[01:42:53] @rhinosf1 since you're the person that apparently blocked the bot in the first place, i just wanna refer you to T12417 [01:43:10] not super high priority but just figured i'd get it on your radar 👍 [02:28:46] was reading https://wikitech.wikimedia.org/wiki/Kafka#Multi_Datacenter and read kafka mario maker [03:37:44] https://issue-tracker.miraheze.org/T12417 [03:37:50] if you want a direct link [06:15:16] Ack, not really convinced yet as to why we should unblock it [06:15:38] If we've blocked it, then we had probable cause to think it was likely to cause an outage if it carried on [06:15:57] I might be able to set a really low rate limit on it [06:16:12] Instead though [06:18:01] I don't think I'm just going to hope it follows robots.txt [06:20:15] I was originally going to post this on the Phorge task, but since I receive an unhandled exception when I do so I'll post it here: https://ahrefs.com/robot and https://yep.com/yepbot/ are relevant sources I found after a quick browse [06:23:40] I'll have a look at logs [06:37:43] <.labster> Every time I think our website has terrible performance, I try to load WikiApiary. I mean, seriously, something over there must be O(N²). [06:39:37] "so how's the database move going?" [07:08:42] as @tali64 mentioned, it does in fact follow robots.txt [07:09:15] currently the way it’s configured, MH wikis are entirely unsearchable on yep [07:10:57] Doesn't mean I trust it to [07:11:54] fair [07:12:02] though it’s following robots.txt as we speak [07:14:42] anyways point being [07:14:54] I feel it’s something we should at least test and experiment with [10:31:32] @rhinosf1 what are the requirements for cloudflare to recognize that X domain is pointed to us? When does Cloudflare generate a certificate for it and all that? [10:32:06] HTTP verification is used [10:32:19] but we can't work out how to do CNAME at the apex using GDNSD [10:32:30] so only people using CNAMEs can migrate at the moment [10:33:22] so when someone using some DNS that's not Cloudflare points to cf-lb.miraheze.org what happens? [10:34:51] we add them as a custom hostname and it does verification over HTTP [10:35:11] we have to add them via CF's dashboard then? [10:35:18] I see... [10:35:52] ye [11:08:42] [1/2] maybe write a gdnsd plugin that does CNAME flattening CF style? I don't think gdnsd will like this though. [11:08:42] [2/2] Use SVCB? Alias mode SVCB is specifically made to allow basically CNAMEs at the root domain: https://www.rfc-editor.org/rfc/rfc9460.html#name-aliasmode, too bad there's no SVCB support yet https://github.com/gdnsd/gdnsd/issues/242. We could write a plugin for it, but BIND does support SVCB records since 9.16.21. [11:09:11] All options [11:54:49] @bluemoon0332 can you get me a list of all custom domains [11:56:47] or @cosmicalpha [11:57:15] I'll add them so we can move mw-lb [11:57:28] I can get you a list of all wgServer fields tomorrow if OS doesn't beat me. [11:58:00] I can't get a list another way, unless you have a way for only mw-lb wikis? [12:00:27] @cosmicalpha manual check against dns repo [12:04:59] huh? dns repo? I thought you wanted non-dns ones, make a python script checking ns if pointed ns skip then get the wgserver for only mw-lb was my idea, easier than manual but that is just me lol [12:05:35] i was just going to look at what domains didn't exist in dns [12:05:38] and then add them [14:11:54] @rhinosf1 I'm going to make a origin certificate on cloudflare for *.miraheze.org [14:12:13] if the Sectigo-signed wildcard works, an Origin CA-signed wildcard should as well [14:12:23] we also save money on having to renew it in the future [14:12:32] We pay for certs? [14:13:01] Origin CA certs for all traffic would be ideal [14:13:58] We don't [14:14:32] It just something something available to CF by default [14:16:47] the amout of security events blocked shot up [14:16:54] We do [14:17:03] from prior to having cf [14:17:32] Hm? [14:17:41] Didn’t we use LE? [14:17:51] not for *.miraheze.org [14:18:03] What [14:18:04] Why [14:18:05] looks to be a mix of Amazonbot and Go [14:18:07] we don't use CF's origin cert? [14:18:13] Go! [14:18:15] nope [14:18:25] yes, Go is really popular with idiots [14:18:40] Like [14:18:43] The lang? [14:19:41] ye [14:20:47] Time to make a config PR while my breakfast cooks [14:20:58] Why is my dining room so dark [14:24:30] @kiju1108 