[00:53:51] <YuviPanda>	 petan: Coren around?
[00:53:57] <YuviPanda>	 awjr:  is having trouble adding me to a service group
[00:54:05] <YuviPanda>	 it tells him he's not part of it, even though I can see that he is
[00:54:05] <awjr>	 you beat me to it YuviPanda
[00:54:10] <YuviPanda>	 'local-bingle'
[00:54:13] <YuviPanda>	 :)
[00:54:21] <awjr>	 https://wikitech.wikimedia.org/w/index.php?title=Special:NovaServiceGroup&action=addmember&projectname=tools&servicegroupname=local-bingle&returnto=Special%3ANovaProject
[00:54:32] <awjr>	 results in 'Not in service group - you must be a member of this service gorup to perform this action'
[00:55:01] <awjr>	 even though i am in the group (i set it up...)
[00:56:44] <Reedy>	 Tried the usual log out/log in for session failure?
[00:56:57] <YuviPanda>	 awjr: ^ log out and back in?
[00:57:18] <awjr>	 harumph
[00:57:45] <awjr>	 ok, tried it, still doesn't work :(
[00:59:17] <YuviPanda>	 awjr: I guess we caught the admins in a bad timezone :)
[00:59:41] <awjr>	 oh well
[00:59:55] <YuviPanda>	 awjr: but become works for you
[00:59:57] <YuviPanda>	 so that's good
[01:00:03] <awjr>	 yeah so i can at least get the tool set up
[01:00:07] <YuviPanda>	 indeed
[01:00:34] <YuviPanda>	 awjr: I'm going to be around for a while now anyway, poke me if you hit snags :)
[01:00:58] <awjr>	 thanks YuviPanda, i need to go soon so i'll probably not get totally done tonight
[01:01:03] <YuviPanda>	 okay :)
[01:01:09] <YuviPanda>	 enjoy your weekend, awjr
[01:23:58] <awjr>	 zz_YuviPanda: bingle now running on tool labs :)
[08:55:39] <YuviPanda>	 awjr_away: :)
[10:38:43] <Ynhockey>	 i have a question: (i know i need to read the manual but it's kind of urgent): how do i get the top-edited articles for a certain wikipedia? does anyone have a ready script or can someone explain quickly how the best query will look like? i understand that the edit count for articles is cached somewhere
[10:43:05] <Amir1>	 how i can run a mysql in labs? I ran this but an error returned 
[10:43:07] <Amir1>	 ladsgroup@tools-login:~/pywikipedia$ mysql nlwiki_p <iwdwoiw.sql> rezanl.txt
[10:43:09] <Amir1>	 ERROR 2002 (HY000): Can't connect to local MySQL server through socket '/var/run/mysqld/mysqld.sock' (2)
[10:49:30] <legoktm>	 Amir1: try $ sql nlwiki_p <iwdwoiw.sql> rezanl.txt
[10:49:32] <fale>	 Amir1: you miss the connect line
[10:49:43] <Amir1>	 ok
[10:49:53] <legoktm>	 sql is a special wrapper around mysql which takes care of authentication and picking the right host
[10:50:04] <Amir1>	 legoktm: thank you
[10:50:15] <legoktm>	 np :)
[10:50:17] <Amir1>	 why you're still awake :P
[10:50:31] <legoktm>	 i was about to go to sleep xP
[10:51:42] <fale>	 Amir1: are you creating some query to create a list of pages that will have to be refreshed periodically?
[10:51:56] <Amir1>	 fale:  no
[10:52:01] <fale>	 Amir1: ok :)
[10:52:14] <Amir1>	 I'm importing articles to wikidata
[10:52:24] <fale>	 :D
[11:54:33] * YuviPanda  considers setting up some form of map reduce environment on betalabs

[11:54:36] <YuviPanda>	 err
[11:54:37] <YuviPanda>	 toollabs
[11:54:38] <YuviPanda>	 not betalbs
[11:54:39] <YuviPanda>	 grr
[12:28:56] <valhallasw>	 YuviPanda: why map reduce and not generic parallel computation stuff?
[12:29:45] <valhallasw>	 YuviPanda: SGE make *grin*
[12:30:05] <valhallasw>	 of course, this exists. It's the /other/ qmake.
[12:33:51] <fale>	 If I have to look all the pages that include a certain string (or regex), I can use three methods: download all pages from wiki and parse one by one (surely deprecated), download a dump and run a pywikipedia/relace.py script or use a mysql query. Whichone of the last two is the best one?
[12:36:01] <YuviPanda>	 valhallasw: primarily for things like Pig or Hive
[12:36:10] <YuviPanda>	 valhallasw: lets you write 'queries' in an SQLish way, but far more powerful
[12:36:21] <YuviPanda>	 what fale asked for could be done that way, for example ;)
[12:36:27] <YuviPanda>	 fale: dump + scan
[12:36:35] <YuviPanda>	 fale: dumps are already on tool labs, so you can just scan them
[12:37:28] <fale>	 YuviPanda: in the mysql db there is the text of every page... so even a mysql query can do the job ;)
[12:37:47] <valhallasw>	 fale: afaik page text is not replicated?
[12:38:16] <fale>	 valhallasw: uhm, I haven't checked, but I think is there 
[12:38:21] <YuviPanda>	 it's not, IIRC
[12:38:35] <YuviPanda>	 IIRC the cluster doesn't even store the text in mysql :D
[12:38:38] <valhallasw>	 fale: for wikimedia wikis, page text is not stored in the text table
[12:38:42] <valhallasw>	 er, page table
[12:38:49] <valhallasw>	 revision.
[12:39:01] * valhallasw  reboots brain

[12:39:38] <YuviPanda>	 valhallasw: something like this would be something that sets up nicely on Hadoop, for example
[12:40:15] <valhallasw>	 YuviPanda: ah, ok.
