[01:01:06] halfak: ok, let me check [01:01:39] valorar calidad sounds right [13:13:08] Oscar_, my concern about "valorar calidad" is that it seems to be too general. E.g. we might load in whole articles and ask editors to assess their quality some time later. [13:13:27] I'd like a term that represents that we're looking at the quality of an editors "edits". [13:13:35] o/ Amir1 [13:13:42] Just getting ready to deploy the new models now. [13:13:52] \o/ [13:13:55] hey [14:17:54] Going to staging [14:24:04] went smooth. Now to go to deploy [14:24:45] * halfak loves these config change deploys. [14:24:47] Nice and simple [14:31:07] If our uwsgi restarts were faster, this would be done already [14:31:18] * halfak waits patiently for 1 minute and 30 seconds ... [14:32:01] OK... deployment to workers complete. [14:32:07] Waiting on last uwsgi restart [15:02:35] OK Done! [15:25:44] \o/ [15:26:00] halfak: we need to get this merged to reduce the wait time to a second [15:26:01] * halfak updates wiki pages with new model details [15:26:12] +1 for reducing wait time [15:26:25] https://gerrit.wikimedia.org/r/281161 [15:26:43] the good thing is that wikilabels are now using die-on-term [15:26:59] so their wait time is reduced now [15:27:00] Yeah. Saw that. Wish we could get both merged :/ [15:29:33] We'll bump our support above 20 wikis this weekend :) [15:29:42] oh about the bug you told me about it: https://phabricator.wikimedia.org/T130872 I have a guess which is not my area of expertise but I think it might be related to database issues, I checked and I think if madhuvishy can take a look at them since she put them up, it would be great [15:29:56] \o/ [15:31:14] +1 have been suspecting database issue too. [15:31:32] It would be great if we could replicate it, but I can't :( [15:31:43] Will be nice if I can make it happen and then look at the logs though. [15:34:55] it happens all the time [15:35:30] these cases are all around the logs [15:35:39] let me show it to you how I found them [15:37:38] less /var/log/syslog | grep -iE "[0-9][0-9][0-9][0-9] +msecs" [15:38:02] shows things that surpassed 10 seconds [15:38:55] today it happened once since I enabled harakiri [15:45:06] Amir1, do you have this pasted in phab somewhere? [15:45:19] Is this the log lines you already have pasted? [15:45:34] I found those lines using this grep [15:46:30] Gotcha. [15:46:39] Maybe we ought to turn up logging for a little while. [15:46:50] It'll fill the logs with debug, but it might help us diagnose the problem. [15:47:00] * halfak posts ^ in phab [15:47:07] awesome [16:05:11] halfak: are u there? [16:05:18] o/ Oscar_ [16:05:23] hi [16:05:42] one question, what do you mean by 'artifacts'? [16:08:04] I'm trying to translate the page :) [16:08:48] artifacts is a fancy word for "things" [16:08:57] It is intended to be just as general. [16:09:20] Wikilabels could be used to label edits, whole pages, user's activities, a talk page discussion, etc. [16:09:44] Whatever you might be able to wrap up and call a "thing" is fair game. [16:11:48] Ok [16:51:35] halfak: I think Amir1's fix for the slow uwsgi restarts just got merged [16:52:59] yup [16:53:03] awesome [20:21:51] So, this happen (after the OAuth authorization in w:es) [20:21:54] { [20:21:54] "error": { [20:21:54] "code": "forbidden", [20:21:54] "message": "OAuth callback" [20:21:54] } [20:21:55] } [20:38:15] o/ Oscar_ [20:38:37] Did this happen in the labeling system? [20:38:43] Is it persistent or did it just happen once. [20:55:43] halfak: I was expecting that if I use same word list in both hindi.py and test_hindi.py tests would pass, but I'm getting https://dpaste.de/UEDK [20:56:11] codezee, can you sumbit a WIP PR? [20:56:17] That'll help me look [20:56:24] ok [21:05:51] done [21:10:13] codezee, it looks like some words have a space at the end. is that just a rendering issue? [21:13:09] halfak: the spaces somehow seem to have gotten introduced while copying :/ I'll fix it and see [21:13:16] kk [21:16:12] halfak: it is persistent [21:16:26] Oscar_, I can't seem to replicate it. [21:16:29] Where do you see the error? [21:17:48] After I click "conect to the server" and allow the OAuth authorization (in the pop-up that opens) [21:18:16] Logging in with User:Oscar_? [21:18:35] yeah [21:18:47] Oscar_.