13:00:54 <mnasiadka> #startmeeting kolla 13:00:55 <opendevmeet> Meeting started Wed Aug 13 13:00:54 2025 UTC and is due to finish in 60 minutes. The chair is mnasiadka. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:55 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:55 <opendevmeet> The meeting name has been set to 'kolla' 13:00:57 <mnasiadka> #topic rollcall 13:00:58 <mnasiadka> o/ 13:01:34 <bbezak> o/ 13:01:38 <amir58118> \o 13:01:57 <frickler> o/ 13:04:47 <mnasiadka> #topic agenda 13:04:47 <mnasiadka> * CI status 13:04:48 <mnasiadka> * Release tasks 13:04:48 <mnasiadka> * Regular stable releases (first meeting in a month) 13:04:48 <mnasiadka> * Current cycle planning 13:04:49 <mnasiadka> * Additional agenda (from whiteboard) 13:04:49 <mnasiadka> * Open discussion 13:04:51 <mnasiadka> #topic CI status 13:05:51 <mnasiadka> It seems the sec bug merge broke our CI, common_opts no_log=True causes getting **** in some output from kolla_container - currently there's a patch up for revert, checking if it comes out green and fast-merging it 13:06:21 <mnasiadka> Otherwise it was good yesterday I guess, so no other CI issues 13:06:31 <mnasiadka> (other than some rax.ord mirror timeouts) 13:06:37 <mnasiadka> which I already reported on #opendev 13:06:55 <frickler> that's docker.io timeouts very likely 13:07:48 <mnasiadka> Oh ok, maybe we should think about mirroring docker rpms/debs in opendev? 13:07:54 <mnasiadka> (but that doesn't show up that often I guess) 13:08:07 <opendevreview> Merged openstack/kolla stable/2024.1: Revert "mariadb: pin to 10.11.13" https://review.opendev.org/c/openstack/kolla/+/957182 13:08:22 <fungi> basically nothing on those per-provider "mirrors" is actual copies of anything aside from distro packages. all the other trees are caching proxies to external services like dockerhub or pypi 13:09:07 <fungi> so if dockerhub delays or blocks responses to the proxy, then you'll get an error client-side 13:10:59 <mnasiadka> In https://zuul.opendev.org/t/openstack/build/94be78a6d6e249cca2fb160f5eef3bac/log/primary/logs/ansible/bootstrap-servers#764 - it was during downloading gpg and docker client packages 13:11:02 <mnasiadka> so not really dockerhub 13:11:06 <mnasiadka> but I get the point 13:12:54 <mnasiadka> let's move on 13:12:58 <mnasiadka> #topic Release tasks 13:13:00 <fungi> ah, if it was rpms or debs then that might be a connectivity problem between the client and the mirror server or with the mirror server reaching the afs backend, i'll have to take a closer look when i get a moment 13:13:11 <mnasiadka> thanks fungi 13:13:29 <mnasiadka> So, we should be switching Kolla images to build from master, but we need to switch Neutron to uWSGI 13:13:40 <mnasiadka> #link https://review.opendev.org/c/openstack/kolla-ansible/+/956785 13:14:00 <mnasiadka> The CI will fail now, due to the no_log=True issue - but once we revert I'll kick the tyres again 13:14:09 <mnasiadka> frickler: would be nice if you could have a look ^^ 13:14:42 <mnasiadka> #topic Current cycle planning 13:15:14 <mnasiadka> Basically the most important thing is getting the Neutron uWSGI patch merged and then the Kolla master sources one (https://review.opendev.org/c/openstack/kolla/+/949755) 13:15:28 <mnasiadka> and then we can move back to sort of stable state with master 13:15:47 <mnasiadka> #topic Additional agenda (from whiteboard) 13:16:16 <mnasiadka> bbezak: did you have any time to have a look in Bridging the gap between community and contributing orgs? 13:16:25 <mnasiadka> I didn't, because firefighting took over my normal life ;-) 13:16:57 <bbezak> nope, last weeks I'm doing patches and reviews all day long, sorry 13:17:01 <opendevreview> Seunghun Lee proposed openstack/kolla-ansible master: Set default external Let's Encrypt cert server https://review.opendev.org/c/openstack/kolla-ansible/+/957141 13:17:09 <fungi> no worries, it was mostly just food for thought, so taking time to think about it is good ;) 13:17:33 <mmalchuk> o/ 13:17:35 <bbezak> btw - I didn't do a monthly releases as promised, because of recent "security" change. let's stabilize branches, and then I will raise release change 13:17:38 <mnasiadka> I'll try to get there next week :) 13:17:44 <fungi> i'm presenting some similar summaries to horizon later today too 13:18:01 <mnasiadka> bbezak: yup, mariadb unpins, sec thing and other things that need back ports - and then let's make releases 13:18:18 <mnasiadka> #topic Open discussion 13:18:22 <mnasiadka> Anything else from anyone? 13:18:35 <Vii> Could someone help me run the Barbican Vault test? https://review.opendev.org/c/openstack/kolla-ansible/+/935704 :) 13:20:56 <bbezak> I'm wondering if we should do openbao instead of hashicorp vault there? 