[12:40:42] <fale>	 All public wikis will be replicated to the LabsDB servers, with private user data redacted.
[12:43:17] <YuviPanda>	 fale: no revision text tho
[12:43:45] <fale>	 YuviPanda: I see... so the WikiTech docs has to be fixed
[12:44:17] <YuviPanda>	 fale: yeah. the reasons were technical and legal
[12:44:56] <fale>	 I can see tech reasons, while Legal... not really. On TS there was the full text replicated, IIRC
[12:45:35] <YuviPanda>	 yeah, something about revdeletes not going through i think
[12:45:46] <YuviPanda>	 unsure tho
[12:47:18] <fale>	 YuviPanda: I see, thanks :). So there is only one way to retrive this kind of data: using dumps
[12:47:30] <YuviPanda>	 yup
[12:47:50] <YuviPanda>	 fale: addtravel was working on 'dumpscan' explicitly for this
[12:49:12] <fale>	 YuviPanda: nice, I'm thinking about expanding lists to support .py fixes file (for creating lists of pages matching certain regex)
[12:49:24] <YuviPanda>	 ah
[12:49:26] <YuviPanda>	 nice
[12:50:22] <fale>	 YuviPanda: now lists only supports mysql query, but if they are not capable to look for strings of text... I think the .py extention is needed
[12:50:23] <YuviPanda>	 valhallasw: http://www.mappian.com/blog/hadoop/using-hadoop-to-analyze-the-full-wikipedia-dump-files-using-wikihadoop/ :)
[12:50:56] <YuviPanda>	 however
[12:50:59] <YuviPanda>	 ' We are running it on a three node mini-cluster with quad-core machines and the job takes about 14 days to parse the entire English Wikpedia dump files.'
[12:51:45] <fale>	 YuviPanda: not really efficint
[12:52:04] <fale>	 YuviPanda: consider that they are talking about a -full-history, I would say
[12:53:52] <YuviPanda>	 fale: they seem to be actually considering all diffs, so that's even more than full-history
[12:53:56] <YuviPanda>	 so... not bad
[12:54:11] <YuviPanda>	 i guess without history it'll be much lesser
[12:54:41] <fale>	 YuviPanda: it's 5.5Tb the version they are using... I think the last version onlyi is way less
[12:56:22] <fale>	 YuviPanda: is 9.2Gb bz2 compressed, so less than 100Gb
[12:56:28] <fale>	 *200
[12:56:40] <YuviPanda>	 yeah, so should be much faster
[12:56:51] <YuviPanda>	 we should play with it, methinks :D
[12:58:00] <fale>	 YuviPanda: also, hadoop increases performances linearly increasing the number of nodes ;)
[12:58:16] <YuviPanda>	 yeah and if we set things up on puppet, should be trivial to spawn new nodes
[12:59:07] * fale  has not really clear what puppet is and does

[13:01:31] <gerrit-wm_>	 New review: coren; "Protection against library injection is actually provided for by ld.so at runtime (it will not honor..." [labs/toollabs] (master) - https://gerrit.wikimedia.org/r/71112
[13:02:53] <gerrit-wm_>	 New review: coren; "Still ok after rebase" [labs/toollabs] (master); V: 2 C: 2;  - https://gerrit.wikimedia.org/r/70733
[13:02:53] <gerrit-wm_>	 Change merged: coren; [labs/toollabs] (master) - https://gerrit.wikimedia.org/r/70733
[13:04:07] <gerrit-wm_>	 New review: coren; "Also still good after rebase." [labs/toollabs] (master); V: 2 C: 2;  - https://gerrit.wikimedia.org/r/70734
[13:04:07] <gerrit-wm_>	 Change merged: coren; [labs/toollabs] (master) - https://gerrit.wikimedia.org/r/70734
[13:07:24] <YuviPanda>	 fale: also I'm not sure about using Hadoop here, since it might be too large a hammer for this particular nail
[13:09:44] <fale>	 YuviPanda: btw: if you want to have good real-time performances, hbase can be a good solution ;)
[13:10:03] <YuviPanda>	 yeah but question is how do we map the dumps into the system in an easy way
[13:10:06] * YuviPanda  looks into HBase

[13:10:25] <YuviPanda>	 i'm just exploring things.
[13:10:29] <fale>	 YuviPanda: surely a calculate have to be done between the costs and the results
[13:10:51] <YuviPanda>	 fale: in terms of tool reliability, having some form of logging / monitoring would be more important than this
[13:11:18] <YuviPanda>	 Hbase doesn't seem that useful.
[13:11:29] <fale>	 YuviPanda: hadoop have awesome results on a huge dataset (our is large but I would not definite it as huge) and with tens or - better -hundreds of nodes
[13:11:37] <YuviPanda>	 yeah
[13:11:43] <YuviPanda>	 but it runs with 3-4 ones too
[13:11:48] <YuviPanda>	 I guess hadoop is overkill here
[13:12:26] <fale>	 YuviPanda: hadoop can even run on a single node... but has comparable performances to any other map-reduce shell script ;)
[13:12:32] <YuviPanda>	 indeed
[13:14:04] <fale>	 YuviPanda: I've used hadoop/hbase only for other kind of data (financial data) so I can't say howfast  it would handle wiki data
[13:14:34] <YuviPanda>	 in this case I think the 'cost' is going to be 'how much time will we spend on setting it up' and the results would be 'what new things can people do with it'?
[13:14:39] <YuviPanda>	 not sure about the answers to both
[13:14:50] <YuviPanda>	 setting up Redis was ridiculously justifiable :)
[13:15:31] <fale>	 YuviPanda: on the cost consider also the time to (eventually) rewrite some rules that already written for other systems (mysql/pywiki) to hadoop
[13:15:46] <YuviPanda>	 true
[13:16:57] <fale>	 YuviPanda: can be interesting if enought nodes are available, since hadoop handle very well a situation in which the same data are used imultaneusly on more than one request
[13:17:19] <YuviPanda>	 fale: indeed!
