| opendevreview | Merged openstack/diskimage-builder master: Remove centos 9 and rhel 8 block in grub2 pkg-map https://review.opendev.org/c/openstack/diskimage-builder/+/825018 | 00:37 |
|---|---|---|
| fungi | oh, thanks! sorry i stepped away for a bit | 00:43 |
| clarkb | no worries, it all went well. Yay for our testing | 00:44 |
| opendevreview | Steve Baker proposed openstack/diskimage-builder master: Don't install python3-virtualenv on redhat family https://review.opendev.org/c/openstack/diskimage-builder/+/825034 | 01:22 |
| opendevreview | Steve Baker proposed openstack/diskimage-builder master: Never call virtualenv directly, use DIB_PYTHON_VIRTUALENV https://review.opendev.org/c/openstack/diskimage-builder/+/825204 | 01:22 |
| ianw | fungi/clarkb: downgrade testing complete; notes @ https://etherpad.opendev.org/p/gerrit-upgrade-3.4 and https://173.231.255.249/ is on hold to play with. if no concerns, i think update is about ready | 03:22 |
| *** ysandeep|out is now known as ysandeep | 04:33 | |
| *** ysandeep is now known as ysandeep|brb | 05:57 | |
| *** ysandeep|brb is now known as ysandeep | 06:09 | |
| *** ysandeep is now known as ysandeep|lunch | 08:13 | |
| *** jpena|off is now known as jpena | 08:38 | |
| *** ysandeep|lunch is now known as ysandeep | 09:16 | |
| *** lbragstad7 is now known as lbragstad | 09:19 | |
| opendevreview | chandan kumar proposed zuul/zuul-jobs master: Introduce iptables_package var https://review.opendev.org/c/zuul/zuul-jobs/+/822503 | 09:41 |
| *** dviroel|out is now known as dviroel | 10:58 | |
| *** ysandeep is now known as ysandeep|brb | 11:01 | |
| *** ysandeep|brb is now known as ysandeep | 11:15 | |
| *** rlandy|out is now known as rlandy|ruck | 11:15 | |
| fungi | ianw: thanks! and i guess holding with the fully-qualified project name worked? | 13:53 |
| *** ysandeep is now known as ysandeep|away | 14:04 | |
| *** sboyron_ is now known as sboyron | 14:46 | |
| *** dviroel is now known as dviroel|lunch | 14:57 | |
| opendevreview | Ananya proposed opendev/elastic-recheck rdo: [DNM] Erbot with opensearch https://review.opendev.org/c/opendev/elastic-recheck/+/825334 | 15:26 |
| opendevreview | Merged zuul/zuul-jobs master: Add support for RPM packages https://review.opendev.org/c/zuul/zuul-jobs/+/825118 | 15:45 |
| opendevreview | Jeremy Stanley proposed opendev/system-config master: Drop gitweb dependencies https://review.opendev.org/c/opendev/system-config/+/825337 | 15:54 |
| opendevreview | Jeremy Stanley proposed opendev/system-config master: Fix mixed spaces and hard tabs in Gerrit config https://review.opendev.org/c/opendev/system-config/+/825338 | 15:54 |
| opendevreview | Jeremy Stanley proposed opendev/system-config master: Use Gitea for Gerrit's code browser URLs https://review.opendev.org/c/opendev/system-config/+/825339 | 15:54 |
| opendevreview | Merged zuul/zuul-jobs master: Introduce iptables_package var https://review.opendev.org/c/zuul/zuul-jobs/+/822503 | 15:59 |
| *** dviroel|lunch is now known as dviroel | 16:15 | |
| clarkb | ianw: fungi: left a note about the reindex, but I think all of that is fine | 16:23 |
| fungi | clarkb: thanks, i expect to make a dnm change on top of those and set an autohold to poke at it some | 16:24 |
| fungi | just want to see if they even pass tests first | 16:24 |
| clarkb | oh I was referring to the downgrade stuff. But the gitweb stack is also exciting :) | 16:25 |
| clarkb | I just haven't gotten that far this morning yet | 16:25 |
| fungi | oh, the gerrit downgrade experiment. yep, hoping to look at that as soon as i'm done checking moderation queues | 16:26 |
| clarkb | I am going to hold one of the zuul streaming function test jobs to try and debug why focal python doesn't work and bionic's does | 16:26 |
| opendevreview | Eduardo Santos proposed openstack/diskimage-builder master: Fix openSUSE images https://review.opendev.org/c/openstack/diskimage-builder/+/825347 | 16:40 |
| clarkb | fungi: https://review.opendev.org/c/opendev/system-config/+/825179 should be an easy one based on what I discoverd about meetpad yesterday and its mute messaging | 16:47 |
| clarkb | fungi: on the last chaneg in the gitweb stack it seems to be missing a root url? | 16:50 |
