[00:46:59] how do I know if my toolserver acct is set up? [00:47:13] Try to log in with it? [00:47:26] ok, what is the hostname again? [00:47:31] You ought to be able to SSH into nightshade.toolserver.org or willow.ts.o [00:47:42] Both should work [00:49:02] hrm... was supposed to happen sunday evening, though, I admit I didn't give DaB my ssh keys till then. [00:49:11] *hexmode heads over to #toolserverf [05:10:48] <^demon> [15:25:15] "Find someone other than you who would like this feature before posting it" <-- Bah. [05:15:08] :D [05:19:59] Silly if for no other reason than Wikipedia is large enough for any two people to think even the worst idea has merit. [05:20:14] But we don't need co-sponsors to every enhancement request. A bug is a bug. Judge on its merits. [07:50:06] *guillom wonders if Shirley is still around. [07:50:15] Barely. [07:50:26] Would you have a few minutes for me please? [07:50:33] Sure. [07:50:35] What's up? [07:51:11] I'm going to do some cleanup with the WMF engineering project pages on mediawiki.org, and I wanted to ask you what you didn't like about them. [07:51:29] I know the name of the categories is one item. [07:51:36] Ah. I don't like initialisms/acronyms in page titles. [07:52:07] And yeah, the category names make it seem as though someone was trying to tweet them or something. 2010q4 or whatever is just nasty. [07:52:19] Ok. Is that all? [07:52:35] As long as they're not going back in the Project namespace... :-) [07:52:39] heh [07:52:40] That's pretty much it, yeah. [07:52:43] ok, good [07:52:45] thnks [07:52:50] thanks, even [07:52:58] <3 [07:54:55] An index of some kind (or master category) would be nice too. [07:55:44] > Categories: WMF Projects 2010q3 | WMF Projects 2010q4 | WMF Projects 2011q1 [07:55:47] Heh. [10:39:39] http://ultimategerardm.blogspot.com/2011/01/openzim-is-not-supported-here.html [22:57:27] ^demon: you around? [22:57:44] <^demon> Just woke up from a nap :) What's up? [22:57:56] I hear you're the man to talk to about ForeignFileRepos [22:58:06] hope the nap was good :) [22:58:31] <^demon> Well I woke up to it snowing outside :) [22:58:41] <^demon> So what's up? [22:58:48] looking forward to that in brussels [22:58:50] well, maybe [22:59:03] I can recognise belgians everywhere in brussels, pdhanda \o/ [22:59:34] ^demon: enabling Commons as a foreign repos on prototype is breaking some stuff and i was hoping you could help with that [23:00:15] config is in /srv/org/wikimedia/prototype/wikis/rccommon/PrototypeSettings.php (commented out for now) [23:00:47] werdna: yes and they will be so impressed when you tell them that :P [23:01:02] <^demon> pdhanda: Lemme log in [23:01:36] The breakage is ... upload form resetting after uploading an ogg , right? [23:02:20] <^demon> Not sure how the ForeignAPIRepo comes into play there [23:03:10] ^demon: when i disable it, i so no errors, everything works as expected [23:03:28] <^demon> Hmm. [23:03:37] i was wondering whether you know whether the foreign repos may affect it [23:03:52] <^demon> Lets enable it, and see what apache has to say :) [23:06:18] <^demon> Well, I do see a bunch of stuff relating to Special:Upload in the error log [23:06:43] <^demon> [Wed Jan 26 22:31:03 2011] [error] [client 216.38.130.167] File does not exist: /srv/org/wikimedia/prototype/images, referer: http://prototype.wikimedia.org/rc-en/index.php/Special:Upload [23:06:44] <^demon> [Wed Jan 26 22:31:03 2011] [error] [client 216.38.130.167] File does not exist: /srv/org/wikimedia/prototype/skins, referer: http://prototype.wikimedia.org/rc-en/index.php/Special:Upload [23:07:28] pdhanda: haha [23:09:00] something to do with detecting duplicates maybe? [23:15:43] actually i see the same behavior on my local box if it detects a duplicate [23:16:34] <^demon> pdhanda: That's my guess. [23:17:00] ugh, ok let me look into that then. [23:17:14] thanks for humoring me ^demon [23:17:19] :) [23:46:21] RoanKattouw: The reason the 'go back "on all wikis" ' for the Vector skin failed was that it used a new and/or seperate connection per-wiki instead per server and switching select-database, right ? [23:46:45] ie. 700 new connections [23:46:49] Originally, it opened separate conns to each wiki [23:46:52] Which killed s3 [23:46:59] So I hacked it up to do .... something else [23:47:07] I think it closed each conn immediately after [23:47:18] And of course there had to be logic to not accidentally close the 'main' wiki conn [23:47:22] But that didn't seem to work so good [23:47:29] So it didn't keep 700 open. but still causes 700 connections. [23:47:39] Yes [23:47:41] But that's fine [23:47:43] although not simultan anymore. [23:47:48] Okay [23:47:51] It was the # of parallel open conns that was the problem [23:47:58] It's limited at like 1000-2000 per DB server [23:50:19] But, as you may now, in my Toolserver tools (and in some other ppl's as well) we query the 'wiki-toolserver-table' which is a table of each wikidb_name, server_number and some other stuff like is_closed etc.. Then for each unique server_number I open a connection (so, usually 5 or 7 connections are opened). Then a foreach() loop through each dbname and check the server_number and reuse that connection (ie. $dbConnect_S1, $dbConnect_S2 etc.). So t [23:50:20] are 700 select_db() calls but only 7 mysql_connect() and mysql_close() calls [23:50:37] Yes, we talked about that [23:51:02] pretty effecient. Perhaps such a table should be added in the database with the other global stuff like globalusage and centralauth. [23:51:26] Nah, we've got it all in a config file [23:51:34] okay. [23:51:37] The data is probably in the LB/DB conn object *somewhere* [23:51:48] So what was the problem ? [23:51:50] I've just not had time to look at it since you pitched the idea like two weeks ago [23:51:58] Well currently the problem is it just don't work [23:52:11] Don't know why, at least not for sure [23:52:40] CheckUsage (toolserver) iterating like this as well. But the extension keeps a seperate table. Could you rougly link to or explain why this was done (if iteration is not a problem on WMF) [23:54:37] globalcontribs and checkusage take like 1-2 seconds on toolserver. I keeps me wondering why a global (with duplicate information and ocasional bugs with ghost or missing entries) table was created.