[13:17:25] <YuviPanda>	 data locality optimization, etc
[13:17:30] <fale>	 yep
[13:17:42] <YuviPanda>	 I guess we can setup 3-4 nodes easily on toolsbeta without Coren hitting me with a stick
[13:17:52] <YuviPanda>	 but they'll be slow, because VM
[13:18:04] <fale>	 YuviPanda: am I wrong or the tool-labs is based on amazon hw?
[13:18:19] <YuviPanda>	 fale: no, it's running in the foundation's data center
[13:18:23] <YuviPanda>	 fale: it's based on OpenStack
[13:18:30] <fale>	 YuviPanda: I see
[13:18:33] <YuviPanda>	 which has some of the capabilities of AWS, but not all
[13:19:03] <fale>	 is there local HD or is all SAN/NAS HD?
[13:19:29] <YuviPanda>	 fale: I *think* there's a local HD, but I'm not completely suer about the architecture
[13:19:37] <YuviPanda>	 I know that we have some stuff on Gluster, and that's 'going away' soon
[13:20:40] <fale>	 that could be a good point to start, I think, unserstanding the physical server structure
[13:21:43] <YuviPanda>	 indeed
[13:21:51] <YuviPanda>	 fale: Coren is also changing it right now
[13:21:51] <YuviPanda>	 so
[13:22:16] <fale>	 YuviPanda: Coren has physical access to the datacenter?
[13:22:46] <YuviPanda>	 he's sorting out what he called the current 'NFS mess', so...
[13:22:51] <YuviPanda>	 not the data center itself
[13:23:12] <fale>	 I see
[14:17:43] <gerrit-wm_>	 New review: Tim Landscheidt; "@Coren: Thanks.  I'll still need to fix Architecture: and the Depends: stuff, so I keep at -1 for a ..." [labs/toollabs] (master) - https://gerrit.wikimedia.org/r/71112
[14:29:23] <YuviPanda>	 Coren: are you part of the ldap/ops group?
[14:29:27] * YuviPanda  needs a small favor

[14:29:46] <YuviPanda>	 need to tick a checkbox (or something) in Gerrit :) Nothing major, I swear
[14:31:58] <YuviPanda>	 fale: http://lintool.github.io/Cloud9/docs/content/wikipedia.html
[15:16:34] <YuviPanda>	 anyone with a tools account who can help me with a small experiment?
[15:16:45] <YuviPanda>	 like, really, small!
[15:25:34] <valhallasw>	 YuviPanda: sure
[15:25:38] <valhallasw>	 !ask
[15:25:38] <wm-bot>	 Hi, how can we help you? Just ask your question.
[15:25:40] <valhallasw>	 ;-)
[15:26:20] <YuviPanda>	 valhallasw: can you read /home/yuvipanda/testing/hi
[15:26:20] <YuviPanda>	 ?
[15:26:27] <YuviPanda>	 from your account?
[15:26:35] <YuviPanda>	 and also from a tool account?
[15:26:41] <YuviPanda>	 should be able to, but I'm just double checking
[15:27:02] <valhallasw>	 YuviPanda: no
[15:27:17] <valhallasw>	 also not from nlwikibots
[15:27:22] <YuviPanda>	 hmm
[15:27:27] <YuviPanda>	 drwxrwxr-x   2 yuvipanda svn       15 Jun 29 15:26 testing
[15:27:37] <valhallasw>	 drwx------ 13 yuvipanda svn 4096 Jun 29 15:27 /home/yuvipanda/
[15:27:40] <YuviPanda>	 but drwx------  13 yuvipanda          svn            4096 Jun 29 15:27 yuvipanda
[15:27:41] <YuviPanda>	 yeah
[15:27:47] <valhallasw>	 set a+x
[15:28:00] <valhallasw>	 read rights are not necessary, execute (enter directory) rights are
[15:28:23] <YuviPanda>	 hmm
[15:28:50] <YuviPanda>	 i'm a little wary of messing with perms on my home dir, doing it wrong and locking myself out :D
[15:29:03] <valhallasw>	 :D
[15:29:08] <valhallasw>	 chmod a+x /home/yuvipanda
[15:29:10] <valhallasw>	 should be OK
[15:29:14] <valhallasw>	 that's what I did :p
[15:29:26] <YuviPanda>	 well, okay ;)
[15:29:27] <YuviPanda>	 try now
[15:29:32] <YuviPanda>	 can you access?
[15:29:37] <valhallasw>	 YuviPanda: yep
[15:30:01] <YuviPanda>	 valhallasw: but you still can't read, for example, /home/yuvipanda/.rnd
[15:30:05] <YuviPanda>	 but you can list it
[15:30:07] <YuviPanda>	 is that right?
[15:30:52] <YuviPanda>	 my (weak) permissions fu says that's correct, but it is 'weak'
[15:47:13] <valhallasw>	 YuviPanda: no
[15:47:30] <valhallasw>	 YuviPanda: no +r so no listing
[15:47:45] <YuviPanda>	 what do you see when you do a ls /home/yuvipanda
[15:47:45] <YuviPanda>	 ?
