| ianw | yeah i was just wondering that ... | 00:00 |
|---|---|---|
| ianw | i could rm everything on the master branch, but not sure that solves the problem of old changeset pages ranking higher than actual gerrit when searching for specific things | 00:01 |
| ianw | i guess we could explicitly robots remove that repo | 00:02 |
| opendevreview | Ian Wienand proposed openstack/project-config master: opendev/gerrit : retire project https://review.opendev.org/c/openstack/project-config/+/833939 | 00:04 |
| *** dviroel|ruck is now known as dviroel|ruck|Afk | 00:05 | |
| ianw | i've abandoned changes; there's no jobs setup, so that retires it from gerrit, i'll take suggestions in there of what we want to do with the actual repo? | 00:16 |
| opendevreview | Ian Wienand proposed opendev/system-config master: gitea: disallow opendev/gerrit https://review.opendev.org/c/opendev/system-config/+/833940 | 00:23 |
| opendevreview | Ian Wienand proposed opendev/system-config master: gitea: disallow opendev/gerrit https://review.opendev.org/c/opendev/system-config/+/833940 | 00:24 |
| opendevreview | Ian Wienand proposed openstack/project-config master: opendev/gerrit : retire project https://review.opendev.org/c/openstack/project-config/+/833939 | 00:46 |
| *** rlandy|bbl is now known as rlandy|out | 00:59 | |
| opendevreview | Ian Wienand proposed opendev/gerrit master: Retire repo https://review.opendev.org/c/opendev/gerrit/+/833943 | 01:00 |
| opendevreview | Ian Wienand proposed openstack/project-config master: opendev/gerrit : retire project https://review.opendev.org/c/openstack/project-config/+/833939 | 01:02 |
| *** kevinz_ is now known as kevinz | 02:33 | |
| wxy-xiyuan | ianw: Hi, can I spend your time for the repo sync problem? it looks that openEuler mirror is down from 14th Feb. It sync package from openEuler mirror in Russia. Is that the problem, the internet is blocked? | 02:45 |
| wxy-xiyuan | https://grafana.opendev.org/d/9871b26303/afs?orgId=1 | 02:46 |
| wxy-xiyuan | Or any place I can get the sync error? | 02:46 |
| ianw | wxy-xiyuan: ahhh, yes that might be a problem :/ | 03:07 |
| ianw | the logs are exported @ | 03:07 |
| ianw | https://static.opendev.org/mirror/logs/rsync-mirrors/openeuler.log | 03:08 |
| ianw | rsync: failed to connect to ru-repo.openeuler.org (159.138.204.168): Connection timed out (110) | 03:08 |
| ianw | i would definitely imagine this has something to do with ... current events | 03:09 |
| fungi | wxy-xiyuan: if there's a different rsync mirror we should be connecting to instead, please propose a change for this line: https://opendev.org/opendev/system-config/src/branch/master/playbooks/roles/mirror-update/files/openeuler-mirror-update#L35 | 03:10 |
| ianw | https://openeuler.org/en/mirror/list/ doesn't appear to list mirrors | 03:11 |
| wxy-xiyuan | OK, there are many mirror. I'll change it Singapore one | 03:11 |
| ianw | ++ | 03:12 |
| wxy-xiyuan | the en url seems broken. I'll let openEuler fix it quick. The cn one https://www.openeuler.org/zh/mirror/list/ | 03:12 |
| wxy-xiyuan | Thanks for help | 03:12 |
| opendevreview | Merged zuul/zuul-jobs master: ensure-podman: add containernetworking-plugins https://review.opendev.org/c/zuul/zuul-jobs/+/831326 | 03:36 |
| opendevreview | wangxiyuan proposed opendev/system-config master: Fix openEuler mirror problem https://review.opendev.org/c/opendev/system-config/+/833950 | 04:04 |
| opendevreview | wangxiyuan proposed opendev/system-config master: Fix openEuler mirror problem https://review.opendev.org/c/opendev/system-config/+/833950 | 04:05 |
| *** ysandeep|out is now known as ysandeep | 05:05 | |
| opendevreview | Merged opendev/system-config master: Fix openEuler mirror problem https://review.opendev.org/c/opendev/system-config/+/833950 | 05:13 |
| *** arxcruz|off is now known as arxcruz | 07:50 | |
| *** ysandeep is now known as ysandeep|lunch | 07:54 | |
| *** jpena|off is now known as jpena | 08:01 | |
| *** ysandeep|lunch is now known as ysandeep | 08:38 | |
| opendevreview | wangxiyuan proposed openstack/diskimage-builder master: Enable Yum mirror for openEuler element https://review.opendev.org/c/openstack/diskimage-builder/+/833969 | 08:46 |
| *** pojadhav- is now known as pojadhav | 08:54 | |
| opendevreview | wangxiyuan proposed openstack/diskimage-builder master: Enable Yum mirror for openEuler element https://review.opendev.org/c/openstack/diskimage-builder/+/833969 | 10:03 |
