Friday, 2025-10-31

tonybI tried the new etherpad and for me it's a regression from 2.5.1 that I tried yesterday.  I'll do more debugging after the i18n PTG session00:06
clarkbtonyb: what sort of regression?00:22
tonybI can't get any pads to load00:24
tonybthe home page works, but they added more text to the already full button on the home page00:25
fungibutton label overrun00:26
clarkbtonyb: huh it worked much better for me in both ff and chrome. Did you do this in a clean brower with no existing cookies?00:26
clarkbno cookeies and no cached data00:26
clarkbsometimes the old version and new version conflict over that stuff00:26
tonybI can look at that.00:28
prometheanfireJayF, clarkb: openrc generally uses netifrc and systemd generally uses networkd or networkmanager, depending on what the user wants00:46
prometheanfireit's hard to write support for everything :|00:46
prometheanfireone day maybe cross distro server stuff will be "standard"00:47
fungiclarkb: wrt the weblate api key, we have one recorded in our usual list so maybe it's that00:48
clarkbfungi: oh til00:48
fungilisted under "openstack.weblate.cloud"00:48
JayFprometheanfire: I've got a good idea, let's write an application suite to standardize all of this stuff across all Linux distributions. We can call it systemd 😂00:50
fungii thought you were going to say lsb00:52
fungithough i guess that was more about standardizing things that already existed, not replacing everything that exists and then declaring it a standard00:53
JayFI'm totally just making the joke there. I'm legitimately glad that lots of options exist even if I choose the boring one.00:54
prometheanfirewhat's debian's new one?  I think that one is declaritive in yaml?00:54
fungiare you talking about netplan? i think that came out of canonical/ubuntu00:56
fungisomeone was pushing for it to become the default in debian but that didn't get very far from what i saw00:56
fungiwhere i saw the discussion peter out is that networkmanager is great for roaming wireless clients, systemd-networkd is great for static server deployments, and trying to force "one true network management layer" was not sensible00:57
fungithough lately on my mobile clients i'm combining networkmanager (for wifi) and systemd-networkd (for wireguard vpn to my house) with no problem00:58
fungiapparently recent networkmanager versions now also support a yaml-based configuration model lifted from netplan too01:00
fungii also still just use ifupdown on a lot of my debian systems, particularly in a cloud vm context01:02
JayFI've moved mostly to a dhcp-everything configuration01:03
clarkbnetworkmanager is my persistent nemesis. Latest problem is it cannot wireguard at all.01:03
JayFif my opnsense box is down I'm likely offline anyway01:03
JayFonly exception are a couple of specific servers I want online in that specific case301:03
clarkbIt at least detects when I configure wireguard separately01:03
tonybclarkb: Okay after removing the cookies and restarting both chrome and ff the new etherpad is "fine".  I still can't switch into light/dark mode as per 2.5.001:08
Clark[m]Ya I think light/dark modes aren't supported with this skin01:09
prometheanfiremobile I'm using networkmanager (with wireguard) and static I use networkd, ya01:16
JayFhonestly my favorite thing about ussing networkd is how easy it is to rename interfaces01:17
JayFhaving a router interface names actually be wan/lan/san/etc is very, very useful01:17
prometheanfireclarkb: I haven't tried NM 1.54 with wireguard yet01:17
JayFI used to do it back in the day with RHEL5, it had syntax to do it in ... whatever /etc/sysconfig/network was :D01:18
prometheanfirerenaming interfaces is very nice, yes, a few coworkers were surprised when I told them to configure the public interface with xyz IP and asked which one was public01:19
tonyb"update constraint for pbr to new release 7.0.2 | https://review.opendev.org/c/openstack/requirements/+/965358" is in the gate pipeline now so we should check in in about 90mins for any new failures that might be related01:39
tonybinfra-root: Should we land 964728: Force gitea http(s) connectivity through the load balancer early next week or after the gitea update.   I suspect we don't want both changes to land at once01:42
tonybmnasiadka, noonedeadpunk: I had some follow-up thoughts about constraints, stable branches and deployments projects.  Other than ${this} channel is there a good channel you're both in to discuss it?06:50
opendevreviewMathieu Parent proposed opendev/glean master: Use systemd-networkd automatically when enabled  https://review.opendev.org/c/opendev/glean/+/96301007:07
mnasiadkaClark[m]:I don’t know if you’re still considering mjollnir - but it seems draupnir is the replacement(?) - https://github.com/the-draupnir-project/Draupnir09:13
tonybmnasiadka: It certainly claims to be a drop-in replacement rewrite09:17
tonybmnasiadka: but worth looking at Thanks09:18