[15:47:48] <valhallasw>	 YuviPanda: and no +r on /home/yuvipanda/.rnd so I can't read that
[15:47:58] <valhallasw>	 ls: cannot open directory /home/yuvipanda: Permission denied
[15:48:07] <YuviPanda>	 but ls /home/yuipanda/testing works
[15:48:16] <YuviPanda>	 so +x is traversal rather than listing permission
[15:48:24] <valhallasw>	 yes
[15:48:32] <valhallasw>	 +r is listing permission
[15:48:32] <YuviPanda>	 makes sense :D
[15:48:54] <valhallasw>	 but I can explicitly ask about .rnd
[15:48:54] <valhallasw>	 -rw------- 1 yuvipanda svn 1024 Jun 28 16:19 /home/yuvipanda/.rnd
[15:49:29] <valhallasw>	 and I can read your .vimrc >:_
[15:49:45] <valhallasw>	 but that was the point of the exercise to begin with iirc
[15:50:01] <YuviPanda>	 :)
[15:50:10] <YuviPanda>	 valhallasw: that's one of the points, but that's not the sole point
[15:50:20] <YuviPanda>	 valhallasw: i'm testing out a way of making ipython notebook available to tool users
[15:50:27] <YuviPanda>	 with public permalinked shareable urls for the outputi
[15:50:46] <valhallasw>	 oooooh.
[15:51:03] <YuviPanda>	 so by default their output will be private, under their userspace. If they chose to publish, it'll move to a properly permissioned folder in their userspace
[15:51:16] <YuviPanda>	 and then the tool will just read those and serve them with appropriate transformations :)
[15:51:43] <valhallasw>	 I see
[15:52:05] <YuviPanda>	 valhallasw: so this... should work
[15:52:24] <YuviPanda>	 ipython brings with it security issues, so we can't easily open it up to random users (no matter how much I'd like that)
[15:52:33] <YuviPanda>	 but restricting it to tool-users who already have login seems okay
[15:52:45] <YuviPanda>	 for now it can run on -login, if that gets too much we can easily setup another system for it :)
[15:52:47] <valhallasw>	 YuviPanda: why not just letting tool users run ipynb theirselves?
[15:53:09] <YuviPanda>	 valhallasw: yeah, that's what this will do. Will need to proxy via ssh tho
[15:53:17] <YuviPanda>	 it uses websockets, so we can't actually just serve them via apache
[15:53:22] <valhallasw>	 right.
[15:53:37] <YuviPanda>	 i'm just writing a tiny script that'll automate the 'ssh in, setup ipy + deps, start server on appropriate port, and setup an ssh tunnel'
[15:53:52] <YuviPanda>	 so instead of telling people 'do x, y, z oh and then a, b, c' you just go 'curl <script> | sh'
[15:53:55] <YuviPanda>	 (or similar)
[15:54:24] <YuviPanda>	 the share feature is also important because nobody seems to appreciate how awesome ipynb is :(
[16:08:31] <fale>	 YuviPanda: sorry, I was afk
[16:08:42] <YuviPanda>	 'sokay fale, valhallasw helped :)
[16:10:19] <fale>	 YuviPanda: cool the cloud9 thing... seems exactly what we need :)
[16:10:26] <YuviPanda>	 yeah :)
[16:10:37] <YuviPanda>	 fale: also the analytics team already uses hadoop internally
[16:10:42] <YuviPanda>	 for log file analysis
[16:11:05] <fale>	 YuviPanda: cool
[16:11:23] <fale>	 YuviPanda: so the hw infrastructure should already be fine with hadoop
[16:11:31] <YuviPanda>	 yeah
[17:41:13] <YuviPanda>	 Coren: if I'm on screen (from -login), then 'become' a tool, I get no colors :(
[17:41:20] <YuviPanda>	 I get colors on -login tho
[17:41:25] <YuviPanda>	 is this a bug I should report?
[17:41:39] <Coren>	 That depends.  Colors /where/?
[17:41:43] <YuviPanda>	 in vim
[17:42:00] <YuviPanda>	 actually, I get colors
[17:42:09] <YuviPanda>	 hmm, just not, enough colors?
[17:42:18] <YuviPanda>	 I can send screenshots
[17:42:30] <Coren>	 tools tend to not get the default profile/settings.  Perhaps you just don't have syntax on in your vimrc for the tool?
[17:42:48] <YuviPanda>	 Coren: i just set that up. the powerline shows up, but just without colors.
[17:43:20] <Coren>	 "powerline"?
[17:43:21] * YuviPanda  makes screenshots

[17:44:00] <YuviPanda>	 Coren: http://imgur.com/ufPtVXK,BR2Qpsb
[17:44:17] <YuviPanda>	 second image is on login, first after I do become
[17:44:27] <YuviPanda>	 the status line, really. powerline is just the plugin's name...
[17:44:51] <Coren>	 I was about to say that I don't recognize that UI.  :-)
[17:45:05] <YuviPanda>	 too much emacs? :P
[17:45:31] <YuviPanda>	 my terminal is usually much bigger, and the status line looks much better
[17:45:38] <YuviPanda>	 I use relative numbering, so status bar is essential
[17:45:40] <Coren>	 I honestly don't see why not; you must have something differently set between the two; I can switch between my account and by bots and get the normal syntax highlighting, lemme check something about screen.
[17:46:20] <YuviPanda>	 Coren: if I just do 'become' sans screen, I get the colors
[17:46:37] <YuviPanda>	 Coren: we already can't run screen as tool accounts. Related?
[17:46:50] <Coren>	 No.  The latter is normal for non suid-root screen.
[17:47:27] <Coren>	 Hm.  Works For Me™  Check something, though, $TERM should be set to "screen" both before and after become.  Is it?
[17:47:32] <Coren>	 (echo $TERM)
[17:47:33] <YuviPanda>	 checking
[17:47:50] <YuviPanda>	 it is
[17:48:12] * Coren  doesn't get it.

[17:48:19] <Coren>	 Have you more than one tool?
[17:48:24] <YuviPanda>	 yeah
[17:48:30] <YuviPanda>	 this is freshly created, less than 10 mins ago
[17:48:36] <Coren>	 Behaviour identical with all of them?
[17:48:36] <YuviPanda>	 just symlinked all my vim related files
[17:48:42] <YuviPanda>	 no, let me test another
[17:48:43] <Coren>	 Hm.
