[00:02] <jimbaker> i will say whitespace separated does has some wiggle room. the spec states "The output is a whitespace separated list of relation ids, if any."
[00:02] <jimbaker> reading     http://docs.python.org/library/stdtypes.html#str.split, it discusses the whitespace separation issue as well
[00:02] <hazmat> jimbaker, newline and whitespace separation aren't commonly understood as the same thing
[00:03] <hazmat> jimbaker, that doesn't talk about newlines
[00:04] <hazmat> jimbaker, i'm fine with it being newlines if thats what we're already using elsewhere and its easy for shell usage, but please update the docs to reflect it as newline separated
[00:04] <jimbaker> hazmat, it doesn't explicitly. but split does split on newlines as whitespace
[00:04] <jimbaker> as does the shell
[00:04] <hazmat> jimbaker, then its rather ambigious
[00:05] <hazmat> jimbaker, if you google around for common usage you'll see most people make a distinction
[00:05] <jimbaker> http://tldp.org/LDP/abs/html/special-chars.html#WHITESPACEREF
[00:05] <hazmat> jimbaker, yes its part of the definition, but its ambigious
[00:06] <hazmat> your outputting something specific, be specific in describing what it is
[00:07] <jimbaker> hazmat, should we adjust the spec accordingly? because it is ambiguous there
[00:07] <jimbaker> presumably, it should say "space separated" or "newline separated"
[00:07] <jimbaker> not "whitespace separated"
[00:08] <jimbaker> if only we chose to CSV, that's a format that's completely unambiguous ;) (and i'm of course completely joking)
[00:13] <jimbaker> hazmat, anyway, reading the review, i think the best thing to do is update the docstring to say "newline separated" and maybe add a comment this implies whitespace per our above conversation
[00:19] <hazmat> jimbaker, sounds good
[00:20] <_mup_> juju/relation-ids-whitespace-separated r534 committed by jim.baker@canonical.com
[00:20] <_mup_> Update docstring/comments to describe more precisely relation-ids, when using smart formatting, outputs newline separated ids; and that this is also whitespace separated
[00:55] <SpamapS> imo, newline separated is more consistent w/ the other tools
[00:55] <SpamapS> relation-list and relation-ids should basically work the same
[01:07] <_mup_> Bug #988115 was filed: upgrade-charm fails to change a symlink to a regular file <juju:New> < https://launchpad.net/bugs/988115 >
[01:15]  * SpamapS rings the bell
[01:15] <SpamapS> charm promulgated: loggerhead
[07:31]  * SpamapS tinkers w/ the mysql charm
[07:56] <imbrandon> morning SpamapS
[07:59] <SpamapS> imbrandon: you wake up with the roosters or just back from the karaoke bar?
[08:01] <SpamapS> the mysql charm is so F***'d ..
[08:01] <SpamapS> I should just rewrite it
[08:01] <imbrandon> just back from blondies's
[08:02] <imbrandon> working on the nginx charm i dident finish last night
[08:02] <SpamapS> imbrandon: which one is blondie, Bo or Luke? :)
[08:02] <imbrandon> hahaha
[08:02] <imbrandon> cooter
[08:02] <imbrandon> i like the overhauls
[08:04] <imbrandon> but it was Beau really :) /me careses the 1/8th scale diecast "General Lee" on his bookshelf
[08:10]  * imbrandon loves old moopars, general probably got me started but , '71 Cuda ... orignal "plumb crazy purple" with polished chrome hood pins shine against a flat black hood so you dont see the cowl induction only the numbers 427 in red along the ridge , chrome lined steel "fish gill" air vents just infront of the door seams on the front quarter , original black vynal interior and foe wood trim on the dash and wheel, 8 track with KISS detroit rock cit
[08:11] <imbrandon> and the body #'s and engine #'s gotta match, no throwin a Hemi in the other one with some sickers later, had to be original rebuild :)
[08:13] <imbrandon> ahhh i found one, .... /me sets his wallpaper for the day
[08:13] <imbrandon> http://3.bp.blogspot.com/_gfXupHOEhH0/SzkT2Urio5I/AAAAAAAANxo/D0pnw7csTPk/s640/1971+Plymouth+Hemi+Cuda.jpg
[08:14] <SpamapS> looks like a batmobile
[08:14] <imbrandon> that one has a full blower and widend drag wheels, but puretty much that car
[08:14] <imbrandon> is my dream car
[08:14] <SpamapS> a redneck batmobile :)
[08:14] <imbrandon> since i was like 15
[08:14] <imbrandon> lol
[08:15] <imbrandon> dude you ever stand next to one, thats pure stock not some crap some guy did to make it loud
[08:15] <imbrandon> and it will for real make your body shake from the low rumble at idle
[08:16] <imbrandon> 427 hemi with a huge carb and basicly no muffler stock
[08:16] <SpamapS> I was traumatized by my father's love of racing as a child, so I kind of hate all big motors
[08:16] <imbrandon> heh
[08:16] <imbrandon> i only like mopar
[08:16] <imbrandon> no idea why really
[08:16] <SpamapS> so nice of my dad to add half of the drag strip he owns as part of my inheritance when he kicks the bucket
[08:16] <imbrandon> but like 67ish to 73ish mopars of any kind
[08:17] <imbrandon> heh
[08:17] <imbrandon> sell it out, its what i did with my dads computer shop last year, i dont wanna run a repair shoow
[08:17] <SpamapS> Fitting, since its in "Clint" Texas
[08:17] <imbrandon> shop
[08:17] <SpamapS> http://elpasomotorplex.com/Home/
[08:18] <imbrandon> oh wow, i think i been there
[08:18] <SpamapS> Wow, gotta get dad to fix that
[08:18] <imbrandon> east tex
[08:18] <imbrandon> like near longview and kilgore ?