| *** rlandy|out is now known as rlandy | 10:26 | |
| *** ysandeep is now known as ysandeep|afk | 10:37 | |
| *** marios is now known as marios|ruck | 10:40 | |
| *** ysandeep|afk is now known as ysandeep | 11:16 | |
| *** dviroel|ruck|Afk is now known as dviroel|ruck | 11:16 | |
| opendevreview | Benedikt Löffler proposed openstack/diskimage-builder master: Use https for downloading ubuntu images https://review.opendev.org/c/openstack/diskimage-builder/+/833997 | 12:43 |
| gthiemonge | Hey Folks, promote-openstack-releasenotes failed in https://review.opendev.org/c/openstack/octavia/+/832604 (after the patch merged), is there anything we can do here? | 13:19 |
| fungi | gthiemonge: it's idempotent so will get rerun the next time a change merges to any branch of that repo | 13:28 |
| gthiemonge | fungi: ok! thanks! it's not a big deal | 13:29 |
| fungi | i haven't looked at the failure, but the most common cause is when uploads of builds triggered by different branches of the same project race one another and the two rsync projects trip over each othe's tempgiles | 13:30 |
| fungi | er, and the two rsync processes trip over each other's tempfiles | 13:32 |
| *** ysandeep is now known as ysandeep|afk | 13:40 | |
| *** rlandy is now known as rlandy|mtg | 14:00 | |
| fungi | need to go run some errands now, but should be back by 16:00 at the latest (hopefully sooner) | 14:04 |
| *** ysandeep|afk is now known as ysandeep|out | 14:43 | |
| *** marios is now known as marios|ruck | 14:50 | |
| *** rlandy|mtg is now known as rlandy | 15:01 | |
| clarkb | fungi: when you get back want to +A https://review.opendev.org/c/opendev/system-config/+/828184 if then is a good time to help monitor? I expect to be around until lunch and can adjust accordingly so no worries on my side | 15:14 |
| fungi | yep, back-ish already and will approve now while catching up on other things | 15:15 |
| clarkb | thanks! | 15:15 |
| * clarkb finds breakfast while the gitea change is in the gate | 15:41 | |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] add public url attribute https://review.opendev.org/c/zuul/zuul-jobs/+/834043 | 15:48 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] set ContentEncoding only if not 'None' https://review.opendev.org/c/zuul/zuul-jobs/+/834044 | 15:48 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] add ipa extension to known mime types https://review.opendev.org/c/zuul/zuul-jobs/+/834045 | 15:48 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] add android mime-type https://review.opendev.org/c/zuul/zuul-jobs/+/834046 | 15:48 |
| clarkb | fungi: in double checking gitea things I think we will want to check the gitea authorized keys file is updated to point to the new command path (looks like newer gitea added an extra option or two as well) | 15:54 |
| clarkb | fungi: the new gitea 1.16 deployment sets this properly but I'm realized we don't check that it updates it for existing keys. Not too worried about it since gitea has been good about upgrades in the past but calling it out as something to check | 15:54 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] add public url attribute https://review.opendev.org/c/zuul/zuul-jobs/+/834043 | 15:55 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] set ContentEncoding only if not 'None' https://review.opendev.org/c/zuul/zuul-jobs/+/834044 | 15:56 |
| clarkb | fungi: /var/gitea/data/git/.ssh/authorized_keys if you want to compare between the test node and production | 15:56 |
| clarkb | manually editing that shouldn't be too bad either | 15:56 |
| clarkb | if it becomes necessary | 15:57 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] set ContentEncoding only if not 'None' https://review.opendev.org/c/zuul/zuul-jobs/+/834044 | 15:57 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] add ipa extension to known mime types https://review.opendev.org/c/zuul/zuul-jobs/+/834045 | 15:57 |
| fungi | i guess if it doesn't upgrade appropriately, we should expect gerrit to cease synchronizing commits to gitea? | 15:57 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] add android mime-type https://review.opendev.org/c/zuul/zuul-jobs/+/834046 | 15:57 |
| clarkb | fungi: yes that would be the impact | 15:57 |
| clarkb | I think our recovery process would be to backup the old file, edit file to use command value and options from testing, possibly restart the services, then trigger replication. Shouldn't be too bad | 15:58 |