[17:49:05] <YuviPanda>	 Coren: okay, become is unrelated. just running vim inside screen loses my colors
[17:49:12] <YuviPanda>	 on login
[17:49:51] <Coren>	 I'm going to wager that whatever you do with your setup doesn't grok the screen terminal type.  Try this:  "export TERM=xterm"
[17:50:13] <YuviPanda>	 no luck
[17:50:16] <YuviPanda>	 (inside screen)
[17:50:20] <Coren>	 dafu?
[17:50:37] <Coren>	 Disable your powerbar thingy to try again?
[17:51:45] <YuviPanda>	 should that affect it?
[17:51:53] <YuviPanda>	 actually, let me try ssh directly, rather than mosh
[17:52:27] <YuviPanda>	 okay, same effect in ssh too
[17:52:29] <Coren>	 Well, I'm admitedly groping in the dark and doing the "let's turn things off until it works again" dance; I can't reproduce your problem with the default vim setup.
[17:53:37] <YuviPanda>	 sigh
[17:54:00] <YuviPanda>	 I.. am guessing it's probably a 16color vs 256 color issue?
[17:54:28] <YuviPanda>	 Coren: tada! doing a set t_Co=256 fixed that
[17:54:32] <YuviPanda>	 inside screen
[17:54:44] <YuviPanda>	 though I've no idea why that would change based on screen or not. it's not set in my vimrc
[18:02:38] <YuviPanda>	 Coren: sorry for the bother.
[18:03:08] <Coren>	 YuviPanda: How else would I earn my somptuous salary (*choke* *sputter*)
[18:03:51] <YuviPanda>	 well, in that case since I found the final solution myself, GIMME MY MONEY BAAAAACKKKK!!!!!!11
[18:04:25] <Coren>	 Heh.
[18:04:37] <Coren>	 WMF: Work for 'em because you like the job, not to get rich. :-)
[18:04:55] <YuviPanda>	 :)
[18:05:37] <YuviPanda>	 Coren: is there monitoring (icinga?) for the toollabs infrastructure?
[18:05:40] <YuviPanda>	 apaches, sge, etc?
[18:05:48] <YuviPanda>	 if there is, I should have redis plug into that
[18:06:06] <YuviPanda>	 Coren: also, why are the packages set to ensure => 'present' rather than ensure => 'latest'?
[18:06:19] <YuviPanda>	 since they're not really packages that are part of the underlying infrastructure
[18:06:22] <Coren>	 YuviPanda: To avoid breaking upgrades that haven't been tested.  :-)
[18:06:48] <YuviPanda>	 tch tch, what's the fun in not having to fix your tool over and over? ;)
[18:06:51] <Coren>	 YuviPanda: As a rule, we use 'present' for things we don't control, 'latest' for things we do.
[18:06:55] <YuviPanda>	 ah
[18:06:57] <YuviPanda>	 makes sense
[18:07:11] <YuviPanda>	 and monitoring?
[18:07:13] <Coren>	 Hence 'jobutils => latest'.
[18:07:46] <Coren>	 Most classes have monitoring set up automagically, there is a puppet class you can add to your reddis manifest, IIRC, that will do so.
[18:08:36] <Coren>	 To be honest, I'm not all that familiar with the icinga setup; but my understanding is that adding a new thing to monitor should be fairly easy.
[18:08:53] <YuviPanda>	 hmm, okay
[18:09:31] <YuviPanda>	 Coren: there is a $monitor variable for the Redis class, but I'm not sure if it is setup separately appropriately for labs / production
[18:09:35] <YuviPanda>	 don't want to break it...
[18:09:40] <Coren>	 It is.
[18:10:06] <Coren>	 http://icinga.wmflabs.org/cgi-bin/icinga/status.cgi?host=tools-login.pmtpa.wmflabs
[18:10:30] <YuviPanda>	 ah
[18:10:34] * Coren  notices the total process limit is set too low for -login

[18:10:47] <YuviPanda>	 well, patch submitted then :)
[18:11:11] <YuviPanda>	 Coren: https://gerrit.wikimedia.org/r/71181
[18:11:48] <YuviPanda>	 fhttp://icinga.wmflabs.org/cgi-bin/icinga/status.cgi?host=tools-mc.pmtpa.wmflabs
[18:11:51] <YuviPanda>	 free ram is unknown?
[18:12:48] <YuviPanda>	 Coren: thanks :) hope that doesn't break stuff
[18:14:07] <Coren>	 Doing a puppet now to see.
[18:14:27] <YuviPanda>	 :D
[18:20:25] <AzaToth>	 Trying to access https://en.wikipedia.beta.wmflabs.org/ I'm getting certificate error untrusted
[18:20:28] <AzaToth>	 feature or bug?
[18:21:20] <AzaToth>	 The certificate is not trusted because the issuer certificate is not trusted. The certificate is only valid for *.wmflabs.org 
[18:21:38] <Damianz_>	 feature
[18:21:57] <YuviPanda>	 probably a NSA MITM
[18:22:33] <AzaToth>	 Damianz_: rly?
[18:23:18] <AzaToth>	 YuviPanda: they can neither confirm or deny the fact that they can neither confirm or deny that fact
[18:23:19] <Damianz_>	 Labs doesn't have real ssl certs atm - it's been suggested for the proxy a few times
[18:23:28] <YuviPanda>	 yo dawg...
[18:24:11] <AzaToth>	 Damianz_: could always whip up some CACerts
[18:25:24] <YuviPanda>	 Coren: another thing I was thinking of is the ability to have things like websockets / SSE from labs.
[18:25:35] <YuviPanda>	 Coren: none of the new 'real time' stuff works through Apache
[18:25:56] <YuviPanda>	 and I don't know if it ever will.
[18:26:30] <Coren>	 Has to be adatped to (a) start the backend on the grid and (b) track where on the grid it is running.
