Monday, 2022-08-01

ianwmy second thought is "that's the type of thing i could find out looking at the build history" :)00:53
ianwwhich doens't look promising00:53
ianwit's varied between timed out and failed00:54
ianwit does end up a grub prompt00:56
ianwthankfully we have a working one
fungioh! i forgot we had jobs exercising that01:13
fungiso it would be possible to compare them after all01:14
opendevreviewIan Wienand proposed openstack/diskimage-builder master: [wip] Revert "Fix BLS entries for /boot partitions"
ianwfor clarity :01:17
ianwworking ->
ianwnot working ->
ianwfrom that i suspect
ianwi don't know why, yet, but that seems to be the difference01:18
ianwi don't really feel like it could be that change, a priori01:58
*** ysandeep|out is now known as ysandeep02:08
ianwit's like /boot just disappeared.  despite nothing seeming to change in terms of installed packages, etc.02:29
opendevreviewIan Wienand proposed openstack/diskimage-builder master: [wip] rockylinux : make /boot directory
*** ysandeep is now known as ysandeep|afk03:40
opendevreviewIan Wienand proposed openstack/diskimage-builder master: [wip] rockylinux : make /boot directory
*** ysandeep|afk is now known as ysandeep05:20
frickler_ptooh, fun, flake8 2.9.0 added E275, insisting on whitespace around keywords. very annoying for all the folks that have been treating things like "assert" and "raise" as functions. the correct form now would be "assert x" or maybe "assert (x)" instead of "assert(x)"05:35
*** frickler_pto is now known as frickler05:35
*** akahat_ is now known as akahat05:39
opendevreviewIan Wienand proposed openstack/diskimage-builder master: rockylinux : create machine-id early
ianwclarkb/NeilHanlon: ^ so I think that fixes it.  I'm open to other suggestions05:59
*** tobias-urdin_pto is now known as tobias-urdin06:06
*** ykarel_ is now known as ykarel06:41
*** ysandeep is now known as ysandeep|afk06:51
*** gibi_pto is now known as gibi07:05
*** ysandeep|afk is now known as ysandeep07:18
*** jpena|off is now known as jpena07:32
opendevreviewwangxiyuan proposed openstack/diskimage-builder master: Upgrade openEuler to 22.03 LTS
*** ysandeep is now known as ysandeep|lunch07:50
*** soniya29 is now known as soniya29|afk08:13
*** ysandeep|lunch is now known as ysandeep09:40
*** lbragstad9 is now known as lbragstad10:30
*** rlandy_ is now known as rlandy10:33
opendevreviewAnnu Mishra proposed openstack/diskimage-builder master: Adds required RHEL9 changes
fungifrickler: oh, i missed the new check because the codebase i tested the new version on was one where i went on a tear not long ago to fix those exact cases, del(x) was another i had mistakenly inserted a lot of11:20
opendevreviewwangxiyuan proposed openstack/diskimage-builder master: Upgrade openEuler to 22.03 LTS
*** pojadhav is now known as pojadhav|afk11:33
*** dviroel_ is now known as dviroel11:45
*** ysandeep is now known as ysandeep|afk12:23
*** elodilles is now known as elodilles_afk12:26
*** mnasiadka_ is now known as mnasiadka12:26
*** pojadhav|afk is now known as pojadhav13:15
NeilHanlonianw: that's a great catch.. i'm confused about the behavior and I'm gonna look into that. but if it works there, let's just call it good, IMO13:19
NeilHanlonseems the gate check failed for a transient reason13:22
NeilHanlonoh, i know what this is... iputils pulls in systemd during the container build 🤦13:23
fricklerNeilHanlon: the gate failure seems related to
*** ysandeep|afk is now known as ysandeep13:33
*** Guest6564 is now known as dasm14:00
*** elodilles_afk is now known as elodilles14:02
opendevreviewMerged openstack/diskimage-builder master: opensuse: better report checksum errors
fungisince regular attendees to the diversity and inclusion working group meeting have had frequent problems with browsers blocking audio for jitsi-meet, i've manually switched on the "prejoin page" in advance of the meeting in order to confirm whether that does help matters14:46
fungispecifically, switching config.prejoinPageEnabled from false to true in /var/jitsi-meet/web/config.js on meetpad0114:47
Clark[m]I'm not sure if you need to restart containers to see that change. But probably not14:49
fungii did not need to, i already tested that it shows up in the browser14:50
fungithe config confusingly says that the prejoinConfig block replaces prejoinPageEnabled, but turning that on instead seemed to cause the interface not to load at all14:52
fungiit's possible i needed to keep it in the "var config = {" block instead of just prepending "config." to the name though14:53
fungibut for now prejoinPageEnabled does seem to still work14:53
fungiahh, yeah that was it14:55
fungiswitched over to that for now14:55