| clarkb | if you think its worth trying to test that 1.15 -> 1.16 does this for us we can -W the change before it merges. Or another option would be to put gitea02-gitea08 in the emergency file | 15:59 |
| fungi | i'm not too worried about it. worst case we find immediately after rollout that it's wrong, we fix up the location and re-replicate everything | 16:00 |
| fungi | the impact to users will be nearly nonexistent, they'll just not see updates reflected in gitea for a brief period today | 16:01 |
| clarkb | https://github.com/go-gitea/gitea/pull/17846 is the related chagne for background | 16:01 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] add ipa extension to known mime types https://review.opendev.org/c/zuul/zuul-jobs/+/834045 | 16:01 |
| clarkb | fungi: wfm | 16:01 |
| *** dviroel|ruck is now known as dviroel|ruck|lunch | 16:02 | |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] add android mime-type https://review.opendev.org/c/zuul/zuul-jobs/+/834046 | 16:02 |
| clarkb | it does say if we execute the correct path that it should automatically set things for us and my change did cahnge the path | 16:02 |
| clarkb | and the test node has the correct value in its authorized_keys file so chances are it will just work | 16:02 |
| fungi | python 3.7.13, 3.8.13, 3.9.11 and 3.10.3 were just tagged. i'll be recompiling for a bit | 16:03 |
| *** marios|ruck is now known as marios | 16:04 | |
| clarkb | in looking at the gitea ssh stuff more closely I've discovered that /var/gitea/data/gitea/avatars is where the avatar files apparently go. I strongly suspect that those are either not in the right format or got lost in production in an update. Chances are if we reset avatars it will find them again after writing them to that dir | 16:11 |
| fungi | yeah, that brokenness may be cruft from an early upgrade which has just been carried forward ever since | 16:13 |
| fungi | we could test the theory on one of the servers | 16:13 |
| clarkb | fungi: ya we should double check what produces a 404 in the browser when fetching an avatar then cross check against that dir's contents | 16:15 |
| clarkb | then if the content is missing we could use a test node to try and just replace the file | 16:15 |
| clarkb | or if the content is in the wrong format try to determine teh correct format and convert it | 16:15 |
| *** frenzy_friday is now known as frenzyfriday | 16:17 | |
| zigo | I recieved a voucher for the summit, but it doesn't seem working, there's nowhere to enter the promo code, who should I complain about this? | 16:22 |
| clarkb | zigo: when you register you select the ticket you want and above that is a "enter promo code" option That is where it goes | 16:23 |
| zigo | Oh gosh... didn't see ... :) | 16:24 |
| zigo | Thanks. | 16:24 |
| zigo | clarkb: Will I have the pleasure to see you in Berlin, as well as fungi? | 16:24 |
| fungi | zigo: that's my hope! | 16:24 |
| zigo | Gr8 ! :) | 16:25 |
| fungi | predicting the future is hard even in reasonably certain times, but i'm going to do my best to be there | 16:25 |
| zigo | BTW: My first Yoga VM is up and running (using Bullseye backports), and pings, thanks everyone for this very smooth release! :) | 16:25 |
| clarkb | nice | 16:25 |
| fungi | i saw the post to the ml, excellent progress! | 16:25 |
| opendevreview | Merged opendev/system-config master: Update Gitea to 1.16.4 https://review.opendev.org/c/opendev/system-config/+/828184 | 16:26 |
| clarkb | I'm going to abandon https://review.opendev.org/c/opendev/system-config/+/714237 since it seems like we're happy consuming gerrit releases and just trying our best to keep up to date with those rather than master | 16:27 |
| fungi | sgtm | 16:27 |
| fungi | looks like the deploy for 828184 is already in progress too | 16:27 |
| fungi | well, the image upload anyway, but the prod service job should start any moment | 16:28 |
| clarkb | fungi: it will just do the image promotion then wait for the hourly jobs to complete. The zuul job is running then eavesdrop, but should be soon | 16:28 |
| fungi | oh, yep, though those are nearly done | 16:28 |
| *** ysandeep|out is now known as ysandeep | 16:36 | |
| clarkb | the gitea job has begun | 16:42 |
| clarkb | it does the load balancer first iirc | 16:42 |
| clarkb | so there is a bit of a delay before it gets to the backends | 16:42 |