[08:18] <imbrandon> oh wow, default wordpress pagelines theme
[08:18] <SpamapS> http://elpasomotorplex.com/
[08:18] <SpamapS> ahh thats the real page
[08:18] <imbrandon> well almost default
[08:18] <imbrandon> heh
[08:18] <SpamapS> ok :)
[08:19] <imbrandon> ahh much better
[08:19] <imbrandon> heh i seen the leaf logo at the bottom of the other and was like ohhhh pagelines, i know their themes
[08:19] <imbrandon> heh infact i own the same one
[08:19] <imbrandon> not the "right one" the wrong one
[08:20] <imbrandon> heh
[08:20] <SpamapS> http://elpasomotorplex.com/images/igallery/resized/801-900/IMG_8260-886-500-400-80.JPG
[08:20] <SpamapS> There's my dad's car
[08:20] <imbrandon> mmmm summit
[08:20] <SpamapS> alcohol :)
[08:20] <imbrandon> SWEET
[08:20] <SpamapS> Yeah, my sons love it
[08:20] <imbrandon> yea those frakers are nuts
[08:20] <SpamapS> I'm like, meh :-P
[08:20] <imbrandon> they are worse than bull riders
[08:21] <SpamapS> 0-200mph before you can blink twice
[08:21] <imbrandon> yea i love it too, well when there i dont follow it at all
[08:21] <imbrandon> and cant stand nascar
[08:21] <imbrandon> but funnycars and stuff like that or drag bikes
[08:21] <imbrandon> those are all fun to watch
[08:21] <imbrandon> in person
[08:22] <imbrandon> tv , meh
[08:22] <SpamapS> yeah I feel the same way about hockey :)
[08:22] <imbrandon> you got to feel the engines as the roar oir it dont count, and smelll the fuel
[08:23] <imbrandon> yea i'm probably the strangest dude you will ever meat sports wise, i was all into it up till i graduated HS, no i never even watch ANY on tv, not football, baseball, dont keep up , no march madness
[08:23] <imbrandon> etc
[08:23] <imbrandon> i'll PLAY baseball, but thats the extent, seriously
[08:24] <imbrandon> like i get adgitated at the bar and everyone is into the football game, i just go home
[08:24] <imbrandon> heh
[08:25] <imbrandon> no idea, why but like i just am , meh about them all, all my buddies are like , "are you gay" ..... even the gay ones !! hahaha
[08:25] <imbrandon> they mean it with love those before someone gets upset :)
[08:25] <SpamapS> so, moodle has to be the most sigusting, sad, slow php app ever made
[08:26] <SpamapS> sigusting.. worse thand disgusting
[08:26] <imbrandon> hahah ever run stock drupal 6 with no mods?
