[08:01:51] Hi! I installed mediawiki core in my instance via the puppetization role::mediawiki-install::labs. How can I get a manual configuration for the exim4, rather than the usual setup of exim4 to switch over SMTP ? [10:28:18] how can I remove puppet configuring my exim server configs in labs instance ? [11:19:00] !log deployment-prep Attempting to reenable SSL support with {{gerrit|124057}} [11:19:03] Logged the message, Master [11:26:13] Does anybody know why the replication lag on s4 / commonswiki.labsdb is increasing? [11:27:50] Jelte: you will have better luck asking on the labs-l mailing list [11:27:59] IRC is very quiet during week-end [11:28:14] * hashar vanishes [14:17:44] Hi.... I would like to use /shared/pywikipedia/compat/tag_nowcommons.py but when I run it the bot starts generating a user_config.py... I used https://wikitech.wikimedia.org/wiki/User:Russell_Blau/Using_pywikibot_on_Labs to set up by bot so I have a .pywikibot and a user_config.py [15:15:30] Is there a job that needs configuring for the dump of Wikidata to be copied to Labs ? [15:16:32] ... based on the normal number of days between dumps there should have been one already [17:30:29] https://wikitech.wikimedia.org/wiki/Setting_up_a_MySQL_replica [17:30:40] this is not Labs I know [17:30:50] Why does it say "It is a violation of federal law to use this product in a manner inconsistent with its labeling" ? [17:31:56] Wanna-be comedians, captive audience? [19:26:08] huh: LOL [19:59:19] petan: enwiki replication is screwed up http://tools.wmflabs.org/betacommand-dev/cgi-bin/replag [19:59:37] @replag [19:59:38] Replication lag is approximately 00:00:01.8741200 [19:59:45] Betacommand: it's not [19:59:54] @replag [19:59:55] Replication lag is approximately 00:00:00.7200340 [20:00:11] petan: wm-bot lies [20:00:22] it's bota heh [20:00:24] :P [20:00:37] do select from RC and check the most recent timestamp [20:00:47] that is what bota does [20:01:53] commonswiki 96623 1 day, 2:50:23 [20:01:56] ^^ o_O [20:02:03] https://github.com/benapetr/wikimedia-bot/blob/master/plugins/rl/rl/Class1.cs [20:02:25] core.DB.Select("recentchanges", "rc_timestamp", "order by rc_timestamp desc limit 1", 1); [20:03:49] Betacommand: even the tool you linked here show 0 second [20:03:58] your own script returns similar values [20:04:10] bota is right [20:05:54] petan: nope, let me get you a link [20:07:05] petan: https://dpaste.de/xboJ [20:07:44] you said enwiki [20:07:47] this is commons [20:07:57] but yes commons are fucked up [20:08:07] unfortunatelly I am not allowed to fix it [20:08:43] petan: sorry got my db names mixed up [20:21:08] @replag dewiki [20:21:11] ah :/ [20:26:30] enwiki seems to have caught up *now*. But earlier it was clearly stopped: http://ganglia.wmflabs.org/latest/graph_all_periods.php?c=tools&h=tools-login&r=hour&z=small&jr=&js=&st=1396729522&v=0&m=s1&z=large [20:26:54] Probably a writing query that locked the tables. [20:47:06] legoktm: why is there no webservice for globalprefs? [20:47:12] Kolega2357 would like to use it apparently [23:14:13] scfc_de: Hi. I don't have problems anymore with the server shut down because of the connection limit, but at the moment I'm getting "backend is overloaded" errors again... [23:15:35] scfc_de: it seems, that the php-cgi backend is overloaded, but I don't really know lighttpd or php-cgi or how this could be solved [23:25:59] scfc_de: it works again, after a restart ... but before every single request produced a "backend is overloaded" error message... so it sounds like some php zombie processes hung somewhere or something like that... any experiences with that? [23:26:00] apper: Hmmm. I'm not very knowledgeable about that either. For the most part, I assume this occurs when runtime(script) > hits * workers. [23:27:25] scfc_de: the load from outside shouldn't be that high at this time and is not lower after a restart, so it has nothing to do with too much requests at this time but with some zombies, I think [23:28:53] apper: "wikihistory", right? tools-webgrid-01 has only five php-cgi processes by that, but no lighttpd, so I assume this is on -webgrid-02? [23:29:07] Yes. [23:29:18] wikihistory, yes [23:29:39] but I already restarted it, so hopefully no everything is in a correct state ;) [23:30:54] hm, I'm just seeing I forgot to close some mysql connections at the end of the most called php script, but I thought this is done automatically at the end, but I will add this [23:30:56] What I notice is that that user has 20 lighttpd process, but only 5 php-cgis as every other tool. I would assume that these scale up as well, but I don't know how the control flow lighttpd -> php-cgi -> something.php is. I'll make a note to read up on that. [23:31:50] I think "PHP_FCGI_CHILDREN" is the point [23:32:06] this is set to 4 child processes... together with the one main process this is maybe the 5? [23:32:27] https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/Help#Default_configuration [23:33:36] does lighttpd disable fcgi if it has to wait for it? this sound weird [23:34:40] but of course: more concurrent connections than fcgi php threads doesn't make sense on a php driven page [23:35:03] No idea. For example http://wohnzimmerhostblogger.de/archives/1031-Warum-man-PHP_FCGI_CHILDREN-statt-auf-1-lieber-gar-nicht-setzen-sollte.html sounds very complex. [23:35:30] but the thing is, that at this time not sooo many people are online, so I think it was something hanging [23:37:16] thanks for your help, now everything works after the restart, so I will look after this the next days... [23:39:11] I'll look into the relationship of the various parameters, and what regulates what. In fact, I'll better file a bug so that this is not forgotten. [23:39:48] scfc_de: thanks a lot! [23:40:06] (Though setting PHP_FCGI_CHILDREN = max-connections seems like a good idea.) [23:48:03] !log sugarcrm created officetools instance [23:48:05] Logged the message, Master