[08:48:58] I sent a report about my Wikifunctions work yesterday. [08:49:36] This will be useful to keep an archive of the discussion. [08:50:20] It is in the mailing list. It is entitled Abstract Wikipedia and Arabic. [08:50:42] However, it involves general comments. [10:47:49] I’ve worked around this for now, mobile view should work again (re @wikilinksbot: T291325 – TypeError: Return value of BaseTemplate::getCopyrightIconHTML() must be of the type string, array returned [open]) [11:29:07] It works. (re @lucaswerkmeister: I’ve worked around this for now, mobile view should work again) [11:29:18] Thank you. [14:32:19] FYI, I’m also working on the way the function orchestrator+evaluator are deployed, so there might be some downtime throughout the day [15:38:49] alright, updates to the function orchestrator and evaluator should work without downtime now (re @lucaswerkmeister: maybe I’ll look into it as a kubernetes exercise) [15:39:51] Nice! [15:40:21] Document that into Setup for us to learn too [15:40:31] right, I should update that [15:40:53] and probably just put ~/k8s/.git on GitHub at this point, so it’s more easily visible [15:41:22] Yeah! [15:42:07] So I can help with my limited docker knowledge :) [15:42:17] @lucaswerkmeister: looks like you've been putting kubernetes probes just about everywhere in the last few days? :D [15:44:20] also, your pygments server is failing with "TypeError: a bytes-like object is required, not 'str'" [15:44:38] yeah I just noticed that [15:44:49] while trying to update the Setup page ^^ [15:44:56] There is a pygments server for notwikilambda? [15:45:18] yeah, because the PHP container that MediaWiki runs in doesn’t have Python 3 [15:45:29] so Pygments runs as a separate k8s deployment/service [15:46:35] Maybe jupyter might be more open to more languages? [15:46:53] probably due to https://github.com/wikimedia/mediawiki-extensions-SyntaxHighlight_GeSHi/commit/af6654e5f9d8dee205c1c129e472e1c7797bbcd4 which started calling pygmentize -V, which prints a str to stdout [15:46:56] and I override stdout with a BytesIO [15:47:15] so I need some kind of Python object that’s like a combination of BytesIO and StringIO, with which I can replace stdout… [15:49:07] Maybe TextIOBase? [15:50:24] Jupyter has the concept of kernels which might help encapsulation of more langs in a container [15:50:49] * of kernels (basically rpc servers) which might [15:51:48] jimman2003: Or one step beliw the hierarchy IOBase [15:52:10] I’ll try TextIOBase now, wanted to temporarily disable SyntaxHighlight and then update https://notwikilambda.toolforge.org/wiki/Not_Wikilambda:Setup first [15:52:25] * Or one step below the hierarchy IOBase [15:52:38] Ok,makes sense [15:56:32] btw, mind using some other times than 00 UTC and 12 UTC? it'd be nice to get k8s cluster load spread out evenly through the whole day [15:56:50] sure, any specific suggestions? [15:57:13] (I’m guessing Kubernetes doesn’t have anything like systemd’s AccuracySec=12h) [15:57:20] feel free to be creative [15:57:35] nope, and I'm hoping to get something in the jobs framework before advertising it more widely [15:57:46] makes sense [16:29:02] oh great, SyntaxHighlight doesn’t write the code to pygments’ stdin anymore [16:29:29] instead it writes it to a file, which of course won’t exist in the pygments-server container… [16:29:58] no idea how this is supposed to work in production, since I’d assume the intention behind this work is to move pygments to a separate container there too [16:30:10] but for now I guess I’ll just freeze the SyntaxHighlight extension at whichever git commit used to work [16:32:34] hm, I guess the answer is that in production, the shellbox server will be in the separate container [16:32:49] and since the file is written by the shellbox server, it’ll exist for pygments [16:33:05] but that requires a container with both PHP and Python 3, which I don’t have in Toolforge, so it doesn’t help me [16:33:18] unless I fancy writing an independent shellbox server implementation in Python [16:34:20] There is also Piston [16:35:01] https://github.com/engineer-man/piston which might help [16:39:51] Is that helpful in any way :) [16:40:12] ? [16:42:04] doesn’t look like it at a glance [16:43:41] I thought that was what you meant with the term shellbox [16:44:08] shellbox is a very specific software https://packagist.org/packages/wikimedia/shellbox [16:44:58] Oh, the same idea for shell.. [16:45:30] Piston does support bash... [16:47:57] changed the cronjob to 9:30 and 21:30 (re @wmtelegram_bot: btw, mind using some other times than 00 UTC and 12 UTC? it'd be nice to get k8s cluster load spread out evenly through the whole day) [16:51:59] k8s files are at https://github.com/lucaswerkmeister/notwikilambda-k8s now