[00:00:29] We fixed it within 6 hours, but MapBox didn't and it bit a lot of their users: https://nymag.com/intelligencer/2018/08/why-was-nyc-renamed-jewtropolis-on-snapchat.html [00:01:06] that's what happens when you only have fancy "ai" systems determining if changes should get through [00:01:33] They have users doing it, but unfortunately not enough people looking at their "recent changes" feed. [00:01:40] Which is a situation I'm sure we can empathise with. :-( [00:01:48] * AntiComposite edits OSM as well [00:14:44] and it's probably not easy at all to validate geo changes if you don't know the zone [00:18:26] If I remember correctly, mapbox uses two methods to determine if a changeset (edit) is harmful: one is stats and AI, the other is time. The more the stats/AI say the changeset is harmful, the longer it will have to stay before Mapbox will include it [00:49:12] AntiComposite: there's not been a policy decision about whether or not to 'fully' block external usage of maps.wm.o -- right now it will still work if you happen to request a maptile that is in the CDN cache (which happens to be a lot of them) [00:50:15] Yeah, I told them that it's currently disabled for an unknown period of time and that a different tileserver would be a good idea [00:50:43] perfect, thanks :) [01:10:28] "Funny", just the other week I sent https://lists.wikimedia.org/pipermail/maps-l/2020-January/001704.html [01:12:01] Because I think that even without a clearcut policy it's helpful to come up with a couple examples of obviously ok and obviously bad usage [14:01:35] hey, i'm currently getting more and more http 429 from maps.wikimedia.org osm-intl tiles. i haven't found matching phabicator tasks yet, but the daily-tile-requests graph on the interactive-team-kpi dashboard in grafana seems to show also a large decrease in requests. was a new policy or config change introduced recently? [16:16:43] robbi5: not a new policy, although we're considering one. Maps has had several outages recently where one of the causes is a chronic underprovisioning of the service -- the servers are often running in excess of 75% cpu utilization and sometimes pegged at 100%. we've temporarily started returning HTTP 429 for non-Wikimedia-wiki users in the event that the tiles they're requesting aren't already in [16:16:45] the CDN cache, as an outage mitigation strategy [16:17:16] I wish I could tell you for how long that will be in place, but I don't know that [16:35:42] cdanis: oh. okay, thank you :) we've using tiles with a high zoom level, so thats why the tiles aren't cached and we get the 429. can you add this note somewhere on a wikipage of the map service? i've seen people with the same question even in the software repo: https://github.com/kartotherian/kartotherian/issues/121 :) [17:02:51] robbi5: ah, good to know, thanks. I'll at least update the Phab task from the outage with details about the current situation, and communicate that around [17:03:47] Does any know an alternative to wgHostname that gives the data centre as well? [17:04:10] I have the code in https://meta.wikimedia.org/wiki/User:RhinosF1/global.js [17:05:23] Which returns 199ms (mw1319) but I want something like 199ms (mw1319.server.wikimedia.org) [17:05:38] Where server is equiad, essams etc [17:07:19] not sure if that exists, but I suppose something like [17:07:21] ( { 1: 'eqiad', 2: 'codfw', 3: 'esams', 4: 'ulsfo', 5: 'eqsin' } )[ mw.config.get( 'wgHostname' ).match(/(\d)\d{3}/)[1] ] [17:07:23] would work [17:07:28] (list based on https://wikitech.wikimedia.org/wiki/Infrastructure_naming_conventions#Servers) [17:07:30] that's the hostname of the mediawiki server which means two things, 1) the FQDN will be something like 'mw1319.eqiad.wmnet' which isn't a real TLD, 2) likely it will only