| clarkb | also I'm not quite sure how we want to test that since we don't do jobs with gitea and gerrit in the same job. Maybe good enough to validate the urls are correct | 16:51 |
| fungi | clarkb: oh, yep thanks, i need to figure out where to set that, thanks | 16:51 |
| fungi | clarkb: on 825179, i don't suppose there's a way to fix the notice text instead? i'm not sure switchign back and forth between those behaviors is a great user experience, and presumably once the message gets corrected we'd undo that setting? | 16:51 |
| clarkb | I personally would prefer that I'm always muted on startup | 16:52 |
| clarkb | also we've had people join before that didn't realize they were unmuted with loud background noise interrupting things. I think generally forcing people to send their audio through explicitly is nicer for calls | 16:53 |
| fungi | okay, i can buy that argument. if we plan to keep it this way then i'm not opposed | 16:54 |
| corvus | i'd like to do a rolling restart of zuul | 16:54 |
| fungi | an announcement might be in order, but i guess the software announces it on connect anyway so maybe no need | 16:54 |
| fungi | corvus: sounds great! i'm happy to assist | 16:54 |
| corvus | i've run zuul_pull on bridge, so all containers are up to date | 16:55 |
| clarkb | I'm around. Looking at zuul ansible python things for the focal test node change. Ping me if I can help | 16:55 |
| corvus | thanks; hopefully this will just be slow and boring :) | 16:56 |
| corvus | i'll start by restarting the scheduler on zuul01 | 16:57 |
| fungi | clarkb: out of curiosity, do you know why it's START_VIDEO_MUTED=0 and START_AUDIO_MUTED=0 to enable? other toggles in those configs seem to enable with =1 | 16:57 |
| clarkb | fungi: the value is the first N connections to start unmuted | 16:57 |
| clarkb | if we set it to 1 then the first person to connect is not muted | 16:57 |
| clarkb | and so on | 16:57 |
| fungi | oh, yep the comment even says so. i should try harder at reading comprehension | 16:58 |
| clarkb | In the past they had a binary muted or not value. But then they changed it to this which is weird but seems to work | 16:58 |
| fungi | right, so =1 would leave the room leader unmuted but mute everyone else who joins | 16:58 |
| fungi | thanks | 16:59 |
| corvus | while it's coming back up, i'll go ahead and restart all the mergers | 16:59 |
| corvus | mergers are restarted; zuul01 is still initializing. next i'm going to gracefully stop half the executors. | 17:02 |
| corvus | (i think our load is sufficiently low to permit that) | 17:02 |
| fungi | very cool | 17:05 |
| corvus | i've run into a potential issue, so i've stopped the process... (i had not yet graceful'd the executors) | 17:06 |
| corvus | looking at logs now | 17:06 |
| corvus | i think the issue may only affect zuul01, so i don't think it's user-visible as long as zuul02 is still running | 17:07 |
| fungi | following along in the zuul matrix channel | 17:10 |
| fungi | clarkb: for the failure on https://zuul.opendev.org/t/openstack/build/3944a2d1bbf14e27b5481896957b3384 does that mean gerrit's bazel config hard-codes an expectation the gitiles plugin will be present anyway? i guess we need to build but then disable it if we don't want to fork that? | 17:11 |
| clarkb | fungi: ya seems that plugin is marked core so expected to always be present | 17:12 |
| fungi | thanks, i got quite confused reading the plugin docs, it seems there are built-in plugins in the gerrit codebase, but also apparently "core" plugins which are developed seperately | 17:17 |
| opendevreview | Merged opendev/system-config master: Start meetpad meetings with muted audio https://review.opendev.org/c/opendev/system-config/+/825179 | 17:19 |
| corvus | okay, it looks like we're going to merge a quick fix to zuul; so i have paused the rollout. current status is: mergers are upgraded (but are safe to continue to run at this version). zuul01 is stopped. zuul02 and the rest of the system are up and on the old version. | 17:24 |
| fungi | thanks! | 17:25 |
| corvus | after the fix merges, i will just restart zuul01 on the fixed version, restart the mergers again, and resume as before. | 17:25 |
