| noonedeadpunk | f0o: oh, yes, that is totally an issue in docs, great catch | 07:11 |
|---|---|---|
| f0o | :) | 07:11 |
| f0o | What a ride :D | 07:11 |
| noonedeadpunk | so distributed fips does work with that? | 07:12 |
| f0o | yup! | 07:12 |
| f0o | there is a period of downtime where the neutron periodic worker hasnt updated reside-on-redirect-chassis option on the LRPs but you can set that yourself in the NB DB to speed it up | 07:13 |
| f0o | ultimately neutron_ovn_distributed_fip was the big setting that I had to set correctly | 07:13 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-os_neutron master: [doc] Use neutron_ovn_distributed_fip instead of the override https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/945944 | 07:16 |
| noonedeadpunk | and then for the overall connectivity you had to change the underlying network name in NB DB? | 07:16 |
| noonedeadpunk | as eventually I'm surprised why it's not updated during sync if neutron contains correct record in it's DB | 07:17 |
| f0o | Yep that was a manual change | 07:17 |
| f0o | identify the provnet-* entries and alter the network_name option to match the new provider (fip->vlan for me) | 07:18 |
| noonedeadpunk | and did you have to update ovn-bridge-mappings as well? | 07:18 |
| f0o | no that's taken care of by ansible after I changed the settings in the mappings variable | 07:18 |
| noonedeadpunk | ++ yeah, as I think it was in code... | 07:19 |
| f0o | but what I needed was: ovn-nbctl lsp-list neutron-98b47402-707d-49a6-9242-91ce389c2ea1 -> Output will have one entry called `provnet-.*` -> ovn-nbctl lsp-get-type provnet-c17a93b2-5e38-4441-a0d7-99bc546b7994 -> verify it's of type `localnet` -> ovn-nbctl lsp-set-options provnet-c17a93b2-5e38-4441-a0d7-99bc546b7994 network_name=vlan -> 'Change from old fip provider to new | 07:20 |
| f0o | vlan provider' | 07:20 |
| noonedeadpunk | I kinda wonder if there's somewhere in mysql for neutron that still keeps reference to old network names for the network | 07:20 |
| f0o | all flows moved almost instantly | 07:20 |
| f0o | Yes in networksegments table | 07:20 |
| noonedeadpunk | as I'd really expect this thing to be done with migration script | 07:20 |
| f0o | so that was the last step for me to "correct" the DB in that table | 07:20 |
| f0o | oddly the migration script did absolutely nothing | 07:21 |
| noonedeadpunk | but you did change mysqwl before or after running it? | 07:21 |
| noonedeadpunk | as I guess before? | 07:21 |
| f0o | it printed the same ports/options/networks needed updating no matter in what order I ran it | 07:21 |
| f0o | no I always did SQL first because I assumed the migration script would aid here but it really didnt matter | 07:21 |
| noonedeadpunk | yeah, exactly... | 07:21 |
| noonedeadpunk | it's really weird.... | 07:21 |
| noonedeadpunk | as I used it once or twice when scrapped raft cluster for nb/sb db | 07:22 |
| f0o | even now I can run log/repair/migrate and it will all print the same output | 07:22 |
| noonedeadpunk | so I basically re-created ovn containers and the scrip[t properly populated it with data afterwards | 07:22 |
| f0o | same entries "need updating" and on repair/migrate it claims it does it but nothing changes, you rerun it immediately and it just complains about the exact same things | 07:23 |
| noonedeadpunk | so maybe it has sdome checks to prevent some actions if elements exists... still cery weird | 07:23 |
| f0o | maybe | 07:23 |
| f0o | ultimately Gemini DeepResearch gave me that network_name pointer that I needed :D | 07:24 |
| noonedeadpunk | f0o: and did it help you with your performance botlleneck after all? | 08:50 |
| f0o | noonedeadpunk: a bit early to answer but subjectively it does seem that way. the GW nodes are idle now and the Compute node's extra load seems acceptable. Right now I see about 1Gbit/s more traffic than before based on the past few hours | 08:59 |
| f0o | will have to observe it over the course of the week to compare it to previous week's traffic levels/patterns to get any conclusive data | 08:59 |
| f0o | not much going on Sunday->Monday ;) | 08:59 |
| noonedeadpunk | ok, so if not gw nodes were acting as core routers - then indeed they won't be needed at all with distributed fips | 09:00 |
| noonedeadpunk | fair enough | 09:00 |
| f0o | pretty much | 09:00 |
| f0o | latency/jitter has definetly gone down now that the CPU cycles of the core-routers are no longer pegged by vwsitchd | 09:00 |
| f0o | this is so far the only big difference | 09:00 |
| noonedeadpunk | but also - I think if you create a RBAC rule to access your public network as shared one - you will be able to connect it directly to VMs now I assume? | 09:01 |
| f0o | That's true | 09:01 |
| f0o | it does make subnetting a bit more tedious | 09:02 |
| f0o | will have to calculate now how big of a subnet I can allocate per rack to not stretch more l2 - but I guess that's where ovn-bgp-agent would help out in the end | 09:03 |
| noonedeadpunk | it depends I guess? | 09:03 |
| noonedeadpunk | ah, from this aspect... | 09:04 |
| noonedeadpunk | yeah | 09:04 |
| opendevreview | Merged openstack/openstack-ansible-plugins master: Stop running functional tests for the repo https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/944929 | 10:54 |
| opendevreview | Merged openstack/openstack-ansible master: Ensure that cinderv3 service type is preserved on upgrade https://review.opendev.org/c/openstack/openstack-ansible/+/942580 | 11:05 |
| opendevreview | Merged openstack/openstack-ansible-plugins master: Update ssh connection plugin paramters to match upstream https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/939899 | 11:11 |
