/srv/irclogs.ubuntu.com/2013/04/09/#juju.txt

=== defunctzombie is now known as defunctzombie_zz
=== defunctzombie_zz is now known as defunctzombie
=== defunctzombie is now known as defunctzombie_zz
=== defunctzombie_zz is now known as defunctzombie
=== defunctzombie is now known as defunctzombie_zz
jcastromramm: is there a release of core between now and 13.04?14:37
mramm2yes14:38
mramm2later this week I expect another one14:38
mramm2and possibly one early next week14:38
m_3jcastro: jokes go here14:56
jcastrohey so14:57
jcastroI panged bruno and patrick to clean up the django readme14:57
jcastrosince it's like a template right now14:57
jcastroI was thinking for release week we'd have 2 big blog posts14:57
jcastroone for core14:57
jcastroone for charms14:57
m_3ack14:58
jcastroto go along with newdocs and better juju AU questions14:58
m_3marcoceppi: jenkins-slave w/ juju sub works great14:59
marcoceppiexcellent15:00
m_3I'll just now see what direction for the external relation makes sense15:00
m_3i.e., config on the slave or the master15:00
m_3prolly elastic-ip on the slave and then config the master15:00
m_3frickin cool15:00
jcastrodon't forget to post your thoughts on rick's metadata thing on the list fellas15:01
m_3the list is so boring15:01
jcastroDeal with It(tm) :)15:01
m_3:)15:01
marcoceppiB)15:01
robbiewlol15:06
=== defunctzombie_zz is now known as defunctzombie
m_3marcoceppi: damn... the auth model for jenkins-master/slave really requires them to be local15:17
m_3at least as far as I can see15:17
m_3:(15:17
marcoceppim_3: You mean local:charm or master slave on the same machine?15:17
m_3marcoceppi: same subnet / juju environment15:17
marcoceppiUgh, what?15:18
marcoceppithat's pretty lame15:18
m_3yeah15:18
m_3well at least the way the charm's written now15:18
m_3can maybe front it with apache2 configured for auth'd reverse-proxy15:18
m_3actually, nm... that would probably work15:18
m_3just won't be as lightweight as we discussed before15:19
=== salgado is now known as salgado-lunch
m_3but hey, it's juju15:19
m_3so that doesnt' really matter as much15:19
m_3marcoceppi: so I'll maybe extend jenkins so that the master has config to: a.) accept creds to b.) point to an "external" slave15:21
m_3anyways, I'll run that by ben15:22
=== salgado-lunch is now known as salgado
marcoceppiI've got a relation, that essentially provides an API over http. I should probably just make it consume a http interface, because that's all it does15:38
marcoceppiBut during graph testing it'll fail, because not just any http interface will do15:38
marcoceppiShould this become it's own interface name even though it's basically just a port + http address?15:38
=== defunctzombie is now known as defunctzombie_zz
jcastrohttps://juju.ubuntu.com/community/weekly-charm-meeting/15:44
jcastrothoughts?15:44
m_3sure man15:45
marcoceppi+115:45
imbrandonmarcoceppi: touch call, but i would say yea15:45
m_3maybe we can visit charm quality and audits across the store15:45
imbrandonbecause like port 80 a lot of services run on it15:45
imbrandonthat are considered diff15:45
marcoceppiimbrandon:  true, the interface defines the service, not so much the protocol which the service is using15:46
m_3marcoceppi: I'd say leave it http... then it can be wired behind proxies and the like15:46
imbrandonright, i mean its a fine line, but thats how i would see it15:46
m_3the fact that it's an api is just like saying it's special because of its content15:46
imbrandonhrm good point … hum15:47
m_3we'll need to exclude relations from some automated testing... jsut make sure they're caught in $CHARM_DIR/tests15:47
marcoceppim_3: What I'm worried about is on the flipside, the charm requires: api: http15:47
marcoceppim_3: gotchya, I think http is one of those too broad to test ones15:47
m_3that's fine imo15:47
m_3b/c that api server might be balanced, proxied, etc etc15:47
m_3external perhaps over time15:48
imbrandonm_3: btw i tried setting up a slave jenkins and charm runner ( running on OpenShift machines ) but could never get it to quite kick off fully automatic15:48