| opendevreview | Eduardo Santos proposed openstack/diskimage-builder master: General improvements to the ubuntu-minimal docs https://review.opendev.org/c/openstack/diskimage-builder/+/806308 | 17:34 |
| clarkb | fungi: re gerrit plugins: yes, I think they made some plugins separate repos but they are still required for building gerrit as they are part of the core. Then there are additional plugins which are optional | 17:41 |
| *** jpena is now known as jpena|off | 17:43 | |
| opendevreview | Eduardo Santos proposed openstack/diskimage-builder master: Fix openSUSE images and bump them to 15.3 https://review.opendev.org/c/openstack/diskimage-builder/+/825347 | 17:45 |
| clarkb | corvus: any reason for me to avoid deleting my autohold while zuul is in the halfway restarted state? | 17:51 |
| corvus | clarkb: nope go ahead | 17:54 |
| fungi | clarkb: i'm at a loss to work out how to disable the gitiles plugin via configuration. it looks like there are cli and rest api options to do it if remote admin is allowed in the config, but i don't see a way to block loading a plugin at start other than renaming or deleting it on the filesystem | 17:55 |
| clarkb | fungi: does gitiles take a config option to disable itself maybe? | 17:56 |
| fungi | the scant docs available for the gitiles plugin indicate it has a config option to tell it not to take over gerrit's weblinks, which i suppose is a viable fallback, but seems silly to run the plugin at all in that case | 17:56 |
| clarkb | or maybe we don't disable gitiles? I think our current thing only shows us gitweb links | 17:56 |
| clarkb | well I don't know that it "runs" in that case. The code will be present without any path to execute it? | 17:56 |
| fungi | right now gerrit's webui has the code browsing links taken over by the gitiles plugin, which effectively shadows the built-in gitweb linking. the gitiles plugin can be told not to do that, but direct gitiles urls would still be available (we could block them in apache) | 17:58 |
| clarkb | fungi: right now I think gitiles is replaced by gitweb links | 17:59 |
| clarkb | But I see what youa re saying there isn't a way to prevent gitiles from loading if you know the urls | 17:59 |
| fungi | if i look at a change in gerrit, the git sha linked next to the patch set takes me to the gitiles view of that commit | 18:00 |
| fungi | rather than to gitweb | 18:00 |
| clarkb | ah neat. All of the other links are gitweb | 18:01 |
| clarkb | I bet that is a bug | 18:01 |
| clarkb | below the votes for example is a links section with gitweb and if you click on a specific file you get gitweb links there too | 18:01 |
| fungi | oh, got it. i wasn't aware it linked anywhere else | 18:01 |
| clarkb | oh but there is also a "browse" link adjacent to "gitweb" when you click specific files and browse == gitiles | 18:01 |
| fungi | anyway, the gitiles docs claim it takes over gerrit's web links unless you set noWebLinks=true | 18:02 |
| fungi | https://gerrit.googlesource.com/plugins/gitiles/+/HEAD/src/main/resources/+Documentation/config.md | 18:02 |
| clarkb | ya I'm not sure what the best thing is. Maybe file a bug against gerrit suggesting that removal of gitiels may be desireable if using another system | 18:02 |
| fungi | the docs for the `gerrit plugin remove` say that it renames the plugin jars themselves: https://gerrit.googlesource.com/plugins/gitiles/+/HEAD/src/main/resources/+Documentation/config.md | 18:05 |
| fungi | er, https://review.opendev.org/Documentation/cmd-plugin-remove.html | 18:05 |
| fungi | maybe the expectation is that you do that yourself if you don't want them loaded at start | 18:05 |
| clarkb | ya that might also work. Except gerrit may expect the jar to be functional since it is a core plugin | 18:06 |
| clarkb | I think the fact that this is a core plugin makes it special | 18:06 |
| clarkb | ya note there "Mandatory plugins cannot be disabled" | 18:06 |
| clarkb | I suspect gitiles meets that criteria based on the bazel complaints | 18:06 |
| fungi | that seems to also be configurable? https://review.opendev.org/Documentation/config-gerrit.html#plugins.mandatory | 18:07 |
| clarkb | ah then maybe it doesn't care | 18:08 |
| fungi | anyway, testing will catch at least that much for us, i'll try the simple approach first | 18:08 |