| clarkb | https://gitea01.opendev.org:3081/ just updated | 16:46 |
| clarkb | it updated its ssh authorized keys file so I think that should be good. Still a good idea to double check replication but far less concerned now | 16:47 |
| *** marios is now known as marios|out | 16:49 | |
| clarkb | 02 is updated now too | 16:49 |
| opendevreview | Andy Ladjadj proposed zuul/zuul-jobs master: [upload-logs-base] set ContentEncoding only if not 'None' https://review.opendev.org/c/zuul/zuul-jobs/+/834044 | 16:50 |
| clarkb | I don't see any obvious errors replicating to gitea01 in the gerrit replication log and I can fetch zuul/zuul-jobs:refs/changes/44/834044/4 which was replicated after the update according to the replication log | 16:54 |
| clarkb | I can fetch that from gitea01 I mean | 16:54 |
| clarkb | the first four giteas are done. | 16:55 |
| fungi | yeah, it's looking okay to me so far | 16:55 |
| clarkb | One thing I noticed is that when you have mutliple replications fro the same project it will delay any extras and then check if they are still needed after the first one completes | 16:56 |
| clarkb | I didn't realize the replication plugin was that smart, but that is a nice optimization | 16:56 |
| *** jpena is now known as jpena|off | 17:01 | |
| clarkb | all but gitea08 are done now. SHould be complete shortly | 17:04 |
| fungi | the theme definitely changed in production for me after the upgrade too, as expected | 17:05 |
| fungi | now it's all light grey and pale green on charcoal | 17:05 |
| clarkb | and now 08 is done. I think the playbook is just finishing up stuff and we should be all set | 17:08 |
| clarkb | Once the playbook ends I can #status log this | 17:08 |
| clarkb | https://review.opendev.org/c/openstack/project-config/+/831837 is the other thing on my list that is starting to grow moss. ianw if you get a chance when your day starts to take a look at that it owuld be appreciated | 17:10 |
| fungi | i've noticed that the element matrix webclient similarly takes on these colors when i set its appearance to "match system theme" so i guess this is a thing browsers expose | 17:11 |
| clarkb | https://zuul.opendev.org/t/openstack/build/8aa32897e69c4b17bdbb42ca96bd67dd it was successful | 17:11 |
| clarkb | #status log Upgraded gitea cluster to gitea 1.16.4 | 17:11 |
| opendevstatus | clarkb: finished logging | 17:11 |
| fungi | thanks clarkb! | 17:11 |
| clarkb | fungi: I'm tempted to set one of firefox's more colorful things now and see if gitea attempts to match :) | 17:12 |
| fungi | that would be a good test | 17:12 |
| clarkb | fungi: I've confirmed the behavior chagne with the dark theme in FF but the others all seem to use a bright background and don't really change much | 17:13 |
| clarkb | my eyes didn't know what to think about "cheese puff" | 17:13 |
| clarkb | "Teal Galaxy" looks interesting but isn't installed and I'm sure I'll disable it within 30 seconds of installing it so won't bother | 17:14 |
| fungi | heh | 17:14 |
| *** dviroel|ruck|lunch is now known as dviroel|ruck | 17:15 | |
| *** arxcruz is now known as arxcruz|off | 17:18 | |
| clarkb | I won't delete the autohold for the test 1.16.4 instance as that may be useful to debug the avatar issues. | 17:23 |
| *** ysandeep is now known as ysandeep|out | 17:38 | |
| clarkb | fungi: I've rechecked https://review.opendev.org/c/opendev/system-config/+/765577 and that one should be an easy land | 17:45 |
| opendevreview | Clark Boylan proposed opendev/system-config master: Use add instead of wget to download things for gerrit images https://review.opendev.org/c/opendev/system-config/+/715087 | 17:50 |
| clarkb | really quickly before lunch I've confirmed that the 404'ing images at https://gitea01.opendev.org:3081/explore/organizations do not have files in the /var/gitea/data/gitea/avatars dir | 18:51 |
| clarkb | the files that are in there appear to be PNGs | 18:51 |
| clarkb | I think we can try to manually upload PNGs for those orgs on each of the backends. Or maybe even just copy them into place? Anyway lunch now | 18:51 |
| fungi | i think a solution allowing people to contribute logos for those in project-config and supplying a placeholder otherwise could be nice, but it's not critical | 19:18 |
| clarkb | fungi: poking around in the web ui on our test 1.16.4 instance it does appear we can login as root then go to /admin/orgs and click the little edit button for an org | 19:38 |