*** frenzyfriday|rover is now known as frenzyfriday|rover|lunch15:01
fungiClark[m]: spotz says she was unable to get the jitsi mobile app to unmute after joining15:06
fungisomething to test later15:06
*** melwitt_ is now known as melwitt15:11
Guest6563yup I can help test that when you have a moment to be on the other isde of the call15:14
*** Guest6563 is now known as clarkb15:16
clarkbAnd now I am me again15:17
opendevreviewClark Boylan proposed opendev/system-config master: Update to Gitea 1.17
clarkbGitea 1.17.0 exists now. I'm not in  rush to upgrade. Seems likethey often do a .1 and/or .2 bugfix release pretty quickly to address issues people find. But Since th erelease properly exists now I've updated the chagne and removed the WIP designation as it is technically mergeable now if reviewers are happy15:32
*** rlandy is now known as rlandy|brb15:40
*** frenzyfriday|rover|lunch is now known as frenzyfriday|rover15:51
*** rlandy|brb is now known as rlandy15:52
*** tweining is now known as tweining|off15:52
clarkbNeilHanlon: frickler fungi ianw I've manually downlaoded that opensuse image and checked the sha256sum. They are indeed different :/15:59
clarkbthis has happened before and they updated the sha256sum file after the fact. Considering this has happened again it wouldn't surprise me if their build system is producing buggy results15:59
clarkbdirk: ^ fyi doesn't match its corresponding sha256sum file16:00
NeilHanlonwhat in god's green earth is that url pattern lol16:03
clarkbI forget what the : denotes but if you drop the : you get different content along those path entries16:08
NeilHanlonhuh, interesting. I've never really looked at opensuse much16:08
*** dviroel is now known as dviroel|lunch16:22
*** marios is now known as marios|out16:24
*** ysandeep is now known as ysandeep|out16:26
clarkbwhat i don't understand is how passed but the rocky fix didn't16:30
clarkbare we getting different versions of files, maybe depending on locality or load balancer luck?16:30
clarkbah yup these may ultimately get served by mirrorsand maybe they aren't in sync for the different files?16:33
opendevreviewClark Boylan proposed openstack/diskimage-builder master: Disable the opensuse functest
clarkbthats sort of the nuclear option16:36
NeilHanlonah, yeah that could be it, especially if it's new. we have a similar problem with rocky due to the staggered way in which mirrors can grab content16:39
fungiusually when dealing with checksummed or signed things is to add the new artifact without removing the old one, then replace the signatures/checksums to point to the new artifact, then clean up the old artifact (usually after a bit of a delay)16:41
fungithat way if a mirror grabs the files at any random time, it should get a coherent artifact+sig combo16:43
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
*** jpena is now known as jpena|off17:14
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
*** dviroel|lunch is now known as dviroel17:24
clarkbfungi: ^ current issue is both mailman core and mailman web say mysql isn't available but I've got a test case indicating it is listening on the specified port. Any idea what might be going on there?18:00
clarkbI did change the mariadb version and use host networking compared to the original docker compose file. I wonder if it is related to one or both of those changes18:01
clarkb(I'm specifying connection to localhost:3306 in the mysql urls so the host nteworking update shouldn't matter18:02
clarkb appears to be the code that is looping18:03
clarkbbased on that I don't think the version mismatch is a likely issue since mysqladmin ping is likely to have compat across versions?18:04
fungissl vs plain?18:09
clarkblooking at the little check there it seems to not do auth its just checking if the port is listening? which the testinfra test says it is18:10
clarkbbut it is doing it fro mthe context of the docker container itself not the host so maybe something with how host networking is setup?18:10
clarkbI might have to hold a node to debug that.18:10
fungiyeah, i would confirm the containers can really reach each other before assuming there's something deeper wrong18:11
clarkbcross checking our other docker compose files we don't have to specify ports when using network_mode:host (which makes sense as we aren't NATing at all)18:13
clarkbhold is created and recheck requested18:21
clarkbtcp        0      0  *               LISTEN19:08
clarkbbah I think it may be that localhost == local socket and that isn't present in the container19:10
* clarkb tries
clarkbthat doesn't work because urlparse with returns None for the hostname19:16
clarkbhuh it works on my local python319:18
clarkbhrm no something else must be broken19:20