[08:26] <imbrandon> about 20 second page load times
[08:26] <imbrandon> like ootb, default content
[08:27] <imbrandon> give moodel some apc and a bit of tweaking i bet it could hum tho, i've always ment to get deeper into it, i've installed it like 5 times over the years but never finish using it/filling it out
[08:27] <imbrandon> moodle*
[08:28] <SpamapS> I'm just poking at the half-done charm somebody submitted to see if I can make it go
[08:28] <imbrandon> ahh
[08:28] <SpamapS> so far.. no luck
[08:28] <SpamapS> http://ec2-50-112-14-19.us-west-2.compute.amazonaws.com/login/index.php#maincontent
[08:28] <imbrandon> :)
[08:28] <SpamapS> thats as far as I've gotten
[08:29] <imbrandon> crap this is slow as hell
[08:29] <imbrandon> its not a micro is it
[08:29] <imbrandon> hhe
[08:29] <imbrandon> ahhh
[08:29] <imbrandon> it loosk like
[08:29] <imbrandon> let me check
[08:29] <imbrandon> one sec
[08:30] <SpamapS> m1.small
[08:30] <SpamapS> trying apc now
[08:30] <imbrandon> oh yea with no apc even phpinfo will be slow
[08:31] <imbrandon> but there is something up with the config and the current domain
[08:31] <SpamapS> yeah
[08:31] <imbrandon> its missing an include it looks like
[08:31] <imbrandon> or something
[08:31] <SpamapS> links are wrong or something
[08:31] <imbrandon> cant spot it tho
[08:31] <imbrandon> wonder if they do it like wp, idiots
[08:31] <SpamapS> yeah I'm reading the code
[08:31]  * imbrandon was glad i could bypass it in wp without a core hack
[08:32] <imbrandon> probably cant in moodle
[08:33] <imbrandon> and holly mother of jesus , look at the way yui is loading with inc in cdata
[08:33] <imbrandon> that another reason its takin so long to render
[08:33] <SpamapS> "render"
[08:33] <imbrandon> heh
[08:33] <SpamapS> thats an overstatement :)
[08:33] <imbrandon> right
[08:33] <imbrandon> hehe
[08:34]  * imbrandon opens dev console
[08:35] <imbrandon> in dev console one of the styles.php is empty and one javascript.php is
[08:35] <SpamapS> no errors of course
[08:36] <SpamapS> we wouldn't wnat it to be easy to track this down
[08:36] <imbrandon> might have something to do with it or lead ya where does if you care that much
[08:36] <imbrandon> lol yea
[08:36] <imbrandon> one javascript.php fills, and all the yui-combos do but one php css dont and oen php js dont
[08:37] <imbrandon> afk a few, gotta get a little fruity pebbles or something brb
[08:39] <SpamapS> just doing one page hit pegs the CPU
[08:39] <SpamapS> even w/ apc
[08:39] <imbrandon> did you restart the php after installing it
[08:39] <imbrandon> e.g is it actually caching
[08:39] <imbrandon> i goof that part regularly
[08:40] <imbrandon> quick phpinfo is the easy way to tell
[08:40] <SpamapS> yeah it is
[08:40] <imbrandon> or php-cgi -m on the cli
[08:40] <imbrandon> but it will segfault
[08:40] <imbrandon> hrm
[08:40] <SpamapS> http://ec2-50-112-14-19.us-west-2.compute.amazonaws.com/phpinfo.php
[08:41] <imbrandon> hrm, nice
[08:41] <SpamapS> I set error_reporting to E_ALL
[08:41] <SpamapS> and display_errors On
[08:42] <imbrandon> fskin sousin is enabled that might screw with it, gotta kill it with drupal , well drupal contrib
[08:42] <SpamapS> ugh but they abuse error_reporting all over
[08:42] <imbrandon> heh
[08:42] <imbrandon> error_reporting.scream
[08:42] <imbrandon> i thing
[08:42] <imbrandon> think*
[08:43] <imbrandon> will overide all of them if you set it to 1
[08:43] <SpamapS> [Wed Apr 25 08:42:58 2012] [error] [client 76.94.217.164] PHP Fatal error:  Allowed memory size of 134217728 bytes exhausted (tried to allocate 72 bytes) in /usr/local/share/moodle/lib/moodlelib.php on line 7837
[08:43] <imbrandon> i *think* its been foreever since i used it
[08:43] <SpamapS> YAAAY
[08:43] <imbrandon> ahhh
[08:43] <imbrandon> yea
[08:43] <imbrandon> that would make it slow as hell too
[08:43] <imbrandon> bump that bad boy to 512 like it should be
[08:43] <imbrandon> and i bet things get better
[08:43] <imbrandon> or atleaste show an error
[08:44] <SpamapS> 24082 www-data  20   0  463m 139m  23m R 92.2  8.4   0:06.88 apache2
[08:44] <SpamapS> so awesome
[08:44] <SpamapS> I forgot how much fun it was to watch PHP eat your server
[08:44] <imbrandon> heh thats only 23m
[08:44] <imbrandon> resident
[08:45] <SpamapS> [Wed Apr 25 08:44:50 2012] [error] [client 76.94.217.164] PHP Fatal error:  Allowed memory size of 536870912 bytes exhausted (tried to allocate 32 bytes) in /usr/local/share/moodle/lib/setuplib.php on line 367
[08:45] <SpamapS> debug is for winning!