[18:26:30] <AzaToth>	 is apache or nginx (or something else) the future?
[18:26:55] <YuviPanda>	 HTTP is the present, past and the future :)
[18:27:16] <YuviPanda>	 Coren: yeah, but it needs a way to have a port open. How would that work?
[18:27:19] <YuviPanda>	 our webproxy is apache, right?
[18:27:24] <AzaToth>	 <center><blink> is the future
[18:27:31] <YuviPanda>	 <blink><center>
[18:27:44] <Coren>	 YuviPanda: It is.
[18:27:51] <YuviPanda>	 yeah, and that doesn't support these
[18:27:55] <YuviPanda>	 it's not in 2.4 either
[18:28:16] <YuviPanda>	 Coren: any particular reason it can't be nginx? we use nginx in production too
[18:28:29] <Coren>	 Bah.  You kids and your newfangled toys.  In my days, we had FTP (and gopher, if you were lucky) and we were content.
[18:28:48] <Coren>	 YuviPanda: Any particular reason why it should be nginx?
[18:28:54] <YuviPanda>	 nginx supports websockets
[18:28:56] <YuviPanda>	 and SSE
[18:29:01] <YuviPanda>	 and all the newfangled things
[18:29:18] <YuviPanda>	 as in
[18:29:21] <YuviPanda>	 it supports proxying them
[18:30:03] <YuviPanda>	 http://nginx.org/en/docs/http/websocket.html
[18:30:15] * Damianz_  re-fangles Coren

[18:30:42] * YuviPanda  defangs Damianz

[18:30:56] <YuviPanda>	 hmm, if you de-mianz Damianz you end up with a Da!
[18:31:05] <AzaToth>	 http://www.goer.org/htmlhorror/htmlhorror1.html
[18:31:13] <valhallasw>	 Coren: here, have something old-fashioned: https://www.youtube.com/watch?feature=player_embedded&v=X9dpXHnJXaE#t=5m10s
[18:32:18] <Coren>	 valhallasw: I had an acoustic 110 baud modem once.
[18:32:35] <YuviPanda>	 the first internet connection I ever got home was 256kbps ADSL
[18:32:39] <YuviPanda>	 :D
[18:32:45] <YuviPanda>	 and I don't remember an internet before Google
[18:32:46] <valhallasw>	 http://williambader.com/museum/modem110/modem110.html < one of those?
[18:33:39] <Coren>	 valhallasw: Nah, this one is fancy and combines the phone with the modem so that the audio doesn't have to be audible.  Mine was acoustic.
[18:34:05] <valhallasw>	 oh, wait, you're right. The horn confused me.
[18:35:19] <Coren>	 One very much like the right one on this: http://www.columbia.edu/cu/computinghistory/couplers.jpg
[18:36:06] <Coren>	 It was, admitedly, very old when I got it; state-of-the art was 300 baud at the time.  :-)
[18:36:49] * Coren  got it cheap from a government acution.

[18:37:24] <valhallasw>	 300 baud ought to be enough for anyone!
[18:37:27] <YuviPanda>	 heh, I like how all these links are from various sites with 'history' in the URL :)
[18:37:50] <YuviPanda>	 so, Coren, nginx? Shall I experiment with it on toolsbeta, to replace apache as webproxy?
[18:37:56] <AzaToth>	 https://www.youtube.com/watch?v=abapFJN6glo
[18:38:06] <Coren>	 YuviPanda: Not on the roadmap for the forseeable future.
[18:38:42] <YuviPanda>	 Coren: so even if I experiment with it, and prove it's stable enough, you're not interested / comfortable enough to move it to tools?
[18:39:02] <YuviPanda>	 I understand if that's the case, just want to clarify so I don't waste my time
[18:39:38] <Coren>	 YuviPanda: More precisely, I wouldn't have the cycles to examine the results of your experiments to do so.  At least not yet.
[18:39:50] <YuviPanda>	 yeah, I understand.
[18:40:01] <YuviPanda>	 well, we'll revisit in a year, I guess.
[18:40:07] <Coren>	 And now, it's the weekend, and I'm going to go away a bit to pretend I know how to relax.  :-)
[18:40:11] <YuviPanda>	 haha
[18:40:22] <YuviPanda>	 haha RnR, sure! :)
[18:40:28] <YuviPanda>	 I bet he's still on the computer :)
[18:41:41] <valhallasw>	 YuviPanda: http://cdn0.meme.li/instances/300x300/39267041.jpg ;-)
[18:41:52] <YuviPanda>	 hahaha!
[18:49:28] <DrAmir>	 irc://irc.freenode.net/#wikipedia-fa
[18:59:58] <YuviPanda>	 valhallasw: can you try doing /data/project/code-essay/tools/start-notebook.bash
[19:00:01] <YuviPanda>	 and see if that executes?
[19:00:21] <YuviPanda>	 should throw an error
[19:00:22] <YuviPanda>	 i think?
[19:34:55] <liangent>	 "Did you know that you can find a complete development environment on tools-dev server?" what does this mean?
[19:35:46] <YuviPanda>	 I' was never really sure :)
[20:17:50] <YuviPanda>	 bleeh
[20:18:03] <YuviPanda>	 hmm
[20:18:08] <YuviPanda>	 I'm getting output from a python process
[20:18:13] <YuviPanda>	 however piping it elsewhere does nothing
[20:18:19] <YuviPanda>	 any ideas what could be the cause/
[20:27:16] <anomie>	 YuviPanda: Is it on stderr instead of stdout?
[20:27:21] <YuviPanda>	 yeah
[20:27:23] <YuviPanda>	 fixed that
[20:27:31] <YuviPanda>	 also i've no idea whytf this is printing on stderr
[20:27:36] <YuviPanda>	 (ipython notebook)
[20:29:29] <anomie>	 Because some programmer did the python equivalent of fprintf( stderr, ... )?
