[11:44] <AskUbuntu> how to deploy charm to a machine which is already installed with some other charm? | http://askubuntu.com/q/321479
[13:55] <jcastro> marcoceppi: can you link me to that pad with the build instructions again?
[13:55] <jcastro> last time I promise
[13:56] <melmoth> jcastro not sure this is what you are looking for http://pad.ubuntu.com/rzudQRvsmw
[13:56] <melmoth> but it  s in my client historry :)
[13:58] <marcoceppi> http://pad.ubuntu.com/rzudQRvsmw ^ that's it
[13:58] <marcoceppi> just added update-alternatives instructions too
[13:58] <jcastro> melmoth: yep, thanks
[13:59] <jcastro> marcoceppi: so like these instructions remind me
[13:59] <jcastro> that we should have dailies
[13:59]  * jcastro knows what to ask mramm for next
[14:02] <kirkland> marcoceppi: jcastro: Okay, I got my charm deploying successfully last night, now I'm just trying to get it scaling properly
[14:02] <kirkland> marcoceppi: jcastro: how can I tell, inside of a unit, how many total similar units are in the cluster?
[14:02] <kirkland> marcoceppi: jcastro: I need to update a configuration file, with the total number of units, and which number unit each of them are
[14:02] <jcastro> from inside an existing unity?
[14:03] <kirkland> jcastro: :-)  "unity"
[14:03] <marcoceppi> kirkland: You'll want a peer relations
[14:03] <jcastro> yeah sorry
[14:03] <kirkland> marcoceppi: oh, interesting
[14:03]  * marcoceppi digs up docs
[14:03] <kirkland> juju just became a swingers party
[14:04] <jcastro> marcoceppi: I need mongodb for local to work right?
[14:04] <jcastro> is it "mongodb" or "mongo-server"?
[14:04] <marcoceppi> jcastro: yeah, install mongodb then run sudo service stop mongodb
[14:05] <marcoceppi> jcastro: err, mongo-server
[14:06] <marcoceppi> I must have already had that installed
[14:06] <melmoth> kirkland line 60 http://bazaar.launchpad.net/~pierre-amadio/charms/precise/ntpmaster/trunk/view/head:/hooks/ntpmaster_hooks.py
[14:07] <melmoth> thats how i list peer in relation
[14:07] <melmoth> may be there s a better way
[14:08] <marcoceppi> So, we're missing that in the docs evilnickveitch we don't mention peer relations at all. You still working on charm-author doc?
[14:08] <jcastro> marcoceppi: and mongo needs to be stopped?
[14:08] <jcastro> I am getting
[14:08] <jcastro> juju open.go:89 state: connection failed, will retry: dial tcp 127.0.0.1:37017: connection refused
[14:09] <marcoceppi> jcastro: do you have the latest juju-core?
[14:09] <jcastro> over an over on bootstrap
[14:09] <jcastro> yeah I did the update and it fetched stuff
[14:09] <marcoceppi> jcastro: oh, it hasn't failed yet?
[14:09] <marcoceppi> wait for it to fail or succeed
[14:09] <jcastro> oh ok
[14:09] <marcoceppi> juju bootstrap -v is crazy verbose
[14:09] <jcastro> I probably cancelled too early
[14:09] <marcoceppi> it's running juju-*-local upstart which starts a mongodb instance
[14:10] <jcastro> error: no reachable servers
[14:10] <marcoceppi> jcastro: yeah...
[14:10] <marcoceppi> I thought you had an ssd?
[14:10] <jcastro> I do
[14:11] <marcoceppi> destroy environment then try again
[14:11] <marcoceppi> kirkland: the peer relation is defined like provides and requires, but it's defined as peer and all units automatically get the relation
[14:12] <marcoceppi> So if you just need to get a list of all the peers you can run get a list of the units with `relation-list` which will print each of the units in a service deployment
[14:13] <marcoceppi> it's strucutre is just like any other relation, so if you needed to get the address of each unit in the peer relation you could write a relation-joined hook that looked like this
[14:14] <marcoceppi> http://paste.ubuntu.com/5887679/
[14:15] <marcoceppi> but with actually syntatically correct bash: http://paste.ubuntu.com/5887680/
[14:15] <jcastro> what version do you have of goju checked out?
[14:15] <jcastro> 1.11.3-saucy-amd64
[14:16] <marcoceppi> that's the version I have compiled, what does bzr revno $GOPATH/src/launchpad.net/juju-core/
[14:18] <jcastro> 1481
[14:21] <marcoceppi> jcastro: what does the line `timeout :=` in $GOPATH/src/launchpad.net/juju-core/environs/local/environ.go (~L539) say?