[08:45] <imbrandon> HAHA
[08:45] <SpamapS> I set the moodle debug = 1 flag
[08:45] <imbrandon> wow thats got to be an infinate look
[08:45] <SpamapS> This yak is just about bald
[08:45] <imbrandon> loop
[08:46] <imbrandon> what version is that , i'm gonna grab it clean and try on my server
[08:46] <imbrandon> just to see
[08:46] <SpamapS> git
[08:46] <SpamapS> Here I'll push the charm up
[08:46] <imbrandon> oh no wonder, git branch or tag ?
[08:46] <imbrandon> kk
[08:47] <SpamapS> stable tag I thin
[08:47] <SpamapS> k
[08:47] <SpamapS> forgot to check ;)
[08:47] <imbrandon> btw dont bother reviewing the drupal one yet, decided to go ahead and tear it down to use the nginx base server as a sub
[08:47] <SpamapS> cool!
[08:47] <SpamapS> remember tho..
[08:48] <imbrandon> if ya get there today
[08:48] <SpamapS> subs can't open ports
[08:48] <imbrandon> k
[08:48] <SpamapS> tho you can work around that by making the primary open ports for the sub
[08:48] <imbrandon> thats fine, i was gonna have the nginx actualy control the whole 80 and 8080 configs
[08:48] <imbrandon> and just dropa drupal specific in sites/enabled
[08:48] <SpamapS> lp:~clint-fewbar/charms/precise/moodle/slow
[08:48] <imbrandon> no need for ports in ther
[08:49] <SpamapS> git checkout MOODLE_22_STABLE
[08:49] <imbrandon> , even so though couldent we kinda cheat and do like on relation-join , drupal config set nginx ineedport80 , nginx config-changed notices and does it ?
[08:49] <imbrandon> hehe
[08:50] <SpamapS> yeah thats how you do it
[08:50] <imbrandon> ahh kk
[08:50] <SpamapS> as long as primary service opens port, its all good
[08:50] <SpamapS> (and devs are working on fixing this)
[08:50] <imbrandon> erm
[08:50] <imbrandon> wait only the primary
[08:51] <SpamapS> why oh why doesn't php have a real debugger yet? :-P
[08:51] <SpamapS> I mean I know it does
[08:51] <SpamapS> but I want it to be easy to use
[08:51] <imbrandon> fudge, i really wanted to do this set on jeos. even if its not for all charms it kinda makes sense here
[08:51]  * SpamapS installs xdebug to see what havoc it plays
[08:51] <imbrandon> SpamapS: xdebug
[08:52] <imbrandon> SpamapS: thats ok, the jeos will just have the job of manageing all the ports then, not giving that up, cuz i'm dropin them all in the store at once TODAY i hope
[08:53] <imbrandon> all being the jeos-webbase with either zend-server-ce 80 and spdy OR nginx with phpfpm for 80
[08:53] <imbrandon> thats 3
[08:54] <imbrandon> then another sub of jeos for the nginx rev prox 30s cache
[08:54] <imbrandon> to one of those others
[08:54] <imbrandon> then drupal or wp base with no default site, just install screen
[08:55] <imbrandon> and then my blog or omg or something on top of that ( but sub of jeos )
[08:55] <imbrandon> thats how i got it worked out
[08:55] <jamespage> morning folks
[08:55] <imbrandon> and then others can be added and removed too but that is some nice default options
[08:55] <imbrandon> morning jamespage
[08:55] <jamespage> hi imbrandon
[08:56] <jamespage> bit of late night charming SpamapS :-)
[08:56] <imbrandon> i'm rubbin off on him, soon he'll be going on 3 hours sleep 4 or 5 nights a week :)
[08:56] <imbrandon> lol
[08:57] <SpamapS> jamespage: indeed.. not sure why I'm poking at moodle when the real problem is mysql :-P
[08:57] <jamespage> lol
[08:57] <SpamapS> imbrandon: I'm training up for having an infant in the house
[08:58] <imbrandon> oh SpamapS been watching the omg load and such, it hit 10.x earlier for almost an hour, it was pushing 25Mb/s out and 13Mb/s in said byobu
[08:58] <jamespage> SpamapS, it will be a real test to see if you can charm whilst settling a baby at 2am!