| clarkb | fwiw we do a bazel build release which outputs a single war iirc. Then when you run that it extracts all the things. Not sure what the best way to not extract the gitiles plugin might be | 18:10 |
| clarkb | in the error log on startup you'll often see it talk about cleaning up old plugins and that is when it replaces the existing content of the plugins dir with its new contents that it extracts aiui | 18:11 |
| fungi | oh, so the plugin jars aren't separate from the war, they're packed inside it? | 18:11 |
| clarkb | yes, that is my understanding. Then when the war executes it has some self extracting early step | 18:12 |
| clarkb | and that step is what will log about cleaning up plugins on gerrit startup. Since it deletes the old versions to replace them with its new versions | 18:12 |
| opendevreview | Elod Illes proposed openstack/project-config master: WIP: Remove ocata branch filters https://review.opendev.org/c/openstack/project-config/+/825372 | 18:26 |
| clarkb | it should be safe to land the limnoria update after 2200 UTC today as there is about a 4 hour window before the nextmeeting. However, tomorrow after 19:00 should be even better so I'll plan to land the limnoria update around 19:00 UTC tomorrow | 18:27 |
| opendevreview | Elod Illes proposed openstack/project-config master: WIP: Remove ocata related definitions https://review.opendev.org/c/openstack/project-config/+/825372 | 18:33 |
| opendevreview | Elod Illes proposed openstack/project-config master: WIP: Remove publish-install-guide ocata job https://review.opendev.org/c/openstack/project-config/+/825374 | 18:34 |
| clarkb | fungi re meetpad muting it seems we don't restart services when those values in the .env file update. Should I go ahead and restart those now? | 18:47 |
| fungi | clarkb: seems fine to me. i don't know how to tell if anyone's using it at the moment | 18:48 |
| clarkb | I think we can dig into logs somewhere. But ya I expect it will be quick and if someone is using it they should reconnect without trouble | 18:49 |
| opendevreview | Jeremy Stanley proposed opendev/system-config master: Use Gitea for Gerrit's code browser URLs https://review.opendev.org/c/opendev/system-config/+/825339 | 18:49 |
| clarkb | hrm I suspect that feature doesn't work as advertised :( | 18:52 |
| clarkb | I don't join automuted and have tried in firefox and chrome | 18:52 |
| clarkb | maybe down then up -d wasn't enough to rewrite teh configs? | 18:54 |
| fungi | we map the configs into the containers though, right? so if they're updated on disk then down/up should be all we need | 18:56 |
| opendevreview | Elod Illes proposed openstack/project-config master: Remove ocata related grafana definitions https://review.opendev.org/c/openstack/project-config/+/825379 | 18:56 |
| clarkb | no, we set env vars and then the jitsi meet templating writes out the configs for us | 18:56 |
| opendevreview | Elod Illes proposed openstack/project-config master: WIP: Remove ocata related definitions https://review.opendev.org/c/openstack/project-config/+/825372 | 18:57 |
| clarkb | config.startAudioMuted = 10; is still set in config.js so it didn't template out the file as expected. I might need to move that file aside and have it rewrite it? | 18:57 |
| opendevreview | Elod Illes proposed openstack/project-config master: WIP: Remove publish-install-guide ocata job https://review.opendev.org/c/openstack/project-config/+/825374 | 18:57 |
| fungi | worth a try. maybe we should always clear the configs on restart | 18:57 |
| clarkb | hrm except it did update the file according to the timestamp. So something is preventing the value from going through and I think I know what it is | 18:58 |
| clarkb | I'll get a change up shortly | 18:58 |
| opendevreview | Clark Boylan proposed opendev/system-config master: Fix meetpad audio mute setting https://review.opendev.org/c/opendev/system-config/+/825381 | 19:01 |
| clarkb | fungi: you have to explicitly set the value, then explicitly allow the value to be passed to the container | 19:01 |
| clarkb | I think ^ then a restart as I did previously should fix it | 19:01 |
| fungi | oh wow. double-book envvar accounting | 19:01 |
| corvus | i just used the zuul admen web api to promote the zuul change which had failed flaky tests | 19:03 |