opendevreviewMerged openstack/diskimage-builder master: Fix openstack-ci-mirrors element for CentOS Stream  https://review.opendev.org/c/openstack/diskimage-builder/+/96534409:41
opendevreviewribaudr proposed opendev/irc-meetings master: Move Nova weekly meeting from Tue to Mon  https://review.opendev.org/c/opendev/irc-meetings/+/96546510:26
*** jroll08 is now known as jroll010:46
stephenfinclarkb: fungi: We're going to need a pbr 7.0.3. https://review.opendev.org/c/openstack/pbr/+/965474 is essential (see linked bug: I'd missed this change before now) and https://review.opendev.org/c/openstack/pbr/+/965470 is a pretty noticeable issue that's affecting pbr itself.   https://review.opendev.org/c/openstack/pbr/+/965313 is not essential but might good to get in also11:31
fungilooking13:08
fungistephenfin: looks like the py27 job is throwing ImportError: pbr._compat.commands.LocalEggInfo when doing an editable install of pbr. do you happen to know why yet?13:12
fungion 96547413:12
fungiat first i thought it might be related to clarkb's https://review.opendev.org/965397 but it looks like that failed out of the gate due to the depends-on not having merged yet13:28
Clark[m]fungi: stephenfin: that py27 error may be related to the change since 965470 passes?13:54
Clark[m]And yes the failed job ran on bionic so not due to switching to jammy13:55
stephenfinHmm, that happened locally when I had a bug in the code. I'll take a look13:56
stephenfinugh, python 2's awful relative import mechanism strikes again13:59
stephenfin>>> from pbr._compat import commands13:59
stephenfin...13:59
stephenfin    from setuptools.command import egg_info13:59
stephenfinImportError: No module named command13:59
* stephenfin rename module13:59
stephenfinClark[m]: fungi: Should be resolved now https://review.opendev.org/c/openstack/pbr/+/96547414:10
Clark[m]Thanks I'll review all three changes properly in a bit. Still booting the morning. Do we need to warn the release team?14:12
noonedeadpunktonyb: we can use #openstack-requirements as well I guess? seems we all are there 14:38
clarkbstephenfin: that latest patchset hit ImportError: No module named setuptools14:48
clarkbI've approved the other two changes as they pass testing and seem fine14:49
clarkbI think we just need to s/setuptools/versions/ on that import line. If I decide that is the acse I can push an update14:51
clarkbya I think that is it I'll go ahead and modify14:52
fungiah yep thanks, i missed that reviewing14:54
clarkbtonyb: re etherpad, ya thats a known thing with updates being weird with overriding /etc/hosts. I typically use an incognito tab to work around it14:57
clarkbtonyb: for gitea I think we can focus on stability changes from our side before worrying too much about 1.25. Typicalyl they will release a .1 or .2 with many bugfixes which are worth waiting for. On the stability side of things we've got the limit access through the load balancer change and I also proposed a change to increase the memory available to memcached to hopefully mitigate14:59
clarkbsome of the effects of crawling as a cache poison14:59
clarkbbut ya maybe early next week we get an etherpad upgrade done to finally catch back up there, then figure out what if anything we want to do with the gitea deployment14:59
clarkbtesting looks good on that third pbr chagne for the develop command so I've approved it now too.15:14
fungithanks!15:15
opendevreviewMerged opendev/irc-meetings master: Move Nova weekly meeting from Tue to Mon  https://review.opendev.org/c/opendev/irc-meetings/+/96546515:15
fungii can propose pbr 7.0.3 as soon as it lands15:15
clarkbdo you think it is worth direct enqueing to the gate or should we just ride it out?15:19
clarkbtoday is the day we all learn guthub releases are not immutable by default (they just added support for this)15:27
fungii think we can just wait for it to merge normally, i'm buried under other things in the meantime anyway15:29
clarkbmy main concern is that if setuptools makes a release in the next 5 minutes we might break the release pipeline to fix things in the first place. But I think we can figure out a workaround for that up to an including manual workarounds up to pypi upload15:29
fungiyeah15:31
corvusdo we know what timezone the setuptools deadline is in?  :)15:42
clarkbcorvus: no, and we don't even know if they've remembered their own deadline15:42
clarkbthat said historically they are fans of the weekend release. I wonder if today is the public deadline then we'll see a new release saturday or sunday15:42
clarkbubuntu has announce x86-64-v3 support but will be doing so via additional packages (and possibly only for packages where the greatest impact will be made?)16:37
fungiheh, blizzard's dockenstack jenkins is e-mailing us about job failures again ;)17:26
clarkbI'll take the emails over the failed ssh connections ever few seconds :)17:27
fungiindeed17:28