[08:58] <imbrandon> i allllllllllmost did a juju add-unit
[08:58] <SpamapS> jamespage: I may need Siri to help w/ that. :)
[08:58] <imbrandon> but it calmed back down, was just before you poped in
[08:59] <SpamapS> imbrandon: hah cool
[08:59] <imbrandon> funny thing was tho with the 10.x load , i still it the page fine
[08:59] <imbrandon> without any extra lag, so thats why i dident rush to add a unit
[09:00] <imbrandon> if it starts to creep tho i will, running one extra for a day or two wont be much
[09:00]  * imbrandon is suprised its holding up on two smalls as well as it has been
[09:01] <SpamapS> jamespage: I may have fixed the problem hive had w/ mysql actually
[09:01] <imbrandon> i mean i know this cruft is good but damn i'm thinking about starting a shared host on a m1.small and selling 100000000 accounts at $1
[09:01] <SpamapS> should probably abandon moodle and try hive instead :-P
[09:01] <imbrandon> and come out ahead
[09:01] <imbrandon> lol
[09:02] <imbrandon> SpamapS: did you see a small patch from marcoceppi for mysql charm
[09:02] <imbrandon> he said he had one for you to fix some utf8 things we got to fix wevery redeploy
[09:02] <jamespage> SpamapS, I'll have to read - that was at least a month about and my ringbuffer memory has looped a few times since :-)
[09:03] <SpamapS> jamespage: I poked at the hive charm in new-charm .. looked good to me
[09:03]  * SpamapS officially gives up on moodle
[09:03] <jamespage> SpamapS, sweet - i'll promulgate it then
[09:04] <jamespage> SpamapS, as hadoop, hbase, zookeeper and hive will all be in the charm store I was going to do a trivial update to all of their readme's to stop using a local repo
[09:04] <SpamapS> cool
[09:05] <SpamapS> I am not so excited about cs:* until we get an ability for people to fork a running service's charm
[09:05] <SpamapS> https://code.launchpad.net/~clint-fewbar/charms/precise/mysql/add-binlog-format/+merge/103427
[09:05] <SpamapS> There's the fix
[09:06] <jamespage> SpamapS, nice
[09:06] <SpamapS> imbrandon: I don't see anything from marco, I'll ping him tomorrow
[09:07] <SpamapS> sad.. us-west-2 seems to be more laggy than us-east-1, despite being baout 1000 miles closer to me. :-P
[09:07] <SpamapS> imbrandon: mysql and charsets is always a nightmare. :-/
[09:08] <imbrandon> yea, he said the fix is in that shelr.tv prov feed he gave me and you
[09:08] <SpamapS> ok, sleep
[09:08] <imbrandon> just fyo
[09:08] <imbrandon> gnight
[09:09] <imbrandon> see ya in a few hours /me will still be here
[09:28] <jamespage> hive charm promulgated - thanks bbcmicrocomputer!
[09:29] <bbcmicrocomputer> jamespage: ah no problem :)
[14:46] <jcastro> imbrandon: you're working on the nginx subordinate right?
[15:57] <imbrandon> jcastro: yea
[15:57] <imbrandon> like right now
[15:57] <imbrandon> will be done before lunch
[15:58] <imbrandon> jcastro: and an apache zend server one too that will be done same time as it shares 90% of config
[16:10] <jcastro> imbrandon: so what I was thinking about, is to make an nginx ppa for 1.2.0 an option in the charm, for people who want to use it, that would be sweet.
[16:11] <imbrandon> yea i'm definaly using the upstream nginx ppa
[16:11] <imbrandon> i do on my site and depends on some of the new stuff
[16:11] <jcastro> <3
[16:11] <imbrandon> its packaged by upstream and stable release
[16:11] <imbrandon> so should be good too
[16:11] <imbrandon> and i've been using it since it came out on my site
[16:12] <imbrandon> ( i have the daily ppa on )
[18:40] <SpamapS> jimbaker: so, I'm experimenting with relation-ids
[18:40] <jimbaker> SpamapS, cool
[18:40] <SpamapS> I can getthe relation id..
[18:40] <SpamapS> but I want to re-run a hook as if it was changed again
[18:41] <SpamapS> so I need to set, like, JUJU_REMOTE_UNIT
[18:41] <SpamapS> any idea?
[18:42] <jimbaker> SpamapS, you can only do settings on the local unit
[18:43] <SpamapS> hm, so do I do relation-list ?