does change this require approval from community or stewards or tech to smt [14:27:45] Probably requires approval from tech but I'm not sure [14:28:01] @reception123? [14:28:08] For https://issue-tracker.miraheze.org/T12526 [14:29:26] @bluemoon0332 so now we just need to deal with our dns [14:29:37] yep [14:30:16] fair drop on https://grafana.wikitide.net/d/arhCmd7Mz/nginx-cache-proxies?orgId=1&refresh=5s&from=now-3h&to=now [14:31:16] Good or bad? [14:31:57] good [14:32:41] coderabbit doesn't like os.system https://github.com/miraheze/puppet/pull/3905 [14:33:05] it's not wrong [14:33:28] it shows how much traffic to cp* dropped [14:34:19] Hel yea [14:34:55] we want it to say 0 eventually [14:35:24] Mod mod [15:45:42] How do I add an interwiki link to my wiki in the [[Special:Interwiki]] table? [15:45:43] [15:50:13] I believe you are already bureaucrat in your wiki and have access to ManageWiki, you will have to give yourself the permission to edit them [15:50:21] You need to grant yourself the interwiki right in special:ManageWiki/permissions [15:50:32] Ok, thank you. [15:51:35] pizzatower.wiki seems to work fine, but gogigantic.wiki straight up won't open [15:51:58] unable to make safe connection etc [15:57:23] `ERR_SSL_VERSION_OR_CIPHER_MISMATCH` at gogigantic.wiki [15:58:23] For how long [15:59:35] [1/2] I checked just 5 min ago [15:59:36] [2/2] 5 hours ago were no problems, I blocked a spam account and reported in #cvt [15:59:51] I'm fixing my TV [15:59:54] Then I can look [16:17:19] fixed (cc @theoneandonlylegroom ) [16:22:34] for some reason it had got stuck [16:29:48] thanks [16:50:19] @rhinosf1 cf sends a header back on a curl right? [16:50:34] ye [16:51:44] Okay so it’s not behind yet [16:51:55] Thought we had put them all behind CF already [16:52:54] everything that's using CNAME [16:53:07] and everything not at the apex of the domain [16:53:17] Apex? [16:53:26] not got aa subdomain [16:53:41] @bluemoon0332 what's your plan for updating SSL [16:53:52] So all root domains should not be [16:53:56] Okay [16:54:05] well we should be able to migrate to BIND before the LE certs expire hopefully [16:54:05] Got confused [16:54:25] then we can just add the origin cert wildcard to all of them [16:54:35] well they expire at different times [16:54:41] unless it's on https://github.com/search?q=repo%3Amiraheze%2Fdns%20%22%40%09%09DYNA%09geoip!cp%22&type=code [16:54:46] everything else is migrated [16:55:35] btw here's that cert: https://github.com/miraheze/ssl/commit/23a47dba1ebe96e9c1e4b272ce08fbf0c3e495ac [16:59:05] @bluemoon0332 anything with a subdomain can be moved now [16:59:25] only thing not moved is stuff using our DNS at the root of the domain [17:01:06] we can switch them all over to that cert then [17:02:13] What’s the difference between CNAME and using our dns servers when [17:02:15] Even* [17:02:26] <- does not understand certs and dns [17:02:29] yup [17:02:42] because GDNSD doesn't support CNAME at the root [17:03:04] what the FUCK is a GDNSD [17:03:09] I’m dumb remember [17:03:44] GDNSD is what both us and wikimedia use for DNS [17:06:18] So software [17:06:29] And again we fall into the trap of being limited because we copied wikimedia [17:06:35] Recurring theme [17:08:13] We can move to something else [17:08:17] we probably have to [17:08:39] Rip [17:09:06] BIND supports what we need for this: https://issue-tracker.miraheze.org/T12518#250374 [17:09:27] how hard would it be to move to BIND [17:09:33] I don't know actually [17:09:55] but it is a quite famous daemon so it won't be for a lack of docs or blog posts about it [17:10:33] @bluemoon0332 are you insane enough to try it? [17:14:20] This is miraheze [17:15:39] So are we going to have to move to a whole other software for the apex wikis? [17:16:11] Probably [17:16:29] Drat [17:16:30] We don't really need to host our own dns tbh [17:16:57] Time for a side quest? [17:18:31] So using a third party service instead of moving to a new software of our own [17:19:19] Just use free cloudflare [17:19:29] oh [17:19:52] so pretty easy to move? [17:20:13] Well you only need to set a CNAME [17:21:56] so would everyone need to switch their domains [17:25:33] not all, except for those who don't want to move, which they will have to repoint their domain to an legacy address [17:28:04] If we stop offering dns at some point then yes but