anskiy | noonedeadpunk: https://bugs.launchpad.net/openstack-ansible/+bug/2023370 | 08:35 |
---|---|---|
*** cloudnull8 is now known as cloudnull | 09:18 | |
*** cloudnull1 is now known as cloudnull | 10:42 | |
noonedeadpunk | anskiy: thanks! next week I will try to talk to nova folks on the summit regarding it, as I assume it might be an issue in their code and it might be more beneficial to fix it there | 10:51 |
anskiy | noonedeadpunk: I've managed to reproduce it in lxc aio 25.2.0, updated report. | 10:53 |
noonedeadpunk | I'd assume you should be able to reproduce that anywhere... | 10:54 |
noonedeadpunk | in the meanwhile, as a workaround, you can run playbook with `--skip-tags common-service` | 10:54 |
anskiy | at least, in newer versions Nova continues to work. On production I just commented out `update_password: always` for now :) | 10:56 |
anskiy | I wonder if running nova-api-os-compute with `nova_wsgi_threads: 16`, `nova_wsgi_processes_max: 1` `nova_wsgi_processes: 1` is safe thing to do | 10:56 |
noonedeadpunk | anskiy: well ,we can add variable for update_password I guess. As it was added like that just to ensure password is valid, as it wasn't breaking anything at times it was added | 11:53 |
noonedeadpunk | so that's another option indeed | 11:54 |
anskiy | noonedeadpunk: this sound a little bit strange to me, as I don't understand reasoning behind always ensuring the password is valid, except if osa supports changing passwords, I guess... | 12:49 |
noonedeadpunk | yeah, that's the idea behind that. If you change password in user_secrets, it will get updated with quite minimal downtime, as other backends are disabled in haproxy, and when running with a tag <service>-config it's not that long from password update to service being re-enabled in haproxy | 12:51 |
noonedeadpunk | though we can improve that as well... | 12:52 |
noonedeadpunk | (and document) | 12:52 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!