[18:43] <SpamapS> ahh relation-list is what I want. :)
[18:43] <jimbaker> SpamapS, good :)
[18:44] <SpamapS> jimbaker: basically I want to write a "refresh *everything*" script
[18:45] <jimbaker> SpamapS, that's a bit challenging in the scope of what is provided - you would need to ensure this runs in some hook on a service unit for every relation
[18:46] <jimbaker> but if you can use "jitsu do" (which i will merge in shortly to juju-jitsu), it's pretty straightforward
[18:46] <jimbaker> given it's rather powerful capabilities ;)
[18:46] <jimbaker> my feeling is that this would be a good place to start at least
[18:46] <SpamapS> jimbaker: jimbaker http://paste.ubuntu.com/946143/
[18:47] <SpamapS> jimbaker: there's my first attempt
[18:47]  * SpamapS tries it
[18:47] <SpamapS> jimbaker: note that is the whole upgrade-charm hook
[18:48] <SpamapS> doh, have to enumerate the relation name myself
[18:50] <jimbaker> SpamapS, yes, there's no generic support for getting relation names from juju itself. the expectation is that charms know the relation name. leaving this issue to some sort of charm tooling that can work with the metadata.yaml
[18:51] <SpamapS> which is what I'm writing right now :)
[18:51] <SpamapS> actually
[18:51] <SpamapS> no
[18:51] <SpamapS> I only have one relation
[18:51] <SpamapS> I can be explicit in this case
[18:52] <jimbaker> SpamapS, sounds good
[18:52] <jimbaker> SpamapS, re line 14 of that paste, does that exec the script?
[18:53] <SpamapS> yes
[18:53] <SpamapS> but I think it may fail because it doesn't do what I would have expected, which is that the presence of JUJU_RELATION_ID would be enough to give me a context
[18:54] <jimbaker> SpamapS, right, the overall script has that flaw
[18:55] <jimbaker> it needs to make changes on relation settings, which is restricted to the local unit
[18:55] <SpamapS> huh?
[18:56] <SpamapS> It actually just needs to do a few relation-get's
[18:56] <jimbaker> SpamapS, if that's the case... then fine, you can do that on remote units of course
[18:57] <SpamapS> yeah I think I just need to add an explicit '-r $JUJU_RELATION_ID' to the hook it is calling
[18:58] <SpamapS> which, seems like a bug
[18:58] <SpamapS> if I have no context, but I do have $JUJU_RELATION_ID .. that should be equivilent to passing -r $JUJU_RELATION_ID
[19:02] <SpamapS> jimbaker: jitsu do seems too general. I don't know exactly what it is for.
[19:03] <jimbaker> SpamapS, the way to think about jitsu do is that it allows you to execute the command + args as if it were a hook
[19:04] <jimbaker> so for example, $ jitsu do mysql/0 relation-get -r db:0 -
[19:04] <jimbaker> this is as if this were called as a nonrelational hook (a relational hook automatically gets an implied relation, including $JUJU_RELATION, $JUJU_RELATION_ID)
[19:05] <SpamapS> so, I had hoped that implied relation would simply be carried in the env vars mentioned...
[19:06] <SpamapS> that would make things more self contained.. I don't want to included jitsu in charms just so they can do the right thing and refresh their relations on upgrade-charm
[19:06] <SpamapS> actually IMO juju should do it automatically.. but .. baby steps
[19:09] <SpamapS> jimbaker: here's the diff I'm adding to the munin charm..
[19:09] <SpamapS> jimbaker: http://paste.ubuntu.com/946171/
[19:09] <jimbaker> SpamapS, agreed on that. jitsu do for now should be considered a useful backdoor for trying out things
[19:09] <SpamapS> jimbaker: I shouldn't need the '-r' .. its implied by the environment.
[19:10] <SpamapS> this works btw
[19:10] <SpamapS> its just that it shouldn't need to be so invasive
[19:13] <jimbaker> SpamapS, i understand your perspective re not needing -r if JUJU_RELATION_ID is specified *or* changed
[19:14] <jimbaker> this could be done in a reasonable way. right now, the implied relation is special since the invoker knows it's in a relation  hook context
[19:52] <hazmat> m_3, jamespage you guys have talks in already for strata ny+hadoopworld?
[19:53] <hazmat> just got  a reminder that their due on may7
[19:53]  * hazmat finishes up his presentation for openstackdc meetup
[20:05] <jcastro> http://podcast.ubuntu-uk.org/live/
[20:05] <jcastro> I am on this podcast talking about juju!
[21:00] <jamespage> hazmat, sure have
[21:00] <jamespage> arosales and m_3 are working on a charm school as well
[21:02] <arosales> ya, hopefully we can have a tutorial session at Strata/Hadoop world on how to build off James' Haddop charms, and also use the charms in different deployments.
[21:03] <arosales> Hope to have that submitted by end of the week.
[22:17] <jcastro> Announcing the UDS Charm Contest!