13:21:10 <mnasiadka> run as in? review your code? 13:22:04 <mnasiadka> Yeah, openbao would be more open - but we should also rework the other hashicorp vault based job 13:22:09 <frickler> mnasiadka: where did the failures from https://review.opendev.org/c/openstack/kolla-ansible/+/957027 show? why did the CI not fail on that change itself? 13:22:13 <Vii> I assume that the text must work for you to accept this change 13:22:41 <mnasiadka> frickler: it seems the uuid nova_cell obfuscation doesn't show up always, but often - we were just too lucky on that one 13:22:53 <Vii> test* 13:23:00 <frickler> oh, heisenlogs, nice 13:23:16 <bbezak> :D 13:23:16 <mnasiadka> frickler: https://zuul.opendev.org/t/openstack/build/fbffac28ac74420089eb5657c4ad3975/log/primary/logs/ansible/deploy#33410 13:23:21 <Vii> but I don't know how to run it, where does the SCENARIO variable come from 13:23:37 <mnasiadka> Vii: it comes from zuul.d/base.yaml - it defined per base scenario 13:23:43 <mnasiadka> *it's 13:24:02 <Vii> I defined it but there is a test and it is skip 13:24:36 <Vii> https://zuul.opendev.org/t/openstack/build/e413d504c9cd4345ab31d57eff0d05e9/console 13:24:40 <mnasiadka> I commented, you have barbican-vault in zuul.d/base.yaml and when: scenario == 'barbican' in tests/run.yml 13:24:42 <mnasiadka> you need to decide 13:24:46 <Vii> Run test-barbican-vault.sh script - Skipped 13:24:55 <mnasiadka> Ansible doesn't read your mind yet 13:25:29 <Vii> aaaaaaa :D 13:25:31 <Vii> thx 13:25:33 <fungi> on the earlier topic of bug https://bugs.launchpad.net/kolla-ansible/+bug/2120302 since the vmt has only recently started to oversee deployment projects we're still somewhat unclear on the degree to which our existing policies and classifications are appropriate 13:25:36 <fungi> since they were originally focused on issues with api services, libraries and client tooling... one thing i'm wondering is exactly how sensitive is the ansible output from k-a expected to be? has there been a lot of work to scrub it? so you have specific policies around that? 13:26:01 <frickler> seems it does implement "no_no_logging" by doing "s/no/********/", interesting 13:26:28 <mnasiadka> fungi: I have an idea we could try to have a check if we leak passwords to any logs, it hasn't been a priority for us until now, but I can see that as being a problem 13:26:32 <fungi> i know in opendev, for example, our deployment logs of ansible output are treated as dangerous to expose, because we don't know what things random modules or playbooks might dump in there 13:27:05 <mnasiadka> Basically Ansible logs a lot of things via syslog it seems 13:27:06 <frickler> fungi: yes, I'd assume the situation for kolla-ansible to be similar 13:27:16 <opendevreview> Piotr Milewski proposed openstack/kolla-ansible master: Added vault support to barbican as back-end secret https://review.opendev.org/c/openstack/kolla-ansible/+/935704 13:27:57 <mnasiadka> Anyway, we'll keep a better eye on that now, but sadly there's also a lot of other things we need to do outside of security-aware space :) 13:28:05 <fungi> like, if nova regurgitates a password or key in its servicve log at any log level above debug, we'd consider that a vulnerability... but deployment tool output doesn't seem like it's necessarily got the same set of risks or user expectations 13:28:46 <fungi> maybe ansible output in this case is akin to debug logging? 13:29:15 <fungi> we don't treat sensitive data in debug level logs as a vulnerability that warrants an advisory, just a hardening opportunity that can be cleaned up/improved 13:29:17 <mnasiadka> well, with the problem caused by merging the no_log=True change - I found out nova-manage prints out rabbitmq and database password in the list-cells subcommand 13:29:45 <fungi> sounds like a deep rabbit hole, pun intended 13:30:15 <Vii> Yesterday I wrote about another problem I noticed, where can I write it down? 13:30:23 <Vii> There's another problem, but it requires deeper analysis and consideration. After adding a new "role" that uses Haproxy or ProxySQL, the Docker services don't restart. This results in an error like "User not in database." And you have to manually restart the Haproxy/SqlProxy service. 13:30:26 <fungi> anyway, just thinking, if the kolla docs don't already make security recommendations about safe handling of ansible output, that could be a good addition 13:30:54 <Vii> on an already running openstack - adding a new service 13:32:10 <frickler> Vii: https://bugs.launchpad.net/kolla-ansible 13:40:06 <mnasiadka> I guess we’re done 13:40:14 <mnasiadka> Thanks for coming 13:40:24 <mmalchuk> mnasiadka thanks 13:41:09 <mnasiadka> #endmeeting