| clarkb | From there you can "choose new avatar" | 19:39 |
| clarkb | But there is also a "delete current avatar" button. I'm going to try that on the test instance and see if it generates a enw random one for me | 19:39 |
| clarkb | no that set the cirros avatar on the test instance to the opendev logo | 19:40 |
| clarkb | oh if I click "Update avatar" without uploading a file it generates a random one. | 19:40 |
| clarkb | fungi: I suspect this means in production the quickest thing to do for now is to do what I just did on the test host. Delete the avatar then click update without a file set to generate a new random one. | 19:41 |
| clarkb | fungi: ianw: maybe one of you want to test that on the test instance and make sure I'm not crazy then we can try on one of the prod giteas | 19:41 |
| clarkb | https://104.239.143.159:3081/ is the test one | 19:41 |
| fungi | how do we login as root, exactly? | 19:43 |
| clarkb | fungi: https://docs.opendev.org/opendev/system-config/latest/gitea.html#create-all-bare-git-repos has directions. In this case you have to use the test credentials in the test groups file in system-config | 19:44 |
| fungi | ahh, okay | 19:46 |
| fungi | in inventory/service/host_vars/gitea01.opendev.org.yam | 19:47 |
| fungi | l | 19:47 |
| clarkb | that won't have the secret in it. Its the zuul template group vars that has it | 19:47 |
| fungi | er, yeah, git grep is failing me but i'll find it | 19:48 |
| fungi | okay, managed to log into the test webui | 19:51 |
| fungi | yeah, so it seems to generate deterministic avatars | 19:53 |
| clarkb | fungi: yes it uses an md5sum hash of some sort based on my quick read of some code | 19:53 |
| fungi | clicking "update avatar" on its own does nothing that i can see if the org already has one | 19:53 |
| clarkb | fungi: ya I suspect we'll have to delete the avatar then click the update avatar button | 19:54 |
| fungi | but deleting and updating does remove and restore it | 19:54 |
| clarkb | to force it to forget the current avatar | 19:54 |
| fungi | seems so, yes | 19:54 |
| clarkb | you know what though | 19:54 |
| clarkb | I wonder if we had set the avatars to relative image file paths for what we host and then pointed it at logos | 19:54 |
| clarkb | then when we refactored the logo hosting we broke things? | 19:54 |
| clarkb | anyway that seems like something to address in a roll forward fashion after fixing the immediate 404 problem with generated avatars | 19:55 |
| fungi | agreed | 19:55 |
| fungi | we'll also need to repeat it on each gitea server individually | 19:55 |
| clarkb | correct | 19:55 |
| fungi | or call the relevant api methods maybe | 19:55 |
| clarkb | fungi: another thing I notice is that all of the non working avatars appear to have double digit file paths but the working ones seem to be primarily hash looking file names | 19:58 |
| clarkb | implying something changed with how gitea generates those | 19:58 |
| clarkb | so it still could be a bug in gitea we tripped over rather than our own reorganization of values | 19:59 |
| fungi | presumably supporting custom logos for avatars would be fairly straightforward if someone ends up wanting to write an implementation, we could simply copy them from a git repo onto the filesystems right? | 19:59 |
| clarkb | fungi: it is probably better to go through the apis for that in case filesystem locations change, but yes potentially | 20:00 |
| fungi | oh, i see, if there's an upload api then that makes sense | 20:00 |
| fungi | presumably the same credentials we use to authenticate repository creation from manage-projects could do it | 20:01 |
| fungi | er, not manage-projects, but however we automate it | 20:01 |
| fungi | getting spacey, it must be almost dinner time | 20:02 |
| *** dviroel|ruck is now known as dviroel|ruck|afk | 20:03 | |
| clarkb | PATCH /api/orgs/$org seems to allow you to set the avatar url | 20:05 |
| clarkb | not seeing a delete avatar or regenerate avatar url similar to what those buttons do | 20:05 |
| fungi | we have 25 orgs, so clicky-clicky 200 times there. not _terrible_ | 20:07 |
| clarkb | fungi: and not all of them are broken | 20:07 |
| fungi | oh, that simplifies things | 20:07 |
| opendevreview | Merged opendev/system-config master: Clean up Gerrit image builds https://review.opendev.org/c/opendev/system-config/+/765577 | 20:19 |
| ianw | retirement lgtm | 21:36 |
