19:00:02 <fungi> #startmeeting infra 19:00:02 <opendevmeet> Meeting started Tue Jun 17 19:00:02 2025 UTC and is due to finish in 60 minutes. The chair is fungi. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:00:02 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:00:02 <opendevmeet> The meeting name has been set to 'infra' 19:01:30 <fungi> #link https://lists.opendev.org/archives/list/service-discuss@lists.opendev.org/thread/SCQGMDGA25UKFCKPKBO2RFTJRETB5XFT/ Our Agenda 19:01:35 <fungi> sorry, took me a sec 19:01:43 <fungi> #topic Announcements 19:05:48 <fungi> #link https://cfp.openstack.org/app/2025summit/all-plans/36/presentations/new/summary The OpenInfra Summit Europe 2025 call for Forum session proposals is open until July 8 19:06:10 <fungi> i clearly should have prepared announcements in advance, that url was a lot more work to find than i anticipated 19:06:37 <fungi> anyway, if anyone wants to propose an opendev-related forum session, now's the time to figure that out 19:07:02 <fungi> and remember, the gerrit folks will be there too so this is an opportunity for some cross-project work there as well 19:07:17 <fungi> any other announcements? 19:08:03 <fungi> #topic Zuul-launcher image builds (corvus 20240910) 19:08:23 <fungi> i know there's a stack of changes pending for label switches in nodesets 19:08:28 <corvus> couple of zuul-launcher things: 19:08:34 <corvus> 1) i have a change up to move the openstack tenant (and others) to use zuul-launcher nodes https://review.opendev.org/952712 19:08:38 <corvus> this would affect some, but not all jobs 19:08:43 <corvus> 2) i have a series of changes to move even more nodesets over to zuul-launcher for the tenants that use it. that series *ends* at https://review.opendev.org/952726 19:09:00 <corvus> we could merge the change to switch openstack over, then merge the changes in the series to use progressively more zuul-launcher nodes as we gain confidence 19:09:01 <corvus> are we ready for that? 19:09:07 <corvus> frickler noted that we can't autohold zuul-launcher nodes yet. i can probably get that implemented this week so it's in saturday's restart. should we wait on that or proceed without it in order to ramp up our load testing? 19:10:02 <frickler> I wouldn't see that as blocker if the expected timeframe is that short 19:10:23 <fungi> yeah, i think if we tell people we can't autohold their job for now it's not the end of the world 19:11:41 <corvus_> i think my homeserver just decided to do something else... 19:11:43 <fungi> i'm on board with approving 952712 after the meeting today, or first thing tomorrow depending on what others are comfortable with 19:11:50 <corvus_> yeah, i think that's reasonable. and i think it's worth getting more data now 19:11:53 <corvus_> 12:11 19:11:58 <corvus_> so i'll plan on switching over the openstack tenant 19:12:05 <corvus_> at the end of the series that moves over nodesets, you can see there are only two images we're missing in zuul-launcher: ubuntu-bionic and ubuntu-focal 19:12:10 <corvus_> should we send out a "last call" for folks to port them? or does someone here want to do it? i don't expect it to be very hard. 19:12:48 <fungi> i think we (opendev) are still using focal nodes for at least some tests right? 19:13:06 <fungi> things that haven't been moved to jammy/noble yet 19:13:49 <fungi> i'm hesitant to commit to finding time to do the focal label migration myself at this time though 19:14:16 <fungi> bionic i'm tempted to just let openstack folks take care of or we drop it 19:14:36 <corvus_> ok. i'm sure one of us opendev folks will end up doing focal if no one else does. 19:14:56 <corvus_> it's looking highly likely that bionic may just not get done unless there's a motivated openstack-focused person 19:15:00 <fungi> or we take it as a sign to replace the lingering focal servers and move testing to later ubuntu ;) 19:15:25 <fungi> which would be better for us overall, but maybe more work in the near term 19:15:57 <corvus_> let's let it sit out there for a while, and when shutting off nodepool is in sight, send out one last notice. sound good? 19:15:58 <fungi> i think when i looked, all the openstack releases still under maintenance were tested on jammy and newer 19:16:25 <fungi> so even openstack may just say people who care about unmaintained branches should either step up asap or drop testing 19:18:14 <fungi> yeah, sounds good to me. any other feedback on this topic? 