[22:17] <jcastro> http://cloud.ubuntu.com/2012/04/announcing-the-uds-juju-charm-contest-for-uds-attendees/
[22:17] <imbrandon> ^^
[22:17] <imbrandon> so mine
[22:21] <SpamapS> hey
[22:21] <SpamapS> somebody put a dell logo on my laptop
[22:22] <imbrandon> heh
[22:22]  * imbrandon is soooooooooooooooooooooo got one of those this time
[22:22]  * imbrandon polishes more turds to combine them alll
[22:22] <imbrandon> :)
[22:23] <imbrandon> and if not least i lost fair and square from the start this time :)
[22:23] <imbrandon> heh
[22:23]  * imbrandon gets food
[22:30] <jimbaker> that xps 13 looks rather rather nice
[22:39] <SpamapS> hrm.. poking at making an autoscaler w/ munin+juju .. one real pain is that you have to know *all* of the SSH keys that you want to be able to use to talk to juju before you bootstrap
[22:40] <jcastro> SpamapS: Got ~5 for G+?
[22:40] <jcastro> I have quick questions
[22:41] <SpamapS> but
[22:41] <SpamapS> the daisies
[22:41] <SpamapS> the pirates will be sad
[22:41] <SpamapS> will try.. fighting pink eye onset and lacking a headset :)
[22:43] <imbrandon> steal a patch from on of dem pirates
[23:12] <lifeless> SpamapS: pink eye... what have *you* been up to ?
[23:14] <SpamapS> lifeless: no good
[23:33] <RichardRaseley> I have a question regarding scaling things like mongodb or cassandra via juju. If you look at the command juju add-unit -n 2 cassandra its purpose is clear - but how does it know what computers to configure those two additional nodes on?
[23:34] <RichardRaseley> Is that specified after the fact (I haven't actually ran these commands, FYI)?
[23:34] <RichardRaseley> Or rather, as part of the input asked for after the command?
[23:39] <SpamapS> RichardRaseley: it decides where to scale onto the same way it decides when you run 'deploy'
[23:40] <SpamapS> RichardRaseley: the only difference is that there is already one unit in that service.
[23:40] <SpamapS> RichardRaseley: so, it uses a combination of the placement policy and constraints
[23:41] <RichardRaseley> If I am SSH'd into "server00" and I do a cassandra deploy - it won't go to that particular server? Where are those policies defined?
[23:41] <SpamapS> RichardRaseley: the default is to place everything on an "available" machine, or to ask to have one provisioned if there are none. The default constraints are different per provider but basically mean "any machine"
[23:41] <RichardRaseley> I am just looking over the juju.ubuntu.com site and I don't see that information...
[23:41] <SpamapS> RichardRaseley: juju is in charge of provisioning, so by the time you can ssh in, its too late. :)
[23:42] <RichardRaseley> Obviously if I have 1000 production machines in an environment - I don't want JuJu grabbing a mailserver and make it a mongodb node.
[23:42] <SpamapS> RichardRaseley: since juju's use case is more "cloud like", it expects to be able to install a new OS on a clean machine every time.
[23:42] <RichardRaseley> What controls those placement policies?
[23:43] <SpamapS> RichardRaseley: the provider defines the constraints that are available.. and there are only two placement policies right now.. the "available machine" and "local". Local is used for doing local dev.
[23:43] <RichardRaseley> What does "provider" define in this context.
[23:43] <RichardRaseley> ?*
[23:43] <SpamapS> well realistically, ec2, or maas
[23:44] <RichardRaseley> If I want to deploy Keystone on the server to which I am currently logged into via JuJu, I cannot do that?
[23:44] <SpamapS> no, juju does not do that currently
[23:44] <RichardRaseley> Similarly - if I want to create a mongodb deployment, then extend it to server00 and server01, I cannot do that?
[23:45] <RichardRaseley> Oh, wow - that is hugely dissapointing.
[23:45] <SpamapS> RichardRaseley: why's that?
[23:45] <SpamapS> Those machines are already busy
[23:45] <SpamapS> If you have clean machines
[23:45] <SpamapS> use MaaS
[23:45] <RichardRaseley> No, they aren't - I just created them.
[23:45] <RichardRaseley> They aren't physical machines.
[23:45] <SpamapS> Ahh, well if you create them w/ MaaS, then yes you can do that. :)
[23:45] <SpamapS> And if they're VM's, then let juju create the VM's.
[23:46] <SpamapS> RichardRaseley: It would not be hard to write a juju provider which SSH's to boxes and runs juju's agent.
[23:46] <RichardRaseley> That just doesn't make sense to me. I mean, that is awesome that it supports that but it seems a no-brainer that I should be able to specify (IP, FQDN, etc.) the machines that I want to extend the services to.