| opendevreview | Merged openstack/project-config master: Finalize batch of opendev repo retirements https://review.opendev.org/c/openstack/project-config/+/831837 | 21:45 |
| ianw | if we can PATCH update the org logos, i guess we might as well update the ansible to copy in a set a "real" logo for those that have it | 21:49 |
| ianw | in *and* set | 21:49 |
| ianw | https://try.gitea.io/api/swagger#/organization/orgEdit doesn't list AvatarURL | 21:54 |
| ianw | yeah, i can't really see the code that would set the avatar url in https://github.com/go-gitea/gitea/blob/main/routers/api/v1/org/org.go#L337 | 22:01 |
| ianw | if err := user_setting.UpdateAvatarSetting(ctx, form, ctx.Org.Organization.AsUser()); err != nil { | 22:15 |
| ianw | @ https://github.com/go-gitea/gitea/blob/main/routers/web/org/setting.go#L145 | 22:16 |
| ianw | so it looks like the web form update "casts" the org to a user and then calls that function. but it doesn't appear to be exported as an API | 22:17 |
| ianw | that function just sets in the db "if err := user_model.UpdateUserCols(db.DefaultContext, ctxUser, "avatar", "avatar_email", "use_custom_avatar"); err != nil {" | 22:19 |
| ianw | all the broken image orgs have "use_custom_avatar" set to 1 | 22:31 |
| *** rlandy is now known as rlandy|PTO | 22:31 | |
| ianw | "old" orgs seem to have a monotonically incrementing "avatar" field | 22:33 |
| ianw | e.g. | 22:34 |
| ianw | | lower_name | avatar | avatar_email | use_custom_avatar | | 22:34 |
| ianw | | opendev | 4 | | 1 | | 22:34 |
| ianw | | openstack-attic | 5 | | 1 | | 22:34 |
| ianw | | cirros | a320c2a3eb4fb45635ae2d338488f58b | | 1 | | 22:34 |
| ianw | but then, one we have added recently, e.g. cirros, has it's md5 as the avatar (a320 == echo -n "cirros" | md5sum) | 22:35 |
| ianw | https://opendev.org/cirros/ has hte logo | 22:36 |
| ianw | pyca is "27" | 22:38 |
| ianw | and we have a file /var/gitea/data/gitea/avatars/27 | 22:38 |
| ianw | u.Avatar = fmt.Sprintf("%x", md5.Sum([]byte(fmt.Sprintf("%d-%x", u.ID, md5.Sum(data))))) | 22:41 |
| ianw | if err = user_model.UpdateUserCols(ctx, u, "use_custom_avatar", "avatar"); err != nil { | 22:41 |
| ianw | when you upload a custom avatar, it saves the file as "u.ID-$(md5sum data)", then sets "use_custom_avatar" | 22:41 |
| ianw | so, in short, the "avatar" field of the user appears to be just a name of a file on disk in /var/gitea/data/gitea/avatars | 22:42 |
| clarkb | ianw: pyca is a working one with the short file path | 22:47 |
| clarkb | I guess we didn't set a custom avatar for tha | 22:47 |
| clarkb | btu it is old enough to have the old format | 22:48 |
| ianw | yes, but it gets odder | 22:48 |
| ianw | ooohhh, i see ... | 22:48 |
| ianw | so, without use_custom_avatar, it seems to hit gravatar | 22:49 |
| clarkb | ah | 22:49 |
| ianw | e.g. https://opendev.org/gerrit/ | 22:49 |
| clarkb | but orgs don't have email addrs so that doesn't quite work for orgs? | 22:49 |
| ianw | we see a logo there because it does | 22:49 |
| clarkb | I don't see a logo there | 22:49 |
| ianw | https://secure.gravatar.com/avatar/b11210e93bb757a1b17636767c31cb46?d=identicon&s=870 | 22:50 |
| clarkb | I get the hashed thing | 22:50 |
| clarkb | but also gerrit is a user not a org | 22:50 |
| ianw | yeah, the "identicon" ... sorry, i'm equating that to the logo | 22:50 |
| clarkb | ah ok | 22:50 |
| clarkb | ianw: areyou sure it is always going to gravatar? if you look in /var/gitea/data/gitea/avatars the files there are pngs | 22:50 |
| clarkb | I haven't copied a png out of there to view it but I assume it is the identicon value? | 22:51 |
| ianw | i think it is *if* the user db has use_custom_avatar==0 | 22:51 |
| clarkb | I see. If use_custom_avatar == 0 go to gravatar else lookup in /var/gitea/data/gitea/avatars | 22:52 |
| ianw | i think so | 22:52 |
| clarkb | ianw: did you see my suggestion that we delete the avatar and then update teh avatar for those orgs? you can test this on the test instance if you like. But that seems to restore the identicon at least. Then we can think about how to set custom logos | 22:53 |
| ianw | i did, but trying to think about how we can not click on all the giteas :) | 22:53 |
| ianw | we might be able to do a custom script that hits the non-api app end-points to update it | 22:54 |
| ianw | to set custom logos would be a hack, but not hard | 22:55 |
| clarkb | ya the old management ansible library did that in the past but gitea has since grown api supprot for all the things it does so we stopped oding that | 22:55 |