ever be eqiad or codfw, 3) mw1* will always be eqiad and mw2* will always be codfw [17:08:12] I'd guess that we don't expose the '.eqiad.wmnet' portion because it isn't meaningful outside WMF private networks, and also doesn't add information [17:08:59] cdanis: then I can probably just note it and remember. I just wanted to be able to tell which centre it was serving it [17:10:24] yeah, what Lucas said will work well for you [17:10:41] this of course doesn't tell you anything about the CDN servers involved, but that's in the X-Cache response header if you need it [17:11:04] yeah, I’m basically exploiting 3) of what cdanis said [17:11:16] that the first digit of the server name gives you the dc [17:13:18] <_joe_> RhinosF1: https://config-master.wikimedia.org/discovery/discovery-basic.yaml tells you which datacenters are active currently [17:13:50] <_joe_> the - dnsdisc: "/appservers-rw" [17:14:08] <_joe_> and - dnsdisc: "/api-rw" are the records you're interested into [17:14:26] <_joe_> will tell you which datacenter is serving mediawiki traffic right now [17:15:12] <_joe_> if you want to know which applayer is responding you [17:15:39] Thanks _joe_ [17:15:46] <_joe_> for what you want, though, the easiest way to get that info is [17:15:51] <_joe_> mw1* => eqiad [17:15:56] <_joe_> mw2* => codfw [17:17:32] What Lucas_WMDE gave works [17:17:53] <_joe_> yep [20:25:37] Commons doesn't like me today [20:25:38] Request from - via cp3064.esams.wmnet, ATS/8.0.5 Error: 502, Next Hop Connection Failed at 2020-02-06 20:24:52 GMT [20:25:54] And even a good old bare: 504 Gateway Time-out nginx/1.13.9 [20:26:05] uh [20:26:08] did wp die again [20:26:15] Nemo_bis: yeah, not just you [20:26:21] It's not personal :) [20:26:27] I'm glad ;) [20:26:46] #rip [20:26:59] gone down for me too [20:28:14] (ops called in en masse; revert upcoming) [20:29:19] who b0rked it [20:30:02] thanks for the reports, team is already working on it <3 [20:30:08] I'm getting major issues trying to get the upload forms to try to load on Commons right now, too [20:30:50] It'll take a little time for the revert to go through. [20:31:22] Nothing works now [20:31:32] :/ [20:31:35] yah this is pretty bad [20:32:17] en.wikipedia works kind of [20:32:38] works for me again [20:34:01] looks like we're back [20:34:36] Meta was just slow actually trying to load the page with my issue on but I broke paladox|UKInEU’s response time script [20:34:49] everything should be recovered, let us know if you're still seeing issues [20:35:04] I’m getting ERRORJavaScript parse error: Parse error: Missing ; before statement in file 'User:RhinosF1/global.js' on line 13 on [20:35:05] script? [20:35:06] https://meta.wikimedia.org/w/index.php?title=User:RhinosF1/global.js&diff=19779560&oldid=19779344 [20:35:18] paladox|UKInEU: from your mh global.js [20:36:04] it's not mine, it's SPF|Cloud [20:36:46] paladox|UKInEU: fair enough [20:40:50] paladox|UKInEU: any idea how to fix it? [20:42:51] maybe a missing comma [20:43:11] paladox|UKInEU: where exactly? [20:43:25] dcas = ( { 1: 'eqiad', 2: 'codfw', 3: 'esams', 4: 'ulsfo', 5: 'eqsin' } )[ mw.config.get( 'wgHostname' ).match(/(\d)\d{3}/)[1] ] -> dcas = ( { 1: 'eqiad', 2: 'codfw', 3: 'esams', 4: 'ulsfo', 5: 'eqsin' } )[ mw.config.get( 'wgHostname' ).match(/(\d)\d{3}/)[1] ], [20:44:30] Nope paladox|UKInEU [20:44:57] https://meta.wikimedia.org/w/index.php?title=User:RhinosF1/global.js&diff=19780060&oldid=19779560 gets the same error [20:45:38] maybe caption = respTime.toString() + 'ms (' + server + '.' + dcas ')'; -> caption = respTime.toString() + 'ms (' + server + '.' + dcas + ')'; [20:48:06] paladox|UKInEU: yep thx [21:59:38] got tired of getting randomly disconnected due to "invalid captcha" so I switched to a standalone IRC client instead of webchat