[01:23] <marcoceppi> heh, so I did a stupid thing. I bootstrapped on EC2 and deployed a crap load of charms to test something. Then I left the office and went home: ERROR Invalid SSH key
[01:24] <marcoceppi> Can I destroy from EC2 without Juju getting sad faced
[03:13] <_mup_> juju/hook-alias-expansion-redux r468 committed by kapil.thangavelu@canonical.com
[03:13] <_mup_> scheduler stop values are ignored if the scheduler is running
[10:04] <jamespage> morning all
[14:04] <jcastro> anyone have the spec for the automagic dependency check handy?
[14:05] <jcastro> lp:~hazmat/juju/auto-magic-dependency-spec seems like it's the canonical location
[14:06] <hazmat> jcastro, that's not quite the same thing..
[14:06] <jcastro> ah ok
[14:06] <hazmat> jcastro, i mean well.. it depends on what you mean ;-)
[14:06] <jcastro> do we have a spec for juju recommending dependencies on a charm?
[14:07] <jcastro> I know we've talked about it and the answer was "some day", just wondering if we had gotten beyond that
[14:07] <hazmat> jcastro, if you mean deploy dependencies automatically.. that's what its doing.. but that isn't practical it turns out
[14:07] <jcastro> hazmat: oh, ok, I am trying to answer this guy's question:
[14:07] <jcastro> http://askubuntu.com/questions/109042/how-would-one-know-a-charms-dependencies
[14:08] <jcastro> and my answer so far is "yeah we don't do that yet." but I wanted to link him to something.
[14:08] <jcastro> or in this case, I guess explain why we don't want that anymore?
[14:09] <hazmat> jcastro, so knowledge is easy to satisfy.. i'd like put the charmrunner plans onto the charm browser, so its more clear
[14:09] <jcastro> ok so the tldr; is "it'll be in the charm browser?"
[14:10] <hazmat> jcastro, but in terms of deploying it into an environment.. say you want to deploy mediawiki into an environment, and you have 3 mysql services already there.. which one do you use to satisify the dep, or do you deploy a new one.
[14:10] <jcastro> ah, right
[14:10]  * hazmat needs some coffee
[14:11] <jcastro> indeed, caffeine time
[14:14] <jcastro> ok I left the guy an answer, any one feel free to submit an edit to clarify or correct.
[14:19] <hazmat> jcastro, re the charm browser currently, you can got to any charm page (probably via search) on a given charm page, you can click an interface and it takes to you a page that shows who provides and requires a it
[14:20] <jcastro> yeah but we can improve that a bit
[14:20] <hazmat> also the search works directly from the chrome/chromium's url bar if you tab on the site name .. which is magically awesome ;-)
[14:20] <jcastro> like, mediawiki says it requires memcache, but it doesn't really
[14:21] <jcastro> the real issue I think here is that we don't have READMEs for the basic older charms
[14:21] <jcastro> that would make all of this go away
[14:21] <hazmat> true, mediawiki doesn't any of those as optional
[14:21] <hazmat> https://bazaar.launchpad.net/~charmers/charms/oneiric/mediawiki/trunk/view/head:/metadata.yaml
[14:21] <hazmat> but that's partly because the default is optional anyways at this point..
[14:22] <jcastro> hmm, maybe we should highlight the Configs and Interfaces more in the web UI than the hooks
[14:22] <jcastro> if I'm consuming a charm do I care about the hooks?
[14:22] <jcastro> I would think I'd care about the config options I can set and if I need something else
[14:23] <jcastro> anyway, for now I think just asking people to go back and README their charms is a good 90% fix
[14:23] <jcastro> and we're heavily recommending it on the new charms
[14:24] <jcastro> still, the guy asked a good question
[14:36] <hazmat> jcastro, docs need some love ;-)
[14:39] <jcastro> I know
[14:46] <_mup_> Bug #944038 was filed: Inform users they have an environment.yaml with the same control bucket value <juju:New> < https://launchpad.net/bugs/944038 >
[17:28] <SpamapS> hazmat: :(
[17:28] <SpamapS> hazmat: you're suggesting dropping my favorite part of Jim's spec
[17:29] <SpamapS> hazmat: juju do is not an anti-use case. I've been asking for it for a *long* time.
[17:30] <SpamapS> hazmat: and it has always been suggested that these commands would be usable by cron jobs, ssh, etc.
[17:40] <hazmat> SpamapS, cron jobs can execute hooks directly and use the cli api.. but the very notion of juju do is pretty crackful imo
[17:41] <hazmat> SpamapS, post the team meeting currently ongoing i was hoping to do a g+ with jimbaker if your up for it
[17:41] <SpamapS> hazmat: cron jobs would have to set the socket path.. and other weird stuff. wrapping that is nice.
[17:42] <SpamapS> hazmat: hrm.. I'm interested but I'm under the gun for a few other things.. so I'll pass for now.
[18:00] <hazmat> SpamapS, fair enough
[18:00] <hazmat> SpamapS, socket path has a default known address, it doesn't need to be set
[18:00] <hazmat> SpamapS, the problem is the other side isn't really listening outside of a hook execution atm
[18:23] <SpamapS> hazmat: I think you misunderstood the desire to have juju do.
[18:24] <SpamapS> hazmat: From what I see, its so you can make it easy to interact with juju outside hooks.