| corvus | s/admen/admin/ | 19:03 |
| fungi | nice! | 19:03 |
| fungi | via curl or zuul-client? | 19:03 |
| corvus | via clicking gui buttons | 19:04 |
| * corvus uploaded an image: (25KiB) < https://matrix.org/_matrix/media/r0/download/acmegating.com/ZZlsAZJOJRHPQfUoArUdrSjU/image.png > | 19:04 | |
| fungi | whoa | 19:05 |
| fungi | that's slightly amazing | 19:05 |
| clarkb | oh nice we have paginated build search results now too | 19:07 |
| fungi | corvus: that reminds me, openstackid.org is in the process of being replaced by id.openinfra.dev, i guess that's set locally on the keycloak poc, not in configuration management yet? | 19:09 |
| corvus | fungi: yep | 19:56 |
| fungi | cool, i'll see about updating it as time permits | 19:56 |
| fungi | seems like 825381 should have been enqueued into the gate 15 minutes ago | 19:59 |
| fungi | oh, we have per-pipeline events now, seems like there's a backup in evaluating results | 20:00 |
| fungi | related to restarting? | 20:01 |
| fungi | higher-priority results are burning down first, so i guess check will start falling once the more important pipelines clear | 20:02 |
| fungi | and there it goes... "Starting gate jobs." | 20:03 |
| fungi | someone's got a 27-change-long series for keystone they're testing, that could be related | 20:05 |
| opendevreview | Merged openstack/project-config master: Remove ocata related grafana definitions https://review.opendev.org/c/openstack/project-config/+/825379 | 20:06 |
| corvus | we're running one one scheduler and suddenly delays are more noticeable. i take that as a good sign :) | 20:12 |
| corvus | zuul fix just merged | 20:13 |
| corvus | waiting for promote jobs | 20:13 |
| fungi | great point | 20:21 |
| opendevreview | Merged opendev/system-config master: Fix meetpad audio mute setting https://review.opendev.org/c/opendev/system-config/+/825381 | 20:29 |
| opendevreview | Jeremy Stanley proposed opendev/system-config master: DNM: Fail our Gerrit testing for an autohold https://review.opendev.org/c/opendev/system-config/+/825396 | 20:32 |
| clarkb | corvus: looks like zuul promoted images successfully | 20:46 |
| clarkb | Looks like the udpate to the docker-compose.yaml file did cause the service to be restart on meetpad | 20:56 |
| clarkb | I'll test things as soon as zuul says it is done deploying | 20:56 |
| clarkb | I think there might be some issue on my local system having firefox and chrome share the mic at the same time. It seems the browser that opens the meeting most recently wins. | 21:00 |
| clarkb | Also it seems that when you are the first person to join a meeting it doesn't mute you. But everyone else is muted. Thats probably good enough I guess | 21:01 |
| clarkb | config.startAudioMuted = 0; is confirmed to be in the config now too | 21:02 |
| corvus | pulling zuul images and starting zuul01 scheduler | 21:04 |
| opendevreview | Gage Hugo proposed openstack/project-config master: Retire security-specs repo https://review.opendev.org/c/openstack/project-config/+/825401 | 21:04 |
| corvus | restarting mergers | 21:06 |
| corvus | once the schedulers are done, i'd like to just do a hard-restart of the executors instead of a graceful restart. all the jobs will abort and restart which will delay results a bit, but the zuul rolling restart will be faster. | 21:10 |
| clarkb | I had missed it but one of the two backup hosts needs pruning again. I did it last time to make sure I had some comfort with it. Does anyone elsewant to do it this time around? | 21:10 |
| clarkb | corvus: I think that is fine as long as we don't have any release job type deals that might not like getting reenqueued. I can give the openstack reelase team a heads up | 21:11 |
| clarkb | release pipelines are also currently empty so should be fine | 21:11 |
| fungi | thanks, and yeah executor restarts seem like a pragmatic option to speed this up | 21:11 |
| fungi | i can try pruning the backups. we've got that process documented, right? | 21:12 |
| clarkb | fungi: yup should be documented and iirc the docs were quite good when I followed them last time | 21:12 |
| fungi | it's not urgent, we only just started to get the 90% warning | 21:12 |
| clarkb | ya we probably have a week or so to get to that at least | 21:13 |