19:18:42 <frickler> how bad would it be to keep those labels running on nodepool for some time? 19:19:03 <fungi> it would mean continuing to maintain nodepool and run the servers for it 19:19:53 <corvus_> i think the effort to maintain nodepool for one label that no one uses is not worth doing 19:20:19 <fungi> i'm with you on that 19:20:35 <corvus> so i'll plan on switching over the openstack tenant 19:20:38 <fungi> i think the effort to maintain anything no one uses (or at least cares enough about to help with) is not worth it 19:20:46 <frickler> can we gather data how often a label is getting used? 19:20:54 <corvus_> (hey the homeserver is back, sorry about the delayed/repeat message ^) 19:21:27 <frickler> I'm just arguing that we don't need to switch things off. if they break somehow, that's another story 19:21:30 <corvus_> it's probably like 15 minutes of work for someone to port over the image, if it's important. 19:22:12 <fungi> but also a good opportunity to stop spending cycles and wasting space/bandwidth on an image that isn't being used, if it's not being used that is 19:22:13 <corvus_> i very much want to switch it off when we get there. i'd like to recover the resources, and start to see what zuul looks like without nodepool. note that they don't cooperate on things like quota... 19:23:35 <fungi> so i guess to repeat/restate frickler's earlier question, what's the best way to figure out how much or often a given label is requested? 19:23:46 <corvus_> should show up in the launcher logs 19:23:46 <fungi> (or how recently) 19:24:08 <fungi> we have ~2 weeks retention on those? or am i remembering wrong? 19:25:16 <corvus> maybe 10 days it looks like? 19:25:33 <fungi> i concur 19:25:35 <fungi> just checked 19:25:52 <fungi> so we could say a label not requested in 10 days is probably not that important 19:26:06 <fungi> and someone could also port it later if that turns out to be an incorrect assumption 19:27:06 <fungi> anything else on this topic? 19:27:14 <corvus> nope 19:27:20 <corvus> thanks! 19:27:28 <fungi> #topic Gerrit shutdown problems (clarkb 20250527) 19:27:40 <fungi> i don't think there's anything new on this, we can revisit next week 19:27:49 <fungi> #topic Gerrit 3.11 Upgrade Planning (clarkb 20250401) 19:28:52 <fungi> #link https://www.gerritcodereview.com/3.11.html Please check this for any concerns with the way we use Gerrit 19:29:03 <fungi> 104.130.253.194 is a held Gerrit 3.11 node for testing purposes 19:29:17 <fungi> #link https://review.opendev.org/c/opendev/system-config/+/882900 Host Gerrit images on quay.io 19:29:42 <fungi> and its parent change 19:30:00 <fungi> #link https://etherpad.opendev.org/p/gerrit-upgrade-3.11 Planning Document for the eventual Upgrade 19:30:23 <fungi> no changes since last week, but all good reminders as we should move forward on this soon 19:30:47 <fungi> #topic Upgrading Old Servers (clarkb 20230627) 19:31:00 <fungi> #link https://etherpad.opendev.org/p/opendev-bionic-server-upgrades 19:31:11 <fungi> #link https://etherpad.opendev.org/p/opendev-focal-server-upgrades 19:31:35 <fungi> those are somewhat relevant to the earlier zuul-launcher discussion as well 19:32:16 <fungi> i haven't seen tonyb so unsure what the wiki (and cacti) situation is, presumably the same as last week 19:32:44 <fungi> similarly i haven't found time to work out what the communication schedule should be (if any) for decommissioning refstack 19:33:31 <fungi> we're already halfway through the hour so i'm going to skip a few older topics for now... 19:33:36 <fungi> #topic Enabling hashtags globally (corvus 20250523) 19:33:39 <fungi> this happened 19:33:46 <fungi> i