| opendevreview | Merged openstack/openstack-ansible-os_cinder master: Align on cinder service naming https://review.opendev.org/c/openstack/openstack-ansible-os_cinder/+/942581 | 13:04 |
| opendevreview | Merged openstack/openstack-ansible-os_tempest master: Switch volume catalog_type to block-storage https://review.opendev.org/c/openstack/openstack-ansible-os_tempest/+/942775 | 16:57 |
| noonedeadpunk | another vote on https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/944922/2 will unblock a bunch of follow-ups :) | 17:32 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Stop skipping yaml-related rules https://review.opendev.org/c/openstack/openstack-ansible/+/942686 | 17:33 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Stop reffering to tests repo for ansible requirements https://review.opendev.org/c/openstack/openstack-ansible/+/945982 | 17:46 |
| jrosser | ah we really need to have another push on the molecule stuff | 17:47 |
| jrosser | would be fantastic to get that all in this cycle | 17:47 |
| noonedeadpunk | I wonder what we have left.... | 17:48 |
| noonedeadpunk | lxc testing itself? | 17:48 |
| noonedeadpunk | and actual connection plugin testing? | 17:49 |
| noonedeadpunk | or well | 17:49 |
| noonedeadpunk | we have planty of things to do with m,olecule :D | 17:49 |
| noonedeadpunk | as we've jsut scraped the surface | 17:49 |
| jrosser | well - i started kind of at lxc_hosts/lxc_container_create and everything spiralled down to the horror with centos in the connection plugin | 17:53 |
| jrosser | kind of here https://review.opendev.org/c/openstack/openstack-ansible-lxc_container_create/+/939257 | 17:53 |
| jrosser | i think that there are some useful sanity/regression tests in there from the old tests repo | 17:55 |
| noonedeadpunk | yeah, I found old functional tests kinda useful and still relevant with some love | 17:55 |
| noonedeadpunk | but also that's is smth what we don't have for a while? | 17:55 |
| jrosser | perhaps yes | 17:56 |
| noonedeadpunk | meaning - we've disabled functional tests there some time ago | 17:56 |
| jrosser | yeah | 17:56 |
| noonedeadpunk | so from prespecitve of the goal - cover what is still covered we're pretty much there | 17:56 |
| jrosser | so we must just be careful with cleanup, to migrate to molecule rather than just delete things | 17:56 |
| jrosser | yes agreed | 17:56 |
| noonedeadpunk | and then ofc, there's plenty of things tbd - like scenartios for haproxy | 17:57 |
| jrosser | right, PKI too would be great to do | 17:57 |
| jrosser | and so on | 17:57 |
| noonedeadpunk | I think PKI is partially covered? | 17:57 |
| noonedeadpunk | https://opendev.org/openstack/ansible-role-pki/src/branch/master/molecule/default/verify.yml | 17:58 |
| noonedeadpunk | at least - that things are created and are somehow valid | 17:58 |
| noonedeadpunk | but sure - there's room for improvement :) | 18:00 |
| opendevreview | Merged openstack/openstack-ansible-os_trove master: Switch volume catalog_type to block-storage https://review.opendev.org/c/openstack/openstack-ansible-os_trove/+/945477 | 18:07 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Move argument parsing for dynamic_inventory to generate https://review.opendev.org/c/openstack/openstack-ansible/+/945025 | 18:38 |
| noonedeadpunk | btw, it seems we indeed to have dynamic_inventory broken | 18:38 |
| noonedeadpunk | https://bugs.launchpad.net/openstack-ansible/+bug/2102132 | 18:39 |
| noonedeadpunk | and our unit tests for it show pretty much same picture' | 18:39 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Move argument parsing for dynamic_inventory to generate https://review.opendev.org/c/openstack/openstack-ansible/+/945025 | 18:43 |
| noonedeadpunk | or well... disregad I guess... | 18:45 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Move argument parsing for dynamic_inventory to generate https://review.opendev.org/c/openstack/openstack-ansible/+/945025 | 19:11 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Install inventory-manage as a console_script https://review.opendev.org/c/openstack/openstack-ansible/+/945032 | 19:13 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Symlink console_scripts to /usr/local/bin/ https://review.opendev.org/c/openstack/openstack-ansible/+/945033 | 19:13 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Move argument parsing for dynamic_inventory to generate https://review.opendev.org/c/openstack/openstack-ansible/+/945025 | 19:58 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Install inventory-manage as a console_script https://review.opendev.org/c/openstack/openstack-ansible/+/945032 | 19:59 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Symlink console_scripts to /usr/local/bin/ https://review.opendev.org/c/openstack/openstack-ansible/+/945033 | 19:59 |
| jrosser | how do we have a inventory test fail and still manage to merge things….. | 20:21 |
| opendevreview | Merged openstack/openstack-ansible-plugins master: Fix ansible linting for functional tests https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/944922 | 21:16 |
| opendevreview | Merged openstack/openstack-ansible-plugins master: Execute functional tests through molecule https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/944923 | 21:16 |
| opendevreview | Merged openstack/openstack-ansible-plugins master: Remove tagfilter strategy from the repo https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/944926 | 21:16 |
| opendevreview | Merged openstack/openstack-ansible-plugins master: Cleanup unnecessary files/templates from tests https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/944927 | 21:16 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!