| fungi | but i'll give it a shot after dinner | 21:13 |
| fungi | or tomorrow worst case | 21:13 |
| clarkb | corvus: I've warned the openstack release team. Should be good to go | 21:14 |
| corvus | cool, thx. | 21:14 |
| corvus | looks like scheduler 01 is up and no tracebacks | 21:14 |
| corvus | i'm going to restart scheduler02 and the web ui now. since it's not load balanced, the web ui will be unavailable, but the scheduler will continue running. | 21:15 |
| opendevreview | Gage Hugo proposed openstack/project-config master: Retire security-specs repo https://review.opendev.org/c/openstack/project-config/+/825401 | 21:15 |
| corvus | zuul02 restarting now | 21:17 |
| ianw | clarkb: sorry, back to your comment from hours ago :) yeah i did try it, thinking it would reindex online, but it fails to start with that error. i don't think it can handle it when the index version is higher | 21:19 |
| clarkb | ianw: gotcha | 21:19 |
| ianw | it looks like "delta-reindex" refers to ... https://gerrit.googlesource.com/plugins/high-availability/+/refs/heads/master/src/main/resources/Documentation/about.md#last-index-update-timestamp-storage | 21:21 |
| ianw | i'm not sure how you would revert the change index ... perhaps he is suggesting replacing with an old copy of index files from a backup you have on disk, maybe? | 21:23 |
| clarkb | ya then if you are HA doing a delta reindex off that backup? | 21:24 |
| opendevreview | Merged openstack/diskimage-builder master: Fixes for centos-9-stream efi behaviour https://review.opendev.org/c/openstack/diskimage-builder/+/824660 | 21:27 |
| ianw | if that is the case, it's something we should probably add to our upgrade steps to copy the indexes temporarily | 21:27 |
| clarkb | ++ though the rename might make things weird. I guess we can do a rename and startup and wait for all reindexing that th rename needs to complete (shouldn't take too long iirc). Then do another short downtime to upgrade | 21:28 |
| clarkb | that process is probably safer anyway | 21:28 |
| corvus | zuul02 back up and looking good so far; i'm going to hard-restart the executors now | 21:30 |
| clarkb | ianw: when you get a chance https://review.opendev.org/c/opendev/system-config/+/821331 is the limnoria update change. It got a new ps to use a tag instead of master. Looks like after 19:00 UTC tomorrow is a good time to land that due to a lack of meetings. | 21:36 |
| clarkb | for the rename and upgrade we should probably start a doc for that and look at rename changes to be sure we are ready. I can probably start on that tomorrow morning | 21:36 |
| *** dviroel is now known as dviroel|out | 21:38 | |
| clarkb | gitea 1.16 just got its first RC. Doesn't look like there will be many issues for us to upgrade after a quick glance at release notes | 21:38 |
| clarkb | Now that I'ev said that testing will show a bunch of things :) | 21:39 |
| fungi | ianw: clarkb: or just delete the indices and let it create new ones? | 21:43 |
| clarkb | fungi: that is slow (about 35 minutes on our current install iirc) | 21:44 |
| clarkb | but ya that is what ianw's document says to do essentially. Just restart | 21:44 |
| clarkb | s/restart/start over/ | 21:44 |
| fungi | that seems fine to me, it's only if we have to roll back anyway | 21:44 |
| corvus | things look good in zuul-land | 21:47 |
| corvus | #status log performed (mostly) rolling restart of zuul onto commit d304f4134f05fa08aab70e9add6ec490370dc6e2 | 21:48 |
| opendevstatus | corvus: finished logging | 21:48 |
| fungi | yay! | 21:48 |
| clarkb | next stop zuul 4.12.0 | 21:48 |
| ianw | fungi: yeah, just trying to decipher https://gerrit-review.googlesource.com/c/homepage/+/328539 :) | 21:48 |
| opendevreview | Gage Hugo proposed openstack/project-config master: Retire security-specs repo https://review.opendev.org/c/openstack/project-config/+/825401 | 21:53 |
| clarkb | I'm trying to ebtter understand the grafana deployment. It looks liek we use a simple sqlite db located in /var/lib/grafana on the container? This db also doesn't seem to be hosted from the system which implies that it is ephemeral and rebuilt from grafyaml anytime we update the image? | 22:00 |
| clarkb | does anyone know if that is the case? | 22:00 |
