7 jobs to non voting https://review.opendev.org/c/openstack/ansible-role-python_venv_build/+/817219 | 20:00 |
| opendevreview | Merged openstack/ansible-role-python_venv_build stable/train: Workaround distro provided pip having old CA certs on centos-7 https://review.opendev.org/c/openstack/ansible-role-python_venv_build/+/816473 | 22:31 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!