[23:46] <RichardRaseley> Not hard for a developer. =]
[23:47] <SpamapS> RichardRaseley: Why exactly are you so interested in creating the VM's yourself?
[23:47] <RichardRaseley> Does it matter why?
[23:47] <RichardRaseley> There could be any number of reasons.
[23:47] <SpamapS> yeah, it might motivate me to advocate for that use case :)
[23:48] <RichardRaseley> Well, for example we are using a combination of VMware & Hyper-V so I don't know if Juju hooks into that.
[23:48] <RichardRaseley> There are change control issues with allowing an automated script to hook into production infrastructure.
[23:48] <SpamapS> No, it doesn't. It needs an API to talk to, and right now, it only speaks EC2 and MaaS
[23:48] <SpamapS> RichardRaseley: only one machine, the juju provisioning agent machine, has access to the infrastructure.
[23:49] <SpamapS> RichardRaseley: and those change control issues are handled by keystone and nova if you use OpenStack.
[23:49] <RichardRaseley> Right, but it is an "unknown" as far as most people would be concerned. Our process for provisioning VMs is "X" and deviation from that is painful for operations to overcome (that is a whole other issue).
[23:49] <SpamapS> Whatever access creds you're using can be limited to numebr of nodes and types and such.
[23:50] <SpamapS> RichardRaseley: So you'd like to manually define the VM's that are available to juju. I think thats a worthy goal, you're the 3rd person to ask for that functionality.
[23:51] <RichardRaseley> Yes, it *could* be done that way - but that isn't the way I would want to use it in this case. I think having the ability to interface with MaaS or with EC2 or OpenStack is awesome, but also being able to say something like "juju mogodb deploy --server server00"
[23:51] <SpamapS> RichardRaseley: one option is to make those VMs owned by MaaS, so they act more like real servers. The only requirement is that they be able to PXE boot from the network where MaaS is.
[23:52] <RichardRaseley> then "juju add-unit -n 2 mongodb --server server01, server02"
[23:52] <SpamapS> ugh
[23:52] <RichardRaseley> PXE and WoL, right?
[23:52] <SpamapS> RichardRaseley: the WoL isn't required, just makes things easier
[23:53] <RichardRaseley> What if you have existing PXE servers in the environment doing other things (WDS for example)?
[23:53] <SpamapS> RichardRaseley: --server is a bit of a problem. You are limiting the scale quite a bit. How about when you deploy, you say 'deploy --constraints mem=10G,cpu=10' and then add-unit will find the machines that suit that?
[23:53] <RichardRaseley> I don't recall if you can define multiple PXE servers in a DHCP option.
[23:53] <SpamapS> RichardRaseley: you can tell DHCP servers to only respond to known MAC's
[23:54] <SpamapS> RichardRaseley: or just --constraints class=mongoserver .. and then have the machines classified arbitrarily
[23:54] <RichardRaseley> But what is creating the pool of available machines?
[23:55] <SpamapS> RichardRaseley: the reason --server server01 is in your head is that you are limited by your organization's unwillingness to actually meter their infrastructure with a real cloud-like solution.
[23:55] <SpamapS> RichardRaseley: the success of amazon comes largely from letting go of "but we can't let just anybody take one of our precious compute units!" and instead just billing everybody who uses things.
[23:55] <RichardRaseley> We aren't "billing" anyone - it is all internal use.
[23:56] <RichardRaseley> And they don't use a chargeback model.
[23:56] <SpamapS> So why would you be so concerned about letting services that need resources take the resources they need?
[23:56] <SpamapS> And then putting governor's on.. a-la "that environment only gets X compute and Y ram"
[23:57] <SpamapS> RichardRaseley: I'm not trying to change your business model, but rather, let you know the business model we've targetted first w/ juju.
[23:57] <RichardRaseley> Not going to disagree with you in principal - but I think having that flexibility would speed adoption of Juju which will have a net effect of moving things faster towards that model.
[23:58] <RichardRaseley> I can say "Look what Juju can do Mr. Manager" "Holy, crap really?" Yes, and if I deployed Maas (etc.) it could do that +X and +Y"
[23:58] <SpamapS> Perhaps, but at what cost? Instead of migrating to OpenStack, users now have this clunky juju thing on top of their old servers that can't scale.
[23:59] <SpamapS> RichardRaseley: MaaS is laserbeam-targetted at making it really easy to deploy a "seed" openstack cloud.
[23:59] <SpamapS> RichardRaseley: something where you take a few real boxes and throw openstack up and then show how juju can manage apps easily ontop of openstack.
[23:59] <SpamapS> (in theory.. :)