| ianw | this seems likely | 22:08 |
| ianw | not sure if it's ephemeral though | 22:08 |
| ianw | well, i guess it's not. it's ... overwrite only? basically we run grafyaml and that adds or overwrites any graph definitions, but will not remove ones that have gone away | 22:09 |
| clarkb | ya, and if we restart the grafana container often enough it will prune the old ones that go away on its own thatway | 22:09 |
| ianw | that may be a happy accident :) i don't think i intended that to happen but it sounds useful | 22:10 |
| clarkb | I'm taking notes on what is roughly required to run services with dedicated users | 22:10 |
| clarkb | https://etherpad.opendev.org/p/opendev-container-maintenance over here as step 0 in prioritizing and planning the next steps there | 22:11 |
| ianw | ok, cool. corvus had something out there at one point to switch this to the raw upstream images; we're rebuilding currently to put grafyaml in the container | 22:12 |
| ianw | i'm not sure where we got to with that | 22:12 |
| corvus | https://review.opendev.org/780128 is relevant | 22:18 |
| clarkb | thanks I'll have to process that. I wonder if that would make it hard to run grafana as a dedicated user (thinking of matrix-gerritbot). Chances are good it will be fine though | 22:20 |
| opendevreview | Ian Wienand proposed opendev/system-config master: Use grafyaml container image https://review.opendev.org/c/opendev/system-config/+/780128 | 22:25 |
| ianw | does service-lists actually require grafana, or is that just a typo i've managed to introduce @ https://opendev.org/opendev/system-config/src/branch/master/zuul.d/project.yaml#L501 ? | 22:27 |
| clarkb | ianw: I cannot think of a relationship between grafana and lists | 22:42 |
| clarkb | centos 8 stream systemd package updated in git about an hour ago to the fixed version | 22:49 |
| clarkb | the packages aren't in the repos from what I can see though. Also I note that the previous update to systemd in git.centos.org for centos 8 stream isn't in our mirrors either. Hopefully 239-55 ends up in there though | 22:51 |
| fungi | clarkb: https://104.239.175.140/ is a held node with the gitea change applied. the gitea link for the commit looks right (revision) but the link in the diff view (file) seems to put the refname in for ${commit} rather than the actual commit id | 22:51 |
| fungi | revision = "${project}/commit/${commit}" | 22:53 |
| clarkb | fungi: do they have a separate ${ref} substitution? | 22:53 |
| fungi | file = "${project}/src/commit/${commit}/${file}" | 22:53 |
| clarkb | if so then that seems like a bug | 22:53 |
| fungi | in the first one ${commit} is the git id and in the second one ${commit} is the refname | 22:53 |
| clarkb | but maybe we can work with it anyway? the refs are pushed to gitea | 22:53 |
| fungi | yep, i'll have to see if i can work out the correct syntax for putting the refname in a gitea url | 22:53 |
| clarkb | hrm I wonder if gitea supports viewing those refs | 22:54 |
| fungi | i think i had previously concluded gitea didn't support doing that, but it's worth revisiting | 22:54 |
| clarkb | ya I think you may be right. The chagne that was made upstream to gitea allowed us to push them and fetch them with git but the web ui is still clueless to them iirc | 22:55 |
| clarkb | fungi: the gerrit docs for gitweb.file say ${commit} should be the sha-1 I think this is a bug in gerrit | 22:57 |
| clarkb | I wonder if gitiles wants the refname instead so internally that got changed | 22:57 |
| fungi | the gitea webui does *show* the refnames at least: https://opendev.org/openstack/whereto/commit/d75f573d8be414002422e6890cd45ea631e17c1f see where it says changes/26/825326/1 | 22:58 |
| fungi | i wish it included them in the commit log view too like git and cgit/gitweb do | 22:58 |
| clarkb | ya I'm not having any luck trying to construct a url based on what I know of git and gitea. I strongly suspect they just don't have a way to expose those via the web ui | 23:00 |
| clarkb | but as you point out it would work if we could get a url with the sha1 | 23:00 |
| fungi | docs say that's what it's supposed to be: https://review.opendev.org/Documentation/config-gerrit.html#gitweb.file "Valid replacements are ${project} for the project name in Gerrit, ${file} for the file name and ${commit} for the SHA1 hash for the commit." | 23:03 |