[14:22] <jcastro> marcoceppi: 10
[14:22] <jcastro> maybe I should blow it away and reDL
[14:22] <jcastro> you changed that to 60 no?
[14:22] <marcoceppi> jcastro: I'm looking at the rev history
[14:22] <marcoceppi> I don't think my timeout fix landed
[14:23] <marcoceppi> even though it says it was merged
[14:23] <marcoceppi> jcastro: blow away and try again just to make sure
[14:28] <kirkland> marcoceppi: hmm, okay;  is there an example metadata.yaml?
[14:28] <kirkland> marcoceppi: so that I get the syntax right?
[14:28] <marcoceppi> kirkland: https://bazaar.launchpad.net/~charmers/charms/precise/wordpress/trunk/view/head:/metadata.yaml
[14:30] <jcastro> marcoceppi: yeah that did it
[14:30] <marcoceppi> jcastro: cool, I can't even get it to compile anymore :(
[14:31] <kirkland> marcoceppi: okay, thanks;  I guess I just want the interface to be ssh
[14:32] <marcoceppi> kirkland: well, the interface can be anything you want
[14:32] <kirkland> marcoceppi: http://paste.ubuntu.com/5887726/
[14:32] <kirkland> marcoceppi: does that look right?
[14:34] <marcoceppi> kirkland: that looks right to me, I'd be weary of interface naming, it doesn't exactly matter, per se, with peer, but each interface is like an agreement for what data will be sent recv
[14:34] <marcoceppi> So I typically look to see if an interface has been created already (there is an SSH interface) and see what the charms expect.
[14:34] <kirkland> marcoceppi: can you suggest something better?
[14:35] <marcoceppi> Since any charm matching that ssh interface can *technically* be connected to each other
[14:36] <marcoceppi> kirkland: I really can't, I'm not up to snuff with john the ripper, it looks like it's correct. I was more just saying "so you know"
[14:36] <marcoceppi> finally, all relations are optional, so explicitly stating that does nothing
[14:37] <kirkland> marcoceppi: I need all units to be able to ssh to one another
[14:37] <marcoceppi> and setting optional to false, probably does nothing
[14:37] <kirkland> marcoceppi: would "john-ssh" be better?
[14:37] <kirkland> thanks
[14:37] <marcoceppi> kirkland: So you're probably going to be sending host, port, and a shared private key between units
[14:38] <marcoceppi> interface name for peer really doesn't matter, was my point, you can name it whatever you want because no other units except for those inside of the service, can connect
[14:40] <kirkland> marcoceppi: does this make more sense?  http://paste.ubuntu.com/5887750/
[14:41] <marcoceppi> kirkland: yeah
[14:43] <kirkland> marcoceppi: okay, now, back to my original question...   how does a unit count how many peers he has?
[14:43] <marcoceppi> kirkland: relation-list will list all units attached to that unit in that relation
[14:43] <marcoceppi> kirkland: http://paste.ubuntu.com/5887680/
[14:44] <kirkland> marcoceppi: okay, and does *every* unit get triggered each time a unit (relation) is added?
[14:44] <marcoceppi> Just needs to be run within one of the relation-* hooks, can't just be run arbitrarily
[14:44] <marcoceppi> kirkland: everytime a peer is brought up, it will fire a john-relation-joined on every john unit deployed
[14:45] <kirkland> marcoceppi: okay, cool, let me hack on this a bit then
[14:46] <jamespage> wedgwood, adam_g: apologies for my completely fudged landing of roaksoax change for one of the openstack contrib helpers earlier today
[14:52] <evilnickveitch> marcoceppi, yes, there is a lot to rewrite. if someone wants to make some notes on peer relations, be my guest
[14:56] <marcoceppi> jcastro: having weird build errors over here, blew away my copy of 1.11.3.1 too soon
[15:05] <jcastro> marcoceppi: how long was local on "Pending" for you?
[15:05] <marcoceppi> jcastro: do a ps and look for wget
[15:06] <jcastro> hmm nope, must be something else
[15:06] <jcastro> I'm going to do some juju website mainteance, will get back to this shortly.
[15:09] <marcoceppi> likewise
[15:25] <kirkland> marcoceppi: okay, next question...  I have juju-ssh'd to a unit, and I want to run relation-list on a command line, so that I can debug the output
[15:25] <kirkland> marcoceppi: I suspect I need to source some environment or something?