imbrandonmight bug ya about that sometimes $soon15:49
m_3imbrandon: hmmm15:51
m_3imbrandon: sure.. lemme see if I can get a standalone slave hooked up to an external master here15:52
imbrandonits at ci-bluecluster.rhcloud.com:8080 , 3 node jenkins setup just for this … but i got it offline for the moment ( was working on some update to WordPress on my own site this morning )15:52
imbrandoni can give you full root to them in a few , if it helps15:52
imbrandonits one master and 2 builders + can run smallish jobs on the master15:53
m_3imbrandon: ack15:55
m_3no need atm... I'm gonna run to shower before dinner15:55
imbrandongot a set of vCloud machines to play with too ( 6 i think ) unmeetered that i can use to add more builders if we can parallelize(sp?) it like that15:56
imbrandonnp, yea i got things to finish up for the next few hours but i'll likely be doing that this evening local time15:56
imbrandonno if we only had a vCloud provider … :)15:57
imbrandons/no/now15:57
jcastromarcoceppi: can you unpromulgate alice-irc? it's dead upstream18:40
jcastrohazmat: hey what's the ballpark time for jujujcharms.com to render an updated README?18:49
hazmatjcastro, 30m max o u t18:49
hazmatmarcoceppi, there's another branch of charmrunner18:50
hazmatwith notion of  building out juju versions for each test18:50
hazmatmarcoceppi, imbrandon, m_3 i think we need a mass change to charms to populate initial tests with rels18:50
marcoceppijcastro: unpromulgated18:54
marcoceppihazmat: which branch is that?18:54
jcastromarcoceppi: ta18:54
hazmatlooking around for it18:54
jcastromarcoceppi: also if you want to rename promulgate/unpromulgate to something more sane while SpamapS isn't looking ...18:55
marcoceppijcastro: duh, juju publish ;)18:55
jcastroI was just looking for a way to mess with clint18:55
hazmatmarcoceppi, ~hazmat/charmrunner/wip18:59
hazmatits got a little bit of cleanup, and some tests, but i realized that doing graph stuff at all is kinda of bogus long term, we should just seed some test scripts19:01
hazmatdigitalocean btw is awesome19:01
marcoceppihazmat: I'll take a look and merge the stuff I have done over the week19:01
hazmatmarcoceppi, what's your focus for it atm?19:02
hazmatgetting the server going?19:02
marcoceppihazmat: generating graph tests externally, making it a subordinate, streamlining and separating graph test from unit testing19:03
marcoceppigiving more granular test statuses19:04
hazmatmarcoceppi, there's a useful script in the branch for fetching all the charms into a nice directory structure that includes ownership (ie not a straight charm repo)19:04
hazmatand mods to the planner to generate plans from that structure19:04
hazmatcool19:05
marcoceppiarg, just realized I've been talking about charm tester not charmrinnwr19:06
marcoceppiyour changes will be great for integrating with charm tester19:07
hazmatah.. makes more sense now19:18
SpamapSpromulgate is the right word. Just because you guys don't want plain english.. ;)19:28
marcoceppiSpamapS: jcastroI think we shouldmake it promulgar19:28
marcoceppiCould just do "charm yo promulgo"19:29
SpamapSasi es mas internacional19:31
ahasenackhi guys, I just tried two charms from the charm store and both don't deploy, I guess it's my lucky day20:31
ahasenackhttps://bugs.launchpad.net/charms/+source/etherpad-lite/+bug/116699720:31
_mup_Bug #1166997: etherpad-lite doesn't deploy <etherpad-lite (Juju Charms Collection):New> < https://launchpad.net/bugs/1166997 >20:31
ahasenackhttps://bugs.launchpad.net/charms/+source/subway/+bug/116698920:31
_mup_Bug #1166989: Subway charm doesn't deploy on precise <subway (Juju Charms Collection):New> < https://launchpad.net/bugs/1166989 >20:31
ahasenackI tried on lxc, but the error is unrelated20:31
=== gary_poster is now known as gary_poster|away
ahasenackhm, I may have messed up, default-series is precise in my environments.yaml file, but I'm running quantal20:34
