[08:12:48] dcaro: re colalign in tabulate, it was introduced in 0.8.3... we run 0.8.2 lol. Skipping this param doesn't mess up alignment so I'll just remove it. I guess this means I need to create a new release with this fix? [08:13:04] yep [08:13:41] good morning btw :) [08:14:02] 😎 morning! [08:14:09] that was meant to be a sun... [08:14:15] ☀️ [08:14:24] the sun was implied [08:14:41] here its more like 🌧️ today [08:15:42] same here yep [08:16:47] random trivia discovered in the process: tabulate <0.9 is not compatible with python >=3.10 [08:34:15] hmm, there is some discrepancy (to say the least) in running ci in gitlab (passes), via `utils/run_ci_locally.sh` (some unit tests fail but pre-commit passes) and via tox (unit tests pass but pre-commit mypy fails) [08:37:48] you can choose now any result you like :) [08:40:33] no time to investigate today, created a ticket :) T353044 [08:40:33] T353044: [builds-cli][ci] Investigate discrepancy between different CI envs - https://phabricator.wikimedia.org/T353044 [08:41:22] the only result that matters is gitlab ci, right? ...right?! [08:43:22] dcaro: quick review when you have a moment: https://gitlab.wikimedia.org/repos/cloud/toolforge/builds-cli/-/merge_requests/32 [08:43:52] done :) [08:44:00] thx [08:44:55] do you have a nodejs only app I can try building? [08:45:12] I can use heroku's I guess too [08:46:17] I only have flask + nodejs [08:48:25] used the sample, it works well too :) [08:48:39] nice [08:49:11] hmm, I have not been "assigning" myself on gitlab MRs, should I? [08:49:36] this is ready for review :) https://gitlab.wikimedia.org/repos/cloud/toolforge/builds-builder/-/merge_requests/21 [08:51:41] sweet :) I'll go through the dance with the packaging gods again then give it a review (with some breakfast in between) [08:52:17] 👍 [09:01:04] gave it a quick pre-look because curious :) so now that we pre-inject the buildpacks, does it mean that the injected packs' detect scripts run as part of the detect step? [09:10:51] yes [09:12:08] that's why in the newer patches to the buildpacks (apt and dotnet) I had to modify a bit their bin/detect script, an the lifecycle expects to get a rc of 100 for "not detected" instead of 1 (that considers as unexpected error) [09:37:14] nice [09:38:29] ugh, seems I'm cursed with small bugs in these last few days: now the utils/bump_version.sh script fails before having pushed the commit & generated the tag [09:41:22] *generating the commit, not pushing [09:53:52] dcaro: quick review https://gitlab.wikimedia.org/repos/cloud/toolforge/builds-cli/-/merge_requests/33 [09:57:36] added a comment, it seems to be adding newlines [09:57:47] maybe macos<->linux stuff? [09:59:43] ah yes, you're probably right. let me see if I can get rid of the newlines [10:00:18] not sure if it's an issue though, if the commit message looks good, feel free to go for it [10:03:58] hmm, I see no difference in formatting before/after the change. This is after. https://usercontent.irccloud-cdn.com/file/j8BSJRxM/Screenshot%202023-12-08%20at%2011.03.43.png [10:04:16] +1 from me then [10:08:33] just to make sure, I tried it on my remote server (bullseye) too and it looks the same [10:11:07] awesome [10:11:11] thanks for the quick review :) [10:17:23] blancadesal: copy-paste of yours https://gitlab.wikimedia.org/repos/cloud/toolforge/envvars-cli/-/merge_requests/8 :) [10:20:57] left a comment :) [10:42:40] thanks! [10:42:52] yay, build quota works on toolsbeta now [10:44:25] \o/ [10:44:45] python 3.7 is already unsupported btw, 3.8 will reach EOL 2024-10 [10:46:56] yep, we will upgrade to bullseye/bookworm that have a newer python [10:54:28] is there a reason we can't/won't have a newer version of python than the os default? [10:58:54] because we don't use venvs, but debian packages to deploy the clients (so we need debian packages for non-default pythons, and that's not so well supported by default) [10:58:58] afaik at least [11:00:53] i'm hoping we'll get around to moving the clis to a single binary 🤞 [11:01:52] I was thinking the same [11:02:03] (also to get rid of the copy-pasting of makefiles around) [11:02:17] yesss [11:19:12] I've been adding random tasks to the current toolforge iteration workboard backlog willy-nilly. Don't hesitate to take a broom and sweep them out of there; I didn't use any priority criteria when spawning them there [11:21:37] dcaro: do I have to apt-install the new builds-cli version on tools too or does puppet take care of that? I did the aptly thing to copy them over from toolsbeta [11:21:38] no problem, I don't use the priority either [11:21:51] blancadesal: it needs manual install on the bastions [11:24:14] ok [11:27:12] hmm, I must have missed a step somewhere https://usercontent.irccloud-cdn.com/file/le3yYuUM/Screenshot%202023-12-08%20at%2012.26.56.png [11:27:40] did you copy over the package to the