mnasiadkaThere’s a question from my downstream how https://opendev.org/openstack/cinder/commit/49f386f3bc5d5a1e16c1b8cf2ef46c6bf453cd42 got in, if https://review.opendev.org/c/openstack/cinder/+/820375 is in merge conflict - I assume there’s a simple answer, but indeed that looks a bit weird from a user perspective17:58
clarkbmnasiadka: commits are replicated to gitea from gerrit. Whether or not that commit is merged to a branch is a separate question18:00
clarkbthat ref doesn't show up in the master branch18:02
fungimnasiadka: all refs in gerrit, in fact, are replicated to gitea18:02
clarkb(which is expected since it wasn't merged)18:02
fungithe named refs for changes/patchsets get replicated to gitea too (though they're harder to access than raw commit ids due to ui limitations), as well as git notes that contain a bunch of the gerrit review metadata18:03
mnasiadkaAh right, I just assumed something wrong happened, and this commit is not tied to any branch18:05
fungibasically everything in gerrit for a project, whether or not it's already merged, or can ever even merge at all (old patchsets for a change that got revised) appear in the repositories in gitea18:05
fungiwe also make direct use of that in the gerrit ui, linking to commits in gitea from patchsets18:06
mnasiadkaRight, thanks for the explanation - just the fact it’s not merged in any branch wasn’t that super easy visible :)18:08
fungifor example, https://review.opendev.org/c/openstack/pbr/+/965474/1 is an outdated patchset, and the "gitea" link immediately below the vote block takes you to it in gitea even though that will never be able to appear on any branch18:08
mnasiadkaActually I never noticed the gitea link in the Gerrit UI before ;)18:09
fungimnasiadka: click the "..." button under the commit message and you'll see a "this commit is container in" list with branches and tags. for things that aren't on a branch it's empty18:09
fungis/container/contained/18:10
fungiso for example https://opendev.org/openstack/pbr/commit/562f4ebb5b3f1280beab1f7d1e3e56906e2169f7 says master branch and 7.0.2 tag18:10
fungiwhile https://opendev.org/openstack/pbr/commit/472f98878123b342558fb35351bb9f8920317cfe is empty there18:10
mnasiadkaYup, noticed that after your initial response :)18:11
fungii agree it's rather hidden18:11
mnasiadkaGreat, I learned a new thing today ;)18:11
mnasiadka7pm Friday - it’s been a long week , have a good weekend :)18:11
tonyb5am Saturday.... it's been a long sleepless week18:16
fungithanks! it's beer-o'clock here18:21
fungialways18:21
mnasiadkafungi: it’s always everywhere beer o’clock18:25
fungisee, i knew we agreed on something18:26
mnasiadkatonyb: for the requirements discussion - we can probably use #openstack-ansible or create a new one for this. You can also use the etherpad we used for cross deploy session18:27
clarkbI think gitea is slow again today and I'm not sure I have the energy to debug and identify who needs their ipaddresses blocked today18:31
fungii'm headed out to beer^H^H^H^Hlunch now that the pbr 7.0.3 release is proposed, will be back in a while18:40
clarkbenjoy!18:42
tonybclarkb: I'll look at gitea after breakfast.19:08
clarkbthanks. I'm going to pop out for a bike ride now myself19:24
tonybclarkb: gitea looks okay to me.  All servers have a load average between 1 and 2.5, though some are coming down from a higher value.19:52
tonybgrafana's view of the LB indicates that there was a spike about an hour ago19:52
fungiokay, back20:04
Clark[m]tonyb: ack thank you for checking 21:13
clarkbas a sanity check all of the zuul components are currently running (since last week's restart failed due to a launcher being down)21:43
fungifingers crossed they all restart again in a few hours21:44
clarkbI've run down the issue with the reestart too and will have a patch up shortly21:52
clarkb(I don't think we need to rush it in before these restarts, but good to have eventually)21:52
opendevreviewClark Boylan proposed opendev/system-config master: Fix zuul service graceful shutdown routines  https://review.opendev.org/c/opendev/system-config/+/96586621:57
clarkbI think ^ that should address the problem. Its a docker-compose vs docker compose output problem when shutting things down21:57
tonybwhat time does the weekly restart happen?   I can be around to monitor/check afterwards 23:19
tonybnm I see it in crontab on bridge23:26
clarkbtonyb: it should start in like an hour?23:30
clarkbwe've found that the late friday into saturday block tends to be quiet so its a good time to wind down the executors (we do so gracefully one by one which requires all jobs on each executor be stopped one after another)23:30
clarkbsince the components list shows a full set of components I'm not too worried about it23:31
tonyb`1 0 * * 6` which I read as 00:01 on Saturday ... so 30mins from now23:31
tonybokay.23:32
clarkbtonyb: and the easiest way to follow along is to just refresh the zuul /components page periodically23:32
tonybnoted23:33

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!