clarkbok I htink is our first case of "the upstream images are problematic"19:30
clarkbI'll push up a change that should illustrate how this fails19:30
clarkbthe previous issue had to do with quoting :/19:30
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
clarkbI'm overriding the allowed hosts list but it is an append not a replacement19:32
clarkband they lookup a thing that isn't in dns because we arne't using docker networking. We can either add records to /etc/hosts for those names to point at localhost or not use host networking or override that file maybe or do our own images19:33
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
clarkbI don't expect ^ to help but if it does work the nmaybe that is an easy workaround20:33
*** dviroel is now known as dviroel|out21:26
*** dasm is now known as dasm|off21:41
clarkbyup the hostname appears to be ignored21:53
ianwclarkb/NeilHanlon: yeah, i played with this a while ago and got different things from the mirrors at different points, even signed up to the bug tracker but then it went away22:04
ianwapropos opensuse images22:04
ianw to be exact22:06
clarkbya Itried a recheck but it seems to have failed. I also pushed a change to rmeove that test entirely22:08
clarkbanother approach may be to pull the opensuse image from a specific mirror. Do you think that would be a whorthwile change?22:08
clarkbI can write that if so22:08
ianwif we have some data maybe we should disable but file a bug report upstream.  at least we can report it, even if nobody fixes it22:10
fungithing is, even if they ignore the bug report it'll likely fix itself at their next mirror push or something22:12
clarkbthe data I have is earlier today I downlodaed the file and got sha256sum b77c0c48f970c00ab9c99a574114f41d54c12a51c4d29e31e2945bfdf28dc9c8 for the actual file but the sha256sum file on the webserver says 864667deb442174c05d5fb8cd2dd9a8165ded0fb13f31ea7e96139ab911a753122:12
ianwyeah so we have it out of sync today and on 2022-07-11 @
ianwdid you end up at
fungianother option might be to add some strategic checksum validation to out rsync script and skip the vos release when it's no good22:15
ianwfor this test, i think we're pulling form upstream directly, i don't think we mirror these bits22:15
fungiahh, okay22:15
clarkbcorrect this is straight from upstream22:18
clarkbI think it might be more stable if we just pick a specific mirror22:18
clarkbunless that mirror is out of sync for similar reasons or goes down entirely22:18
ianwin that paste above i was pulling from the provo- mirror from a us-based vm.  but locally i got sent to some very unreliable .au mirror that cut out most of the time22:20
ianwwill file something in a bit and we can disable for now22:24
clarkbsound sgood22:26
clarkbI've updated the meeting agenda for tomorrow. Anything else to add before I send it out?22:41
* fungi checks his open reviews22:42
fungisome reviews have come up for git-review22:43
clarkbFeel free to add an agenda item if git review needs meeting coverage.22:44
fungii've got a pair of changes about making its test rebase process clearer to users, and there's another change to add support for a newer gerrit feature we can't test without increasing the gerrit version we test against22:44
funginot sure if those rise to the level of an agenda topic22:44
clarkbthe reabse thing is a fun one since people seem extremely confused about it. I really thing git review -x leads to that22:45
clarkbhopefully your chagnes can make that more clear22:45
fungileave some comments on 850054 or its child if they don't22:45
fungiincreasing the gerrit version we test with is always complicated though, since the git-review testsuite does a bespoke gerrit deployment that invariably needs care and feeding with each new version22:46
fungii suspect the proper solution is to remake the test setup for git-review22:47
clarkbonce it gets to 3.2 and newer it should be pretty consistent fro mthen on22:47
clarkbfungi: the second change related to rebases is workinprogress. were you looking for feedback ro still editing it?22:55
fungiyeah, that needs some consensus building, i guess. i think that's why i did wip initially22:57
fungibecause it's a behavior change22:57
clarkbgotcha I can leave a note about me being ok with the behavior change. I almost always git rebase --abort when that happens22:58
clarkbthe one thing that maybe would make it a good idea to keep as is is if people are less comfortable knowing how to start a rebase but once started can finish it22:58
ianwclarkb: filed for the images23:48
opendevreviewIan Wienand proposed openstack/diskimage-builder master: Disable the opensuse functest
ianw^ just a respin with the bug link23:51

Generated by 2.17.3 by Marius Gedminas - find it at!