/srv/irclogs.ubuntu.com/2014/03/19/#juju.txt

josehey guys, any known way to run 1.18 on saucy?00:19
=== thumper-afk is now known as thumper
=== CyberJacob is now known as CyberJacob|Away
tvansteenburghis it legit to call `juju set` from within a charm hook?02:02
tvansteenburghfor example, if i wanted to change on a config value on the "other" service when a relation is joined02:03
tvansteenburghchange a config value02:04
marcoceppitvansteenburgh: you can't actually do that. Charms can't set configruation for itself or other services02:14
marcoceppitvansteenburgh: if you need to commuicate from one charm to another in a deployed environment you should use relation-set and relation-get02:14
marcoceppijose: 1.18 isn't released yet02:14
joseoh02:14
josemarcoceppi: mind a quick PM?02:15
tvansteenburghmarcoceppi: thanks02:15
marcoceppi1.17.5 is latest devel, 1.16.6 is latest stable02:15
marcoceppijose: sure02:15
josemarcoceppi: hey, do you know approximately how much time does the mysql charm take to set the variables (database, user, password...)?03:24
marcoceppijose: not very long, shouldn't take more than a few seconds. If you're thinking about putting a sleep in your script to wait for variables then you're doing something wrong03:34
josemarcoceppi: that's just what I did03:35
marcoceppidon't03:35
joseI did `while [ -z "$DB ]; do sleep 10; done`03:35
marcoceppiyou won't get them, because relation-get variables are locked at the start of the run of the hook03:35
josehmm, how should I set it, then?03:35
marcoceppithey're only updated when a hook is done executing03:35
marcoceppiuse relation-changed hook03:36
joseinstead of a relation-joined?03:36
marcoceppiright, relation-joined is basically anything you need to do pre-relation03:36
marcoceppiit's like an initial handshake03:36
josecool!03:36
marcoceppiif you're just waiting for data03:36
jose(yeah, you read my mind)03:36
marcoceppiput your code in relation-changed, and add an idempotency guard03:36
josea... what?03:37
marcoceppilike, db=`relationg-get database`; if [ -z "$db" ]; then exit 0; fi03:37
joseoh, gotcha03:37
marcoceppithat way, if the relation settings haven't been set the hook will just exit03:37
marcoceppiand when relation data does change, the hook will fire again03:37
josebtw, let me say this as I *just* did my first juju (unsuccessful) deploy on EC2: it's AMAZING.03:38
marcoceppi:D03:39
bodie_:}03:40
joseif this does good, then you should be have another charm on the queue03:41
=== vladk|offline is now known as vladk
josemarcoceppi: still around? I'm kinda stuck with a hook03:57
=== timrc is now known as timrc-afk
=== vladk is now known as vladk|offline
=== CyberJacob|Away is now known as CyberJacob
=== vladk|offline is now known as vladk
=== CyberJacob is now known as CyberJacob|Away
=== Ursinha is now known as Ursinha-afk
=== Ursinha-afk is now known as Ursinha
jamespagemarcoceppi, is amulet still pip installing cherrypy?11:02
marcoceppijamespage: bleh, yes11:03
marcoceppijamespage: next patch will address that11:04
jamespagemarcoceppi, because its not avaliable on 12.04 right?11:04
marcoceppiright11:04
marcoceppiwell, py3 cherrypy isn't at least11:04
jamespagemarcoceppi, please - I really like the work mbrusek has done on rabbitmq-server for amulet tests11:04
jamespagebut I can't run that on our QA cloud with pip installs....11:05
marcoceppiright, jamespage i have a 1.4.1 patch release planned for friday, I'll make the change then11:05
jamespagemarcoceppi, thanks - are you dropping the requirement for it?11:05
marcoceppijacekn: going to use a web framework that's in archive11:06
marcoceppijamespage: ^^11:06
jamespagemarcoceppi, +111:06
jamespagemarcoceppi, how does amulet choose which series to test on btw?11:09
marcoceppijamespage: the test set it (considering the fact that charms are currently tied to series) that might change where series is set by the test runner though11:10
marcoceppino one has had a use case for it yet11:10
marcoceppijamespage: when you create a new amulet.Deployment you can set series as an option, defaulting to precise11:11
jamespagemarcoceppi, I do - cause althought the branch for the charm is 'precise'11:11
jamespageit actually works on 12.04->14.0411:11
jamespagemarcoceppi, OK - cool11:11
jamespagemarcoceppi, I can see how I can make that work - thanks!11:12
marcoceppijamespage: can you file a bug for amulet about muti-series testing. I'll target it to 1.511:12
marcoceppiwe'll probably need a way to override via env variable, mobile atm so cant file it myself11:13
jamvila: just checking in with you that things seem happier on HP for you today11:25
hazmatjose, ping.. resolved?11:32