I'm not sure whether we will [17:28:57] One possible idea is move to custom nameservers in Cloudflare [17:29:04] So we control them [17:30:00] Please only use mw-lb-legacy if things are broken until we fix it [17:30:15] what would that mean effectively [17:30:28] Doesn't matter if we don't do it [17:30:47] mw-lb-legacy is going as soon as all wikis are successfully working [17:32:56] I mean what would not offering dns mean [17:33:59] Either that we create everyone cloudflare tenancies for their domain which is the likely option or you find your own dns [17:34:37] honey the cache servers are on fire again! [17:34:40] dear god [17:34:42] 7 down [17:35:16] How [17:35:25] They are getting half the traffic they were this morning [17:35:44] we breached 2k unclaimed jobs [17:35:57] oh hell nah [17:36:16] actually [17:36:17] Is it controlled though? [17:36:29] the jobs seem to be the cache regenerating [17:36:55] All but one of the top jobs os parsoidCacheRewarm [17:37:06] huh [17:37:11] the chart says otherwise [17:37:24] never mind the pie chart is dumb [17:38:12] Meh [17:38:40] this is fine? [17:38:50] Meta is a wee bit slow but the error dropped off [17:38:57] Yes [17:39:08] ye [17:41:35] Unclaimed jobs are dropped [17:41:39] 1.5k [17:41:48] all cp are up [17:41:52] we’re good [17:43:20] woo [17:45:32] @pancake.aurora do you still have your script for cleaning the dns repo out? [17:45:49] ProTip: it's 100% manual process [17:45:59] I just did it all by hand [17:46:06] @pancake.aurora wait, that wasn't automated at all? [17:46:13] basically `whois`ing them all [17:46:18] Cause I was gonna say can you run it again [17:46:36] To save me trying to work out what domains in dns actually still point at us [17:48:13] I think it can be quickly automated but not in front of my keyboard rn [17:48:59] (At least quickly check for `dig ns $domain`) [17:58:09] I cleaned up dns a couple months ago, but was all manual [18:00:31] Meh more than a couple months ago [18:00:36] (April) [18:00:44] https://github.com/miraheze/dns/pull/511 [18:01:05] But it removed a total of 79 zones [18:03:02] Some may be pointed but not actually attached to a wiki [18:22:58] Seems like we still need to get things such as Mirabeta under CF [18:29:25] MacFan4000: mirabeta is a domain that is using our dns so it'll be moved at some point [18:30:07] test151 should probably be renamed as test151.mirabeta.org tbh [18:30:16] Not wikitide.net [18:30:35] MacFan4000: ye I want to clean them up again [18:30:45] So we can see wikis that aren't working and remove them [18:31:01] So I don't have to deal with migrating their dns and it not working [18:34:39] Beta is a special case though [18:34:47] Cause it doesn't use the load balancer too [20:17:57] rhinos can you publish my msg in ma [20:18:06] discord being an ass [20:18:18] No [20:18:22] Don't see the option [20:19:51] what in the nine hells [20:26:55] Discord moment. Guess replies can't be published or something [20:27:08] Hey @orduin [20:27:20] 👋 [20:27:52] @orduin we have stacks more behind cloudflare [20:27:59] And amazonbot really was badly scraping [20:28:08] I'm not surprised that caused issues [20:28:16] We have a lot left to do though [20:30:11] blocked? [20:30:19] Yup [20:30:25] Cloudflare blocks it [20:30:27] A lot [20:30:27] actually what is the pending fix [20:30:35] It got merged [20:30:45] if not just the butchering of user agents [20:30:48] what was it then [20:31:29] https://cdn.discordapp.com/attachments/1006789349498699827/1279176691238834206/IMG_8406.png?ex=66d37da1&is=66d22c21&hm=c89571ce0841b0015cc93e1f58426538cfa91a10159477030b2a53467d09aa67& [20:31:46] https://cdn.discordapp.com/attachments/1006789349498699827/1279176761799741492/IMG_8407.png?ex=66d37db1&is=66d22c31&hm=e7982b43f47a599b6fe04890bc3cb6d960247da91102212810900fbfa6828cd4& [20:31:58] Thats the fix in action? [20:32:01] @pixldev that huge blue jump is amazonbot [20:32:11] From when we switched custom domains earlier today [20:32:32] what the [20:32:33] That increase in events is just from amazonbot on cname'd custom domains [20:32:34] explain [20:32:55] But that’s when they were moved? [20:32:57] im confused [20:33:21] Well cloudflare didn't block Amazon when traffic wasn't going through cloudflare [20:33:36] Ooooooh [20:33:37] jesus [20:33:51] you mean it's actually started dedicting amazonbots? [20:33:59] so