[20:31:37] <YuviPanda>	 anomie: so this looks like output from the logging module, and by default that goes to stdout.
[20:31:41] <YuviPanda>	 *I* think
[20:31:45] * YuviPanda  checks

[20:33:45] <YuviPanda>	 anomie: tah! by default logs go to stderr
[20:33:50] <YuviPanda>	 not stdout
[20:34:07] <YuviPanda>	 which... now that I think about it, makes sense
[21:37:56] <valhallasw>	 YuviPanda: data to be piped goes to stdout, other stuff to stderr ;-)
[21:38:22] <YuviPanda>	 so i'm starting ipython notebook, which doesn't exit
[21:38:27] <YuviPanda>	 but it outputs the port number it uses
[21:38:30] <YuviPanda>	 which I want to capture
[21:38:32] <YuviPanda>	 and do things with
[21:38:37] <YuviPanda>	 I've no idea how to accomplish that
[21:38:46] <YuviPanda>	 without resorting to trickery like named pipes
[21:39:26] <YuviPanda>	 the alternative is for me to figure out explicitly which port to use by testing each of them
[21:39:32] <YuviPanda>	 and... I don't want to do that :|
[21:39:45] <YuviPanda>	 valhallasw: thoughts?
[21:40:02] <YuviPanda>	 the code is publicly accessible under the code-essay project, under tools/start-notebook.bash
[21:40:07] <YuviPanda>	 (and now notebook.py)
[21:40:29] <valhallasw>	 YuviPanda: can't you just tell it which port to use?
[21:40:48] <YuviPanda>	 valhallasw: I can, but it essentially is 'pick an unused port'
[21:40:54] <YuviPanda>	 that logic is already built into it
[21:40:59] <YuviPanda>	 so 1. I'll be duplicating code
[21:41:04] <YuviPanda>	 and 2. Race conditions
[21:41:08] <valhallasw>	 yes, because that's port ''. Good point.
[21:41:35] <valhallasw>	 so iirc it starts a webbrowser user webbrowser.open() or smth
[21:41:51] <YuviPanda>	 well, I've --no-browser
[21:41:54] <YuviPanda>	 so it doesn't start elinks :P
[21:41:56] <valhallasw>	 it makes sense to wrap that, because there is not really a webbrowser
[21:41:58] <valhallasw>	 right
[21:42:07] <valhallasw>	 but you can do BROWSER=your_script
[21:42:21] <YuviPanda>	 aaah
[21:42:26] <YuviPanda>	 that's.... very clever.
[21:42:35] <YuviPanda>	 since that'll pass me the URL
[21:42:39] <YuviPanda>	 I can extract the port from that
[21:42:47] <valhallasw>	 *and* it only does that once the server is running
[21:42:51] <YuviPanda>	 yeah
[21:42:58] <YuviPanda>	 but... why can't I just grep stderr
[21:43:06] <valhallasw>	 you should be able to
[21:43:11] <valhallasw>	 but this is much nicer! ;-)
[21:43:18] <YuviPanda>	 not necessarily
[21:43:27] <YuviPanda>	 the idea is that you run a simple command from your personal machine
[21:43:42] <YuviPanda>	 which ssh's to tools-labs, starts this, and returns the port number, and closes that ssh connection
[21:43:53] <YuviPanda>	 and then you open up another ssh connection, and port forward
[21:43:58] <YuviPanda>	 actually holy shit why do I need two
[21:44:00] <valhallasw>	 why not change the current connection?
[21:44:05] <YuviPanda>	 haha
[21:44:11] <valhallasw>	 should be possible (not sure how though under linux)
[21:44:27] <YuviPanda>	 well, I don't think that is too possible.
[21:44:29] <YuviPanda>	 and clearly will be convoluted
[21:45:11] <valhallasw>	 you can also consider dynamic port forwarding
[21:45:19] <YuviPanda>	 hmm?
[21:45:29] <valhallasw>	 then ssh acts like a socks5 proxy
[21:45:46] <YuviPanda>	 just doing that with -L should be easy enough, no?
[21:45:46] <valhallasw>	 more crap to set up on the client side though
[21:45:53] <YuviPanda>	 open up localhost:something and bam, you're done
[21:46:02] <YuviPanda>	 yeah, I want to minimize user work.
[21:46:46] <YuviPanda>	 valhallasw: actually both ssh connections need to remain open, sortof
[21:46:51] <YuviPanda>	 hmm, actually, maybe not
[21:47:07] <YuviPanda>	 I can just randomly choose a port number on the *client* side
[21:47:23] <YuviPanda>	 and in one connection, setup both the port forwarding and the running of ipython
[21:47:30] <YuviPanda>	 if that port fails I'll just try again
[21:47:41] <YuviPanda>	 but can I do port forwarding and setup with one connection?
[21:51:23] <YuviPanda>	 tada! that wroks
[21:51:42] <YuviPanda>	 woo
[21:56:15] <YuviPanda>	 valhallasw: can you run:
[21:56:16] <YuviPanda>	 ssh -L8888:127.0.0.1:8555 tools-login.wmflabs.org "/data/project/code-essay/tools/start-notebook.bash 8555"
[21:56:23] <YuviPanda>	 on your machine?
[21:56:28] <YuviPanda>	 and go to localhost:8555
[21:56:29] <YuviPanda>	 ?
[21:56:52] <valhallasw>	 YuviPanda: somewhat difficult, because I run linux as VM under windows
[21:57:12] <YuviPanda>	 why would that be difficult?
[21:57:25] <YuviPanda>	 you can run this on linux and that should work...
[21:57:38] <valhallasw>	 yes, but then I have to port forward to linux :p
[21:57:57] <valhallasw>	 and that doesn't work by default because ports are not further forwardable by default
[21:57:57] <YuviPanda>	 you don't have a GUI in linux?