haven't heard any complaints 19:33:56 <corvus> i think we can drop from agenda now? 19:34:00 <fungi> likely 19:34:21 <fungi> should be resolved, i just wanted to double-check there's been no fallout that escaped my attention 19:35:42 <fungi> #topic Adding CentOS 10 Stream Support to Glean, DIB, and Nodepool (clarkb 20250527) 19:36:13 <fungi> i don't think there's anything new on this one, but as there are a bunch of moving parts and multiple people involved i want to make sure 19:37:17 <fungi> sounds like no 19:37:29 <fungi> #topic projects.yaml normalization (frickler 20250617) 19:37:31 <fungi> a (possibly intentional) regression in ruamel.yaml has started adding spaces to the end of wrapped strings, akin to format=flowed E-mail wrapping 19:37:46 <fungi> #link https://sourceforge.net/p/ruamel-yaml/tickets/546/ Trailing whitespaces in some generated YAML outputs 19:38:06 <fungi> #link https://review.opendev.org/952006 this is impacting our normalization job output 19:38:53 <corvus> i really hope that's not intentional :) 19:39:06 <fungi> yeah, it seems odd to me 19:39:17 <corvus> ruamel does allow for a lot of control, i wonder if we could code around it 19:39:31 <fungi> as i mentioned in one of the comments, the only place i've ever seen that is in format=flowed e-mail 19:40:03 <corvus> (which is a mitigation for the quirks of a 40 year old standard) 19:40:09 <fungi> the lack of maintainer feedback on the linked ticket means i don't know whether it's worth hacking around something they may just fix in the next release 19:40:14 <corvus> (not something anyone should be doing this century) 19:40:56 <fungi> and yeah, wrapping is explicit in yaml, no need for inference like in e-mail message bodies 19:41:22 <fungi> probably just missed chomping the whitespace at the wrap point in some update 19:41:49 <corvus> my thoughts: 1) this is not a good format, we shouldn't adapt to it; 2) pinning for a couple months to see if things shake out seems okay; 3) if it doesn't get reverted, we should code around it... somehow. 19:42:08 <fungi> so anyway, the choices as currently presented to us are to temporarily pin/skip versions of that library, accept the new output, or keep ignoring it and hope something changes 19:42:22 <fungi> i agree with your take, corvus 19:42:35 <frickler> +1 19:43:18 <frickler> the second point I wanted to mention: why does the bot keep updating the change when there is no delta to the previous PS? 19:43:22 <fungi> also worth keeping an eye on (and linking in an inline comment for the pin/skip) the upstream bug report 19:44:10 <fungi> frickler: without having dug into it, i'm guessing it either doesn't attept to check whether the diff is equivalent to the existing change already in gerrit, or the attempt it makes is flawed in some way 19:44:41 <fungi> someone needs to figure out which that is if we don't want to continue getting new identical revisions daily 19:45:32 <frickler> I think I saw a check, but I don't have the ref handy 19:46:27 <fungi> #link https://review.opendev.org/952315 Block broken ruamel.yaml versions 19:46:52 <fungi> that's worth calling out as the (currently abandoned) option we seem to be favoring 19:47:06 <fungi> perhaps worth restoring and updating if necessary 19:47:11 <corvus> ++ 19:47:23 <frickler> restored now 19:47:48 <fungi> thanks 19:47:56 <frickler> will check whether the denylist is still current tomorrow 19:48:03 <fungi> thanks! 19:48:15 <fungi> i think that's still current, fwiw 19:49:01 <fungi> it has 2x+2 already, feel free to self-approve if you confirm there aren't newer releases yet to worry about 19:50:04 <fungi> okay, i don't know that this is worth spending more time on since those of us present seem to have reached a quick agreement 19:50:20 <fungi> #topic Open discussion 19:50:45 <fungi> we have 10 minutes left, anyone have updates on the older topics i skipped or anything else they want to bring up? 19:54:24 <fungi> seems like no. i return the balance of 5 minutes, don't spend it all in one place! 19:54:34 <fungi> #endmeeting