the main slowless issues should be stopped? [20:34:14] We've fixed them yes [20:34:19] We fixed them yesterday [20:34:32] What was the fix proper [20:34:55] We moved 350 or so custom domains behind cloudflare today so ye cloudflare's firewall took effect on them [20:35:00] Banning amazon [20:35:03] Oh urlshortern is global [20:35:07] I'll shout at them at the end of the month [20:35:09] oh lmao [20:35:15] abolish amazon [20:35:27] okay so url allows you to shorten any domain already [20:35:36] you just need to be on it [20:35:44] ie on my wiki i can make links for my wiki [20:35:54] i dont like that but is it worth changing [20:40:53] As far as I understood, there was some breaking changes for the UploadWizard that we completely missed in REL_42 [20:41:36] That's only for custom domains isn't it [20:45:06] Let me know if there's anything that needs doing immediately, otherwise, I'm probably just going to continue taking it easy while seasonal allergies do horrible things to me. [20:45:57] Did your multi purge patch ever get merged? [20:46:53] haven’t tried, this was on a non custom [20:51:12] @rhinosf1 appearntly making short urls isn’t a default permission [20:51:14] i dont like this [20:53:51] It's in their develop branch, not sure if that's stable for use in production. One of the commits on the branch is explicitly marked as WIP. Might be worth looking over, or might be worth getting custom patches done. [21:02:56] Could we look at deploying it in some way? Event if it's temp using a fork you have of it. Or deploying a way to do temp patches or something. [21:03:32] We do need to consider moving to the packaged version of the deployment tool eventually but I'm fine with merging your puppet patch if you test it for now to unblock that [21:03:40] What you gonna do about it [21:06:27] wait some years or until the board allows U18 NDAs run for steward and change the setting? [21:07:43] Can't you just change the defaults in managewiki [21:07:54] On my wiki yea [21:08:00] Globally [21:08:09] Yea [21:08:12] You can change what ManageWikiDefaultPermissions does [21:08:13] A steward can [21:08:15] I ain’t one [21:08:20] Via config [21:08:24] I think [21:08:30] Or is that only on wiki [21:08:32] Hmm [21:08:55] idk i think so [21:09:18] plus it don’t change existing ones [21:10:43] Cough interwiki [21:41:16] erm, just making sure, is https://issue-tracker.miraheze.org/R9:601834a3d0ed2efc8576adef80ed6a10f4e7471e already deployed? [21:41:29] because curl -v https://rainverse.wiki/w/index.php?title=MediaWiki:Gadget-JWB.js\&action=raw%O still redirects me [21:41:47] thanks konversation for handling that properly: `curl -v https://rainverse.wiki/w/index.php?title=MediaWiki:Gadget-JWB.js\&action=raw` [21:42:17] lol not anymore: https://github.com/miraheze/mw-config/commit/d6487cebb4e6 [21:42:54] it was never actually deployed as the deploy failed [21:43:12] oh okay--what for though? [21:43:31] oh amended commit message: Didn't do anything. I think this may need done in decode.php instead. [21:46:54] I deployed it to test151 manually and it didn't do anything. I think it needs to be in decode.php instead. [21:48:18] ah, I was going by the puppet failures on mwtask* [21:48:49] Oh yeah index.php only redirects /wiki to / and vice versa, decode.php handles queries, @BlankEclair [21:50:32] oh wow [21:53:20] i wonder if we could do it in nginx... [21:53:22] https://stackoverflow.com/questions/26133592/how-to-get-query-parameter-in-lua-or-nginx [21:54:04] I tried for like a month when first making that lol, I decided I knew PHP better so went with this fairly messy method. [21:54:11] perhaps something like if ($http_action = raw) { break; } [21:54:43] You can just do the same change you did to index.php but apply to decode.php I think? [21:55:00] but all of what decode.php does is redirect you [21:55:07] if i stop the redirect there... the user gets nothing [21:55:23] or perhaps i could call index.php from decode.php? feels hacky-ish though [21:55:56] The whole index.php/decode.php we have is already very hacky [21:57:13] i mean, i guess i could try to do it in php if you want me to [21:57:21] i feel like it's just rather hacky [21:57:38] Idk lol, nginx may not work. [21:57:45] but if it does we can do that [21:57:55] Anyway it is 6AM and I didn't sleep... whoops [21:58:23] worth a try i guess [21:58:25] also oopsie