tools-* repos? [11:29:08] https://www.irccloud.com/pastebin/y0qc1Fb0/ [11:29:18] s/VERSION/0.0.8/ [11:29:21] xd [11:29:41] facepalm, etc. [11:30:01] Maybe we should set it in a variable or something [11:30:05] (might be clearer) [11:30:40] nothing can save the person doing lazy copy-pasting xd [11:31:17] A script with input :) `please enter the version>` [11:31:34] going for lunch, cya in a bit [11:32:39] cya, enjoy! [11:46:54] going for lunch too, next up will be the dotnet patch [14:09:25] dcaro: for the maintain-harbor quota job to work, the maintain-harbor user needs to be admin (currently isn't). Do you see any issue with this? [14:10:24] hmm... my main concern is that the script is doing cleanups and removing projects and such, so a bug might remove the `toolforge` project that has all our infra images [14:10:50] if the account is not able to remove things from there, then a bug would just make it fail if it tries [14:11:21] iirc the permissions settings were rather limited, can we add an account that does not have permissions on the toolforge project? [14:12:13] not sure, the actual bits it needs system admin permissions for is getting/setting the default project quota [14:12:36] for individual quotas, it's enough that it is a project admin (which is already the case) [14:13:19] hmpf [14:13:36] yeah that's annoying [14:14:02] agree it's dangerous for the jobs to run as system admin [14:14:17] hmm, what if we don't change the default quota? [14:14:22] just the per-project quotas? [14:14:39] (that we have to do anyhow if the default ever changes) [14:15:02] indeed [14:15:35] sounds good to me [14:15:43] 👍 [14:16:37] am running out of time before going on vacation though and don't want to do sloppy changes. this is not urgent though [14:17:04] dcaro, for php + mono, ala what is being asked here, T320157. What would you recommend for buildpacks? php buildpack + mono apt? Something else? [14:17:04] T320157: Migrate wikihistory from Toolforge GridEngine to Toolforge Kubernetes - https://phabricator.wikimedia.org/T320157 [14:17:22] no problem, feel free to ping me (or anyone) in the task or the MR for followup [14:18:01] 👍 [14:19:14] balloons: looking, on a quick check, I'd use mono/dotnet buildpack + php package (but might depend a lot on the libraries needed and if they are available as system packages) [14:19:26] we don't support both buildpacks (php+mono) currently [14:19:28] ahh, ok, so the inverse of my suggestion :-) [14:19:40] btw. the dotnet buildpack is not yet merged, so not available currently [14:19:58] if they are able to get the mono package and the php buildpack running that's also ok [14:20:04] right.. If it has to be mono buildpack, then it's waiting on that work.. [14:20:19] though afaik mono needs some building beforehand no? [14:20:22] (not very familiar) [14:20:38] Why would you suggest the opposite (which language to buildpack, which to use system lib)? I suppose it doesn't matter, as long as it works [14:23:17] yep, I think it would not matter, but if one needs compiling/pulling dependencies, I suggest using that as buildpack (otherwise you'd have to hack around it by running something like `npm build && npm run ...` in the jobs and such and pulling all the dependencies every run) [14:23:52] I can try playing with it, see if I make it work [14:25:10] I'll write a reply; you are free to comment as well. Maybe the mono buildpack will be the easier route [14:26:29] Ahh, I think I understand the logic now.. Mono is more likely to need compilation and depends versus php just needs it's interpreter present. [14:27:00] ... I think that they don't have the code on git at all [14:27:14] balloons: yep, :) [14:27:45] `tools.wikihistory@tools-sgebastion-10:~$ find . -iname .git` comes empty xd [14:27:53] so I guess that might be a first step [14:30:27] dcaro, don't worry about searching too hard. I just left a comment asking where the source code was. It's so hard to give general advice specific enough to fix the problems they see without source [14:31:01] yep [14:39:38] note to self: it is indeed a good idea to click on 'submit review' after leaving review comments – no need for them to marinate for hours [14:41:12] I'm glad to see that other teammates have similar hobbies to mine [14:42:10] xd [14:51:53] * dcaro off for today [14:52:05] cya! [17:18:32] It would be awesome to update https://gitlab.wikimedia.org/toolforge-repos/precise-tools to include how many tools have migrated over the last few years. In March 2022 when we deprecated (for the last time :-) ) grid engine, ~850 tools were on the grid. Since that announcement, roughly 407 have migrated in total as of today. Since we formally asked [17:18:32] everyone to migrate in October 2023 (650 tools present), roughly 207 have migrated. And finally since Komla sent the shutdown date information (490 tools present), roughly 47 have migrated.