| clarkb | but the history there may show us how to do that again | 22:55 |
| ianw | 1) dump logo file in /var/gitea/data/gitea/avatars/<name-of-org.svg>" | 22:55 |
| clarkb | I just think it is likely to be a one off so not too bad to click things | 22:55 |
| clarkb | updating the avatar url does seem to have an api if you want to set it to somthing else | 22:56 |
| clarkb | ianw: https://opendev.org/api/swagger -> organization -> PATCH /orgs/{org} | 22:56 |
| ianw | 2) run db query avatar="<name-of-org.svg" use_custom_avatar="1" where name="<org>" | 22:57 |
| clarkb | oh hrm that only shows avatar url in the response not the PATCH input, maybe they don't accept it | 22:57 |
| ianw | clarkb: yeah, that doesn't seem to actually expose it, see my notes above | 22:57 |
| clarkb | why would they put it in the response if it isn't editable... | 22:57 |
| clarkb | ianw: re 2) aren't they all already set to use_custom_avatar=1? | 22:58 |
| clarkb | we get a 404 because the files are not on disk | 22:58 |
| ianw | clarkb: well, yeah, they are | 22:58 |
| clarkb | ianw: they aren't on gitea01 | 22:59 |
| clarkb | 27 is there so pyca works but 4 is not so opendev does not work. | 23:00 |
| ianw | sorry, i think two diferent things -- yes the orgs all have use_custom_avatar=1 | 23:00 |
| ianw | and yes, i agree the files are not on disk | 23:00 |
| clarkb | I think if we just put a PNG at /var/gitea/data/gitea/avatars/4 that meet the requirements of gitea (let me dig up a link for that) we'll be good | 23:00 |
| ianw | yes, we can do that | 23:00 |
| clarkb | https://docs.gitea.io/en-us/config-cheat-sheet/#picture-picture has limits for those files both in pixel dimensions and file size | 23:01 |
| ianw | i'm thinking though since we've figured this all out, we may as well dump a real logo in there for each org, and update the "avatar" field | 23:02 |
| clarkb | ianw: what do you mean by update the avatar field? | 23:03 |
| clarkb | switch it to the md5sum version? | 23:03 |
| clarkb | but ya no objections to using real logos | 23:04 |
| ianw | so we've established that in the gitea db that the "avatar" field of the "user" table is just a plain-text file-name to a file in /var/gitea/data/gitea/avatars | 23:04 |
| clarkb | yup | 23:04 |
| ianw | so we could dump real logos in there, and have ansible run a manual mysql update of that field during deployment | 23:05 |
| clarkb | right, but why do we need to update mysql is my questions I guess | 23:05 |
| clarkb | the entries are already all in the db we just have to write a file to the correct location? | 23:05 |
| clarkb | we'd want md5sum paths for everything if we do that though I guess since the order of the old file format may differ between backends | 23:06 |
| ianw | right, it's to override this old v new thing we've got going on | 23:06 |
| clarkb | got it | 23:06 |
| ianw | the old ones have been created with "avatar = uid" i guess, and new ones with "avatar = md5sum(username)" | 23:07 |
| clarkb | this will work until gitea ends up getting FIPS'd and then we have to sha256sum :) but until then ++ | 23:07 |
| ianw | i don't think that will make a difference --- i'd suggest we put the svg/png on disk as the org name, and then just set "avatar = <name>" | 23:08 |
| clarkb | I wonder if that will cause any other problems. I guess not since they already have multiple schemes in there | 23:09 |
| ianw | yeah, i don't think so -- cause it's just a plain-text entry ultimately | 23:09 |
| ianw | the confusing thing is the way they've setup the default values, which appears to have changed | 23:10 |
| ianw | i'm just looking at how we create orgs now to see if this can work | 23:10 |
| clarkb | currently we create orgs entirely through the api | 23:11 |
| clarkb | we do a listing first and then for each org not already in the listing create it iirc. The other consideration is when we rename projects across orgs eg x/foo to openstack/foo if it ends up creating a new org then this would potentially cross paths with that too | 23:11 |
| ianw | where is the list of orgs? | 23:12 |
| ianw | i know i have looked at this all before but i've forgotten :) | 23:12 |
| clarkb | ianw: it is generated from openstack/project-config/gerrit/projects.yaml | 23:12 |
| clarkb | that file is an input to jeepyb on the gerrit side and our gitea-git-repos role on the gitea side | 23:13 |
