[13:17:18] Urbanecm: You have been giving change 557838 a +2, but Wiktionary and Wikiquote for “no” anounce their site as “norsk” and not “bokmål”. It is likely they want to change to bokmål only, but that should be decided by the community and not by a technical fix. [13:18:40] That is, the label “norsk” is correct for those sites until a community vote exist for a change to “bokmål”. [13:19:31] * Jhs waves [13:19:35] thank you Jhs [13:19:44] jeblad asked a question about https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/557838 [13:19:50] 14:17 Urbanecm: You have been giving change 557838 a +2, but Wiktionary and Wikiquote for “no” anounce their site as “norsk” and not “bokmål”. It is likely they want to change to bokmål only, but that should be decided by the community and not by a technical fix. [13:19:52] It could be that a vote exist for Wiktionary and Wikiquote, but I can't find anything about it. [13:20:30] I believe Jhs can answer better than me, so I invited Jhs to discuss that here [13:22:31] I'm not sure there are anything to discuaa, start by asking the communities what they want, don't do this without asking them first. It is pretty clear what the communities at Wikipedia want, but it is not clear what communities at Wiktionary and Wikiquote want. [13:27:32] This patch is merged, but this should NOT be done without asking the communities. [13:43:22] jeblad, looks like you're right about nowiktionary and nowikiquote, i didn't look into that closely enough before making the patch [13:44:00] so I can revert those two in the next window, and ask at their village pumps which one they prefer [13:44:39] Post a note at the sites and wait a few weeks. You will not get a reply overnight. And don't listen to people from Wikipedia, let the communities at those projects decide. [13:44:48] * Jhs knows [13:45:06] It should not be rolled out at all [13:45:28] Going from a patch to swap in two hours are plain stupid. [13:46:04] just a small config change, that's how they're usually done :) [13:48:59] This is NOT something that should be rolled out as a SWAT change. Use SWAT for fixing critical bugs, this is not a critical bug. [13:50:01] In particular, using SWAT for contentious pet projects are not wise, and you knew this was contentious. [13:50:02] that’s… not true at all? [13:50:11] (that’s a reply only to the first message) [13:50:19] normal config changes are fine to do in SWAT, as far as I’m aware [13:50:25] https://wikitech.wikimedia.org/wiki/SWAT_deploys#Guidelines also agrees [13:52:42] Then the page should be rewritten. You don't roll out non-critical changes as swat changes on a production site. That is bad at so many levels. [13:53:32] jeblad, it's not a contentious pet project, it's something that all of nowiki + nnwiki have been complaining about for years and filed multiple tasks about, and now it's finally fixed. but yes, the inclusion of wiktionary and wikiquote might have been too hasty, so we can roll that back and ask the communities just to be safe [13:58:04] * jeblad often wonder if change management, change request, change schedule, and configuration item is known at all [13:58:29] But me gone. [15:12:33] About Toolserver : How to activate web directory on my new tool ? I logged to SSH, become codexbot. mkdir "public" and "public_html", adding index.html but 404 on https://tools.wmflabs.org/codexbot [15:13:33] Can't find the related question on doc [15:14:42] Sorry, juste find Help:Toolforge/Web :| [15:16:05] I should try "webservice start" (but 2min for results after a command like "ls"...) [15:16:23] (#wikimedia-cloud is the better channel) [15:17:12] Yeah, one of the Toolforge login bastions is having trouble, usually due to people running tasks there instead of using the grid or kubernetes [15:18:30] It works ! :D [15:18:55] AntiComposite Yeah, I read the Grid doc [15:19:40] (https://lists.wikimedia.org/pipermail/cloud/2019-November/000847.html for details) [15:21:44] My bot runs some workers continuously from day to day (importing data, with quota of max requests by hours and days). The process can be stopped and rerun. What's the best strategy with Grid ? stop/rerun every 2 hours ? [15:22:11] The process sleep a few hours when quota are depassed [15:23:11] Running on the grid, it's best to have jobs end if they can. That will release all of their resources back to the pool when they're not running and will balance load between servers better [15:23:24] I read Grid doc a few months ago, I must re-read. [15:24:29] AntiComposite : ok, I understand. I use same tactic in case of memory leak or high resources overload. [15:25:31] with supervisor (sort of naive Grid, in my mind)