[21:58:00] <valhallasw>	 anyway
[21:58:03] <valhallasw>	 no
[21:58:05] <YuviPanda>	 aaah
[21:58:08] <YuviPanda>	 ok
[21:58:14] <valhallasw>	 but I can try using putty
[21:58:30] <YuviPanda>	 I need to find a similar '1 click' or so solution for Windows
[21:58:35] <YuviPanda>	 valhallasw: sure!
[21:58:50] <YuviPanda>	 the port numbers need to be randomized, however.
[21:59:31] <valhallasw>	 oh my, all kinds of stuff being installed in my /home =p
[21:59:40] <valhallasw>	 aaaaaaand ot
[21:59:44] <valhallasw>	 aaaand it's broken
[21:59:55] <valhallasw>	 libzmq gives errors
[21:59:59] <valhallasw>	 but let's see how far we get
[22:00:35] <YuviPanda>	 that's not really errors
[22:00:38] <YuviPanda>	 it's only sortof errors
[22:00:45] <valhallasw>	 ok :p
[22:00:47] <YuviPanda>	 valhallasw: it's all self contained into a ~/code-essays
[22:00:53] <YuviPanda>	 rm -rf it and it's all gone
[22:00:54] <valhallasw>	 also can we please get all this stuff just installed as packages?
[22:00:59] <YuviPanda>	 valhallasw: yes yes!
[22:01:02] <valhallasw>	 and then virtualenv --use-system-site-packages
[22:01:05] <YuviPanda>	 yup
[22:01:27] <YuviPanda>	 especiall zm
[22:01:28] <YuviPanda>	 q
[22:02:10] <valhallasw>	 connection closed by remote host
[22:02:27] <valhallasw>	 restarting connection seems to work
[22:02:40] <valhallasw>	 aaand it's an ipython notebook
[22:03:07] <YuviPanda>	 valhallasw: wooot!
[22:03:11] <YuviPanda>	 :D
[22:03:18] <valhallasw>	 ImportError: No module named matplotlib
[22:03:19] <valhallasw>	 :-(
[22:03:26] <valhallasw>	 so much for trying my existing notebooks ;-)
[22:03:27] <YuviPanda>	 yeah, forgot to include that :D
[22:03:29] <YuviPanda>	 let me include
[22:03:51] <valhallasw>	 %pylab inline
[22:03:53] <valhallasw>	 should work
[22:04:05] <valhallasw>	 once it's installed
[22:04:06] <YuviPanda>	 yeah
[22:04:22] <YuviPanda>	 doesn't come by default?
[22:04:34] <YuviPanda>	 apparently not
[22:04:37] <YuviPanda>	 err
[22:04:50] <valhallasw>	 and then that wants numpy, and then you're screwed :p
[22:05:00] <valhallasw>	 because you *really* *do not* want to build numpy from scratch
[22:05:23] <valhallasw>	 unless you enjoy compiling fortran code
[22:05:34] <valhallasw>	 or have other masochistic tendencies
[22:05:52] <YuviPanda>	 haha
[22:06:02] <YuviPanda>	 well, we can install them as packages
[22:06:04] <YuviPanda>	 soon enough.
[22:06:17] <YuviPanda>	 once everything else works, getting that should be simple.
[22:06:27] <valhallasw>	 oh yes, and of course the issue of getting it to compile to the /right/ BLAS library. You know, Fortran include paths. *rants some more*
[22:06:37] <valhallasw>	 the packages luckily just work
[22:06:38] <YuviPanda>	 shouldn't ubuntu take care of that for me?
[22:06:40] <YuviPanda>	 yeah
[22:07:04] <valhallasw>	 I think, at least. I run it under windows, where there is just the 'one big magic scientific python distribution installer' button
[22:07:07] * YuviPanda  is actually compiling numpy now ;)

[22:07:20] <YuviPanda>	 oh building stuff on Windows is a lost cause, I think
[22:07:53] <valhallasw>	 that's why I enjoy the fact someone does this for the rest of the world
[22:08:00] <YuviPanda>	 :D
[22:08:08] <valhallasw>	 it's not that bad by the way - just install the free VC++ 2008
[22:08:17] <valhallasw>	 2010 might also work, not 100% sure about that
[22:08:23] <valhallasw>	 and then stuff just compiles
[22:08:30] <YuviPanda>	 :D
[22:08:36] * YuviPanda  has an original VS2005 launch t shirt

[22:08:49] <YuviPanda>	 2008 was the last one before the WPF move, IIRC
[22:16:15] <YuviPanda>	 valhallasw: so, pip can install numpy :D
[22:16:17] <YuviPanda>	 i just did
[22:18:47] <YuviPanda>	 valhallasw: also how do I kill the server once my ssh connection terminates?
[22:18:55] <YuviPanda>	 valhallasw: I guess they'll get SIGHUP and should terminate?
[22:19:18] <YuviPanda>	 but they don't
[22:19:26] <YuviPanda>	 there's a ton of zombie processes i'm cleaning up now
[22:19:27] <YuviPanda>	 by hand
[22:20:31] <valhallasw>	 YuviPanda: weird. 
[22:20:35] <valhallasw>	 no clue, really
[22:20:36] <YuviPanda>	 mmm, wrote an awk based kill -9 after a long time
[22:20:37] <YuviPanda>	 feels good
[22:20:38] <YuviPanda>	 :)
[22:20:54] <YuviPanda>	 valhallasw: can you rm -rf ~/code-essays, and run the script again?
[22:20:58] <YuviPanda>	 should get you pylab
[22:21:00] <valhallasw>	 tomorrow
[22:21:25] <YuviPanda>	 :)
[22:21:25] <YuviPanda>	 ok
[22:30:39] * ExcitedPanda  watches numpy build