| ianw | ahhh, that's right, gitea-git-repos | 23:13 |
| clarkb | with renames we give it a rename specific input file since renames are a one shot | 23:14 |
| clarkb | those are recorded in opendev/project-config/renames | 23:14 |
| clarkb | and they are consumed by the rename playbook and the things it calls | 23:14 |
| ianw | we could do it "out-of-band" ... hit opendev.org/api/v1/user/orgs and parse out the usernames | 23:16 |
| clarkb | ianw: ya maybe as a followup to gitea-git-repos and a last step of the rename process? | 23:16 |
| ianw | write out a file for each org username (if no specific file, use a default place-holder) | 23:16 |
| clarkb | basically let creation and renaming happen. Then come back around and set logos as appropriate | 23:16 |
| ianw | that would be idempotent | 23:16 |
| ianw | if any files change, then run the db update query directly on the gitea hosts | 23:17 |
| ianw | to update the avatar field of each username that the logo file changed for | 23:17 |
| ianw | if there was a rename, well we'd just write out it's correct logo on the next run | 23:18 |
| ianw | in the mean time, it would have whatever default one gitea set | 23:18 |
| ianw | (we've established that new logos are setup correctly to the defaults, it's just the old ones that a borked for unknown reasons) | 23:18 |
| clarkb | note the "default gitea set" is different for each one, but ya it would be a short period then correct | 23:19 |
| ianw | yep, the thing not to have is a broken image at any point, since that looks lame | 23:19 |
| ianw | should gitea gain a proper api for this, it could be switched from a db update easily | 23:20 |
| clarkb | ++ | 23:20 |
| ianw | i think that can all work | 23:21 |
| ianw | the great thing is that can all be done in a speculative fashion thanks to our job setup :) | 23:21 |
| clarkb | yup, that has become super valuable. | 23:22 |
| clarkb | related, any reason to keep my hold for the 1.16.4 test node at this point or should I delete it? | 23:22 |
| clarkb | I think we've run down the details on avatars and sounds like the next step may produce holdable nodes if we need them | 23:22 |
| ianw | yeah, no need for the node i don't think | 23:25 |
| clarkb | ianw: thank you for adding checks around being able to set the reviewed flag on files https://review.opendev.org/c/opendev/system-config/+/715087 failed on that check and it updates how we install the jdbc driver for mariadb | 23:27 |
| clarkb | tldr our testing is working. THough i don't know why it faield yet I'm sure I did something wrong :) | 23:27 |
| clarkb | "Cannot load JDBC driver class 'org.mariadb.jdbc.Driver'" So I did something wrong with the switch to ADD I guess | 23:28 |
| clarkb | https://docs.docker.com/engine/reference/builder/#add says that you can use a src that is a url. This will need more digging I guess | 23:30 |
| clarkb | maybe permissions related | 23:30 |
| ianw | huh, yeah it looks right | 23:33 |
| clarkb | https://zuul.opendev.org/t/openstack/build/c4faa8db575f40378a9814470d6006ca/log/job-output.txt#1560-1563 it seems to have run it | 23:34 |
| clarkb | ianw: -rw------- 1 root root 632979 Jan 29 2021 mariadb-java-client.jar | 23:39 |
| clarkb | so ya I need to make that 644 instead | 23:39 |
| clarkb | I'll work on an update | 23:39 |
| clarkb | I've deleted the 1.16.4 autohold | 23:41 |
| opendevreview | Clark Boylan proposed opendev/system-config master: Use add instead of wget to download things for gerrit images https://review.opendev.org/c/opendev/system-config/+/715087 | 23:44 |
| *** ysandeep|out is now known as ysandeep | 23:46 | |
| clarkb | ricolin: I'm doing cleanup of our old system-config changes and noticed https://review.opendev.org/c/opendev/system-config/+/760770 is that still something you would like to change? IF so you'll need to edit https://opendev.org/opendev/system-config/src/branch/master/inventory/service/host_vars/lists.openstack.org.yaml#L283 instead. Also that won't actually update the list that will | 23:54 |
| clarkb | have to be done manually after we update the record keeping | 23:54 |
| clarkb | fungi: elodilles I'm going to abandon https://review.opendev.org/c/opendev/system-config/+/779546 since the openstack meta project handles that now | 23:54 |
| clarkb | er I guess I can wait for you to confirm that is a safe abandonment | 23:55 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!