[18:24] <SpamapS> but I digress
[18:24] <SpamapS> I think its two different things
[18:24] <SpamapS> and should be discussed as such
[18:25] <hazmat> SpamapS, it is two different things, using the hook cli api outside of hooks is the one we need to support
[18:26] <hazmat> having juju client support for executing arbitrary hooks.. is not
[18:26] <SpamapS> its not so much the client on your remove machine though..
[18:26] <SpamapS> right?
[18:26] <SpamapS> Its just so that while you're ssh'd into the machine, you can use relation-get or call a hook
[18:27] <SpamapS> anyway, I don't want juju do to delay the introduction of all the other goodness
[18:27] <hazmat> SpamapS, 'juju' is the client.. regardless of where it is..
[18:27] <SpamapS> because the other stuff.. being able to relation-list one relation from another.. is *critical*
[18:28] <hazmat> SpamapS, agreed the entire hook cli api offered by hooks should be executable easily outside of hooks
[18:29] <hazmat> SpamapS, what's a use case for executing a hook directly?
[18:32] <SpamapS> hazmat: re-asserting configuration
[18:32] <SpamapS> hazmat: admin logs in, screws everything up debugging.. wants to leave it back in the state the charm asserts
[18:33] <hazmat> SpamapS, the problem is not every hook is idempotent.. calling something like -depart could mess things up more
[18:37] <SpamapS> hazmat: I don't have a use case for departed
[18:39] <hazmat> SpamapS, if we enable the capability.. we enable it for everyone.. a user can just execute config-changed directly at that point
[18:39] <SpamapS> *that* would be fantastic
[18:40] <hazmat> jimbaker, SpamapS g+ invites out
[18:40] <jimbaker> ok
[19:09] <jcastro> m_3: Hey, how's the node charm these days?
[19:09] <jcastro> https://github.com/thedjpetersen/subway#readme
[19:09] <jcastro> saw this on HN and I wanted to chase them down, it looks cute
[19:12] <SpamapS> jcastro: that does look cool
[19:13] <jcastro> especially since it uses node and mongo
[19:13] <jcastro> that's 2 other charms it can interact with
[19:13] <jcastro> SpamapS: you have any idea how complete the node charm is?
[19:14] <SpamapS> jcastro: Its a framework, like django and rails.. needs subordinates to be really amazing.
[19:15] <jcastro> ah dude I get it
[19:15] <jcastro> i'd need 3 instances for IRC
[19:16] <jcastro> bootstrap, subway, and mongo right?
[19:17] <SpamapS> jcastro: yes and no
[19:17] <SpamapS> jcastro: Since this is almost certainly a "low scale" app, you can just have it install mongo locally.
[19:17] <SpamapS> jcastro: I was doing the same thing w/ mediagoblin before they decided to rip out mongodb ;)
[19:17] <jcastro> ah
[19:18] <jcastro> ok I'm going to go just join their channel
[19:18] <jcastro> they seem pretty cool
[19:18] <SpamapS> jcastro: when are we going to start working on project fido again? Next week?
[19:20] <jcastro> yeah
[19:20] <jcastro> we totally forgot to plan for Strata running into this week
[19:30] <adam_g> hmm is there some hard limit on the # of settings that may be passed via relation-set in a single hook?
[19:31] <adam_g> cant seem to get anymore than 7 of 10 to the other side
[19:35] <SpamapS> adam_g: shouldn't be
[19:36] <SpamapS> adam_g: are the values large?
[19:36] <SpamapS> adam_g: zk does have a 1MB limit on any single node, and I believe relation-set sticks all of them in a single node.
[19:36] <SpamapS> But I'd think you would have something much more horrible if you hit that.
[19:36] <SpamapS> dunno if it truncates or errors
[19:37] <adam_g> SpamapS: actually, nevermind. verifiedby flooding the relation with a thousand settings before i realized i fat fingered something in the charm :{
[19:37] <SpamapS> adam_g: lol
[19:38]  * SpamapS sends adam_g's fingers to The Biggest Loser to get yelled at
[19:38] <adam_g> fat finger summer camp
[19:42] <m_3> jcastro: the node charm is actually in good shape
[19:42] <jcastro> good to hear
[19:42] <m_3> it's rpetty much ready for review... had to revamp it a couple of weeks ago for the mongodb talk
[21:09] <jcastro> SpamapS: pastebin/lodgeit charm is ready for round #2 of review. Fight!
[21:14] <flaviamissi> Hey! Is there a way to execute a command in an unit with `juju ssh`? I've tried the ssh's approach of passing the command in the end, but juju doens't interprets it the way I expected..
[21:47] <robbiew> SpamapS: m_3: -> http://ubuntuone.com/3fNyLF45rp016mWud0KMBc
[21:47] <robbiew> juju-jitsu!!
[21:47]  * robbiew returns to work now
[21:50] <SpamapS> robbiew: nice, is that hong kong phooey?!
[21:52] <robbiew> yup
[22:14] <m_3> robbiew: awesome!
[22:14] <m_3> I'm sitting in a talk entitled "Data Jujitsu"
[22:14]  * m_3 harumph!
[23:17] <_mup_> Bug #855989 was filed: Document usage with orchestra <juju:Confirmed> < https://launchpad.net/bugs/855989 >