* [21:20:15] Oscar_, can you try fully logging out of Wikipedia and back in? [21:20:20] In the meantime, I'll check on other things. [21:21:40] ok [21:21:54] Yeah... everything looks OK. [21:21:58] Checking where that error comes from [21:23:37] Here's where it comes from: https://github.com/wiki-ai/wikilabels/blob/master/wikilabels/wsgi/routes/auth.py#L45 [21:23:50] If this isn't working, something is seriously broken with OAuth [21:24:01] If logging out and back in doesn't work, I still have ideas [21:31:05] halfak: still getting it [21:31:15] OK. What browser are you using? [21:31:23] Chrome [21:31:27] Awesome. [21:31:33] Are you familiar with the developer console? [21:31:49] yes [21:32:37] Cool. So I want you to try the OAuth "connect to server", but when the popup appears, open the developer console in that tab and select the "network" tab before approving Wiki labels. [21:33:59] ok, I'm on it [21:34:39] done but tests are failing still, I'm new to this system so unable to see the issue :( [21:37:01] codezee, push your changes to the PR and I'll take a swing at it :) [21:37:11] halfak: done :) [21:37:14] already [21:38:31] Cool. Will get back to you Saturday morning (UTC-5) if not sooner. [21:38:37] codezee, ^ [21:39:09] sure, I'll amend after seeing your feedback [21:44:32] halfak: got the log error, what do u need? [21:45:20] OK. Under network, the last request -- the one that goes to labels.wmflabs.org/auth/callback/ I need as much detail as you can grab. [21:45:24] So maybe a screenshot [21:49:53] ok [21:53:21] Oscar_, cookies! [21:53:31] Do you have cookies blocked for labels.wmflabs.org? [21:53:40] I think that might be the problem. [21:54:45] checking [21:57:57] halfak: https://drive.google.com/file/d/0B2zaSAywvwH2WkRLUVo1cnNqczA/view?usp=sharing [21:58:26] Hmm... it seems to be sending the beaker session ID [22:00:49] Yeah! I can replicate it by deleting my beaker cookie mid-flight [22:01:01] You might have to enable cookies for labels.wmflabs.org [22:01:21] Oscar_, try to go to http://labels.wmflabs.org/gadget/ [22:01:40] If you have a browser extension that manages your cookie permissions, it should turn on once you get there [22:01:57] Actually, better to do the https path: https://labels.wmflabs.org/gadget/ [22:02:01] Finally! [22:02:05] \o/ [22:02:26] I'm just about to deploy a change that will flag cookies as an issue on this error message :) [22:02:30] Thanks for working with me on the issue! [22:03:03] np :) [22:06:28] o/ YuviPanda [22:06:45] I want to make puppet write a file with the DB password for Wikilabels. Where do I configure that? [22:06:59] unfortunately you can't. on labs we have no private things for puppet [22:07:13] lol k [22:07:19] I guess I'll just manually copy the file! [22:07:20] halfak: me and madhuvishy worked out a compromise using gerrit private repos [22:07:29] halfak: I think manually copying file is easiest atm [22:07:33] kk [22:38:16] YuviPanda, I'd like to push this change out to wikilabels soon: https://gerrit.wikimedia.org/r/#/c/282296/1/modules/wikilabels/manifests/web.pp [22:38:26] I'm trying to standardize how it works so that it looks like ores. [22:38:38] I imagine this might cause problems with the live service [22:39:53] Maybe we could disable puppet on wikilabels-01, merge this and then let me test things out on staging [22:40:02] halfak: just need to co-ordinate it I guess. you can do the deploy first (since uwsgi doesn't check files without restart) and then run puppet [22:40:26] Deploy will restart though, right? [22:40:39] And is uwsgi is looking in the wrong place... [22:41:21] halfak: ah, just manually do a git pull maybe? [22:41:41] That'd work. I'd like to get this merged and test on staging though [22:41:51] on quarry I have 'update_git' and 'restart_uwsgi' as separate functions for these cases [22:41:51] Maybe I should just manually modify the uwsgi config on staging? [22:42:40] YuviPanda, even the separate update_git doesn't address the puppet change that, I assume, would also result in a restart of uwsgi [22:43:05] halfak: indeed, so you 1. update git, 2. merge puppet change, 3. run puppet manually [22:43:17] restarting uwsgi shouldn't be a service disruption [22:43:41] YuviPanda, yeah, but I want to test on staging before going to deploy. I don't see how that happens without disabling puppet on deploy. [22:43:53] ah. indeed. disable puppet, do it by hand etc [22:44:30] OK. i think I've got this. sudo puppet --disable, merge puppet, run tests on staging, sudo puppet --enable [22:45:22] *sudo puppet agent --disable [22:45:34] OK. I have puppet disabled on wikilabels-01 [22:46:18] have a minute to merge https://gerrit.wikimedia.org/r/#/c/282296? [22:46:24] YuviPanda, ^ [22:46:42] * halfak would bug madhuvishy but it seems she is away [22:46:49] halfak: merged. she can't merge anyway [22:47:06] Gotcha. Thanks. [23:04:23] OK here we go [23:06:00] And done! [23:06:05] Thanks YuviPanda!