hazmatsidnei, re canaries and upgrade.. alternatives environment snapshots20:36
=== gary_poster|away is now known as gary_poster
=== markramm is now known as mramm_
paragladeis it possible to set values of keys defined in the config.xml during a hook execution?  I know how to get the values just not sure how to set them.21:39
marcoceppiparaglade: You can't change the charm's configuration (or at least the values of a config-get) from within the charm itself21:40
paragladehmm ok.  marcoceppi so I am trying to build a zabbix charm, the agent (a subordinate service) can be setup to "talk" to a server via a config set or with a relationship to a server or proxy.  I am kinda struggling on trying to implement this.  I looked at other charms but nothing really seems to model this setup exactly.21:46
marcoceppiparaglade: Zabbix! Very exciting. What it sounds like you're tying to do is get the "master" server which can either be a relation or a configuration value, correct?21:47
paragladeyes thats correct21:48
marcoceppiCool, there are a few ways to tackle this. You'll have to forgive me, I ususally use Nagios so I'm not sure the inner-workings of Zabbix, but if you can only designate one "master" then I would just choose a pecking order of which is more important. User defined configuration or relation data.21:50
marcoceppiIn my opinion configuration data takes precendence over relation data. So I would have the relation hook check if the configuration value exists. If it doesn't then use the data from the relation, otherwise just exit silently.21:51
marcoceppiIt gets a bit more complicated if you can have multiple "master" monitoring services. In that case I would create a file in the charms root that gets updated by the hooks respectively. So it would just add or subtract values depending on if configuration was changed or if a relation was changed21:52
paragladeok thanks.  I think the configuration over relation data will actually work well with this charm(s).21:55
marcoceppiparaglade: As long as you clearly document that in the README (and unless someone else corrects me) that should be sufficient21:56
paraglademarcoceppi: if you don’t mind I think I have another question that maybe you can give some wisdom on.  I was thinking when a *-relation-changed *-relation-join event fired I would just call config-changed hook in those event hook scripts.  The idea is that I would only need to create the config implation in one script file.  However it seems that I am able to get the relation data just the config data22:11
paragladewhen I do this.22:11
paragladeany suggestions on a22:11
paragladeclean way to handle this?22:11
marcoceppiThat sounds like a great idea! I know a few charms that do this already. The problem with relation data is the environment variables that exist in order for the relation-get to work are only loaded when the relation is actioned upon. So you can't just call relation-get in any other hooks without supplying which _exact_ relation you're referring to. This gets messy when trying to track what relation_id goes where. What I usually recommend is to22:13
marcoceppijust persist relation data to files and have other hooks look in those files for their respective values22:13
marcoceppiHere's a small example:22:13
marcoceppihttps://gist.github.com/marcoceppi/534984222:14
marcoceppiIn this case you can just replace update_configuration with executing hooks/config-changed22:14
AskUbuntuDoes my charm need both a relation-changed and relation-joined hook? | http://askubuntu.com/q/27970622:19
sarnoldmarcoceppi: these are cool snippets, I'd hate to lose them.. are they documented anywhere else? or implemented anywhere? :)22:27
marcoceppisarnold: I was actually self documenting paraglade's question on to Ask Ubuntu to help keep them alive22:28
marcoceppiHopefully help fill up Google's juju/charming questions with better information22:28
sarnoldmarcoceppi: oh! excellent :)22:28

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!