| clarkb | yup hence my expectation that this is actually a bug. | 23:05 |
| clarkb | They should rename whatever they currently fill to ${ref} and have ${commit} point to the sha | 23:05 |
| fungi | of course, there's every chance that's a breaking change for a bunch of users who rely on the currently incorrect behavior, so they may instead leave ${commit} broken there and add a new ${no_really_commit} instead | 23:07 |
| clarkb | ya | 23:07 |
| clarkb | I've found the code locally let me get you a link | 23:07 |
| clarkb | this might take a minute due to java's deep tree structure | 23:07 |
| fungi | oh, thanks. java usually leaves me cross-eyed | 23:08 |
| clarkb | https://gerrit.googlesource.com/gerrit/+/refs/heads/stable-3.3/java/com/google/gerrit/server/config/GitwebConfig.java#322 that is the utility method that does the interpolation. | 23:12 |
| clarkb | Call point momentarily | 23:12 |
| clarkb | https://gerrit.googlesource.com/gerrit/+/refs/heads/stable-3.3/java/com/google/gerrit/server/restapi/change/GetDiff.java#194 and https://gerrit.googlesource.com/gerrit/+/refs/heads/stable-3.3/java/com/google/gerrit/server/restapi/change/GetDiff.java#230 appear to be how we construct that value | 23:13 |
| clarkb | the java lambdas still confuse me. The language has come a long way since I had to use it regularly. | 23:14 |
| *** rlandy|ruck is now known as rlandy|out | 23:15 | |
| clarkb | fungi: reading that it seems to be side A is the old side and if there is a patchset for that then get that refname otherwise the commit. Then for the new side always use the refname since it is always going to be a patchset? | 23:16 |
| clarkb | so ya I think it should be possible to alwys get the commit id for both sides and store them as different values then interpolate that as you suggest as a No really the commit please var | 23:17 |
| clarkb | maybe we should start with an upstream bug? I think coding this shouldn't be too difficult if we want to take a stab at it. Then we can use a patch file in our CI to ensure it works too. | 23:17 |
| clarkb | fungi: something like private final String commitA; this.commitA = sideA.fileInfo().commitId(); then later plumb through commit as a separate var alongside revision. And then interpolate that with ${commit_sha} | 23:19 |
| fungi | any indication if it's still the same in master? | 23:22 |
| fungi | but yeah, i can probably get a bug report filed tomorrow. eod has caught up to me | 23:22 |
| clarkb | ya it appeas to still be in master but with slightly different code | 23:23 |
| clarkb | https://gerrit.googlesource.com/gerrit/+/refs/heads/master/java/com/google/gerrit/server/restapi/change/GetDiff.java#235 | 23:24 |
| clarkb | I'm hacking up a thing because I already debugged most of what was necessary to write some code | 23:24 |
| fungi | according to the docs, the roottree option also takes a ${commit} so i wonder if it's correct or similarly broken... i can't find where that actually gets used in the ui | 23:24 |
| clarkb | the hardest part of this is always figuring out the right git diff syntax to get a patch that patch can apply | 23:32 |
| clarkb | ok first real issue is going to be that gitiles implements this interface | 23:35 |
| clarkb | figuring out compat across the two codebases will be fun | 23:35 |
| clarkb | fwiw I strongly suspect that the reason it is this way is for gitiles since gitiles depends on this interface directly | 23:40 |
| fungi | which is not so great if you want to avoid serving up a code browser view in gerrit separate from your primary code browser | 23:42 |
| clarkb | anyway my code compiles I just have to figure out git diff really quickly then can push up a patch that builds this updated gerrit | 23:43 |
| opendevreview | Clark Boylan proposed opendev/system-config master: DNM this patches gerrit to do gitea links for files https://review.opendev.org/c/opendev/system-config/+/825409 | 23:48 |
| clarkb | fungi: ^ I think you can rebase on that and try ${hash} | 23:49 |
| clarkb | one thign I didn't check is that commitId is the sha1 but it really should be | 23:53 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