[15:25] <marcoceppi> kirkland: you kind of can't
[15:25] <kirkland> marcoceppi: ?  that seems unkind
[15:26] <marcoceppi> well, you can, but it's going to require quite a few things. You're better off putting juju-log `relation-list` in to your hook so you can just tail the log instead
[15:26] <marcoceppi> kirkland: there used to be a pyjuju command juju debug-hooks that would allow you to, but that's not been ported to juju-core yet
[15:26] <kirkland> marcoceppi: does that mean re-deploying?
[15:26] <kirkland> marcoceppi: or can I manually hack it into /var/lib/juju/somewhere
[15:26] <kirkland> marcoceppi: yeah, I've used juju debug-hooks before
[15:27] <marcoceppi> kirkland: well, you could edit the charm live and then add a unit and watch the output (what I sometimes do during development)
[15:27] <marcoceppi> one sec
[15:27] <marcoceppi> /var/lib/juju/agents/unit-*/charm
[15:52] <marcoceppi> jcastro: http://imgur.com/a/L6QJk
[16:59] <ahasenack> charmers! :)
[16:59] <ahasenack> hi, can someone review https://code.launchpad.net/~stub/charms/precise/postgresql/bug-1187508-allowed-hosts/+merge/174771 please?
[16:59] <ahasenack> 52 lines of diff
[17:00] <ahasenack> just 13 new lines
[19:22] <hazmat> jcastro, re pending local, about 3-5m on ssd and fios
[21:24] <thumper> marcoceppi, jcastro: how is your local provider playing going?
[21:25] <marcoceppi> thumper: got juju-gui running on it, have a bug to report if I can reproduce it though (destroy-environment does weird things now)
[21:25] <thumper> weird things in what way?
[21:26] <marcoceppi> thumper: it complains that it can't delete the /var/lxc/auto files
[21:26] <marcoceppi> because it seems to delete it before it checks for it
[21:26] <marcoceppi> I had to touch them all before it let me actually destroy environment
[21:26]  * marcoceppi bootstraps again to verify it wasn't a one time thing
[21:26] <thumper> marcoceppi: let me try here...
[21:29]  * thumper twiddles thumbs waiting for things to spin up
[21:30] <thumper> marcoceppi: I'm wondering if you have a larger test deployment that I could try
[21:30] <thumper> wondering how much stress it can handle
[21:31] <marcoceppi> thumper: I had 6 machines running earlier without much issue
[21:31] <marcoceppi> but it's an i7 and 32G of ram
[21:31] <thumper> destroy machine seemed to work without issue
[21:32]  * thumper waits for the wordpress agent to start
[21:32] <thumper> before killing it
[21:33] <marcoceppi> thumper: here we go
[21:33] <marcoceppi> http://paste.ubuntu.com/5888895/
[21:33] <thumper> yep I get an error too
[21:34] <thumper> I wonder...
[21:34] <marcoceppi> thumper: http://paste.ubuntu.com/5888896/
[21:34] <marcoceppi> It's like it deletes it, then checks if it exists
[21:34] <marcoceppi> pretty minor, keep running destroy-environment and it clears out eventually
[21:34] <thumper> dude, that isn't minor
[21:35] <thumper> that is freaking annoying
[21:35] <thumper> I'll fix-it right now
[21:35] <marcoceppi> well, relatively speaking
[21:35] <thumper> I think it was an assumption on my part
[21:35] <thumper> it appears that lxc-destroy removes the auto restart symlink for us
[21:35] <thumper> and we are trying to remove it too
[21:35] <marcoceppi> I haven't tested reboot survive-ability
[21:35] <marcoceppi> but I'll probably give that ago later tonight
[21:36]  * thumper tests manually
[21:36] <marcoceppi> Otherwise I haven't found anything else with local provider just yet
[21:36] <marcoceppi> outside what we talked about last night
[21:37]  * thumper nods
[21:37] <thumper> that is it
[21:37] <thumper> lxc-destroy removes the link
[21:37]  * thumper submits fix
[21:38]  * marcoceppi trembels at the thought of having to recompile gojuju
[21:38] <thumper> marcoceppi: what is wrong with "go install ./..."
[21:38] <thumper> ?
[21:38] <marcoceppi> thumper: I've just not had a lot of luck with it since I started using it yesterday
[21:46] <thumper> marcoceppi: now to get someone to review it...
[21:48] <marcoceppi> thumper: So I've been using `go get -v launchpad.net/juju-core/..." should I be doing something different?
[21:48] <thumper> marcoceppi: hmm...
[21:48] <thumper> that works the first time