vilajam: thanks for checking ! Things stabilized enough yesterday for me to be able to deploy and use the deployment.11:32
jamgreat11:32
vilajam: being able to use 1.17.5 without uploading it was key. I'm still talking with hp about the upload issue I face11:33
hazmatmarcoceppi, why does amulet need a web app? re cherrypy11:33
hazmatmarcoceppi, oh.. ic.. xmlrpc server is builtin and roughly the same11:33
josehazmat: hey, not exactly... if you have a min I have the links around12:05
hazmatjono, sure12:19
hazmater. jose sure12:19
josehazmat: so, I deploy https://code.launchpad.net/~jose/charms/precise/phpbb/trunk and install, config-changed and start run good, but when a DB starts a relation with it I get an error when running db-relation-changed, says something about unbound variable on line 1812:20
lazyPowerif anyone has time can i get a +1 on https://code.launchpad.net/~lazypower/charms/precise/nrpe/fix-lp-1287393/+merge/209137 ? I need to land this as a dependency on some other stuff in the queue12:23
joseoh, morning, lazyPower!12:24
lazyPowerjose: if a variable is reported as unbound, you're referencing it before it's set12:24
hazmatjose. line 18 you've got an extra > at the end afaics12:25
lazyPowerand good morning :) o/12:25
josehmm, let's change that12:25
hazmatjose,  re http://bazaar.launchpad.net/~jose/charms/precise/phpbb/trunk/view/head:/hooks/db-relation-changed#L1812:25
josehazmat: looking at it now! I'm just wondering, does relation-get host with the mysql interface provides the host address? or what's the parameter set?12:26
hazmatjose host does work for mysql interface12:28
hazmatthe mysql server side  sets database, username, password, host atomically12:28
josethen I wonder what's the variable not set?12:29
* jose deploys and does debug-hook12:30
jamespagemarcoceppi, hey - when I wrote the percona-cluster charm I decided to default all databases to UTF812:37
jamespagemarcoceppi, any reason why we should not do the same in mysql?12:37
jamespagemarcoceppi, I was so busy testing pxc with openstack for icehouse b3 I missed https://bugs.launchpad.net/glance/+bug/127900012:38
_mup_Bug #1279000: db migrate script to set charset=utf8 for all tables <glance> <Glance:In Progress by lzy-dev> <oslo:In Progress by lzy-dev> <glance (Ubuntu):Triaged> <glance (Ubuntu Trusty):Triaged> <https://launchpad.net/bugs/1279000>12:38
=== timrc-afk is now known as timrc
marcoceppijamespage: I don't see why not12:47
joseguys, I'm having a problem when SSH'ing into a machine... looks like juju is getting the private IP instead of the public IP so it never goes through13:05
jose(using EC2)13:05
lazyPowerthats strange, i've not seen that behavior13:09
lazyPowerjose: how are you attempting to access the unit?13:09
joselazyPower: juju ssh phpbb/013:09
joseand juju status is the one giving me that info (I'm checking my console atm)13:09
lazyPowerwhich release of juju are you running?13:10
jose.1613:10
lazyPowerhmm i dont have a .16 release to check on13:10
joseI'm going to try adding an entry to my hosts file13:10
josethat may solve it in the meanwhile13:10
lazyPowerjose: if you can reproduce it, make sure you file a bug on that.13:11
josewill do - it's not the first time I've seen this13:11
rbasakmgz: around? I think bug 1248283 is critical for MAAS+juju in Trusty, but it isn't milestoned to make Trusty currently.13:22
_mup_Bug #1248283: juju userdata should not restart networking <dbus> <deploy> <landscape> <maas-provider> <midway> <ubuntu-openstack> <juju-core:Triaged> <MAAS:Invalid> <dbus (Ubuntu):Invalid> <juju-core (Ubuntu):Triaged> <juju-core (Ubuntu Saucy):Triaged> <juju-core (Ubuntu Trusty):Triaged> <https://launchpad.net/bugs/1248283>13:22
jamespagefwereade, ^^13:23
mgzrbasak: looking13:29
mgzrbasak: I'll set back to 1.1813:31
mbruzekjamespage, Thanks for the review on the amulet tests, working on addressing the issues now.13:31
rbasakmgz: thanks!13:31
jamespagembruzek, did you see what I was poking at? we should be able to confirm the correct data is set on the relations13:32
jamespagembruzek, excitingly I have a full redux in flight for rabbitmq - want to land your tests first tho!13:32
mbruzekHonestly did not, but running them now to see what ceph is passing on the relation13:32
mgzI think our issue with that bug is we're actually not sure what to change on the juju-core side... or we have already poked it and our change didn't work, we tried moving when the bridge-utils package gets installed for... basically that reason?13:33
mgzbut the last comment looks hopeful13:33
mbruzekjamespage, What key should I be looking for in the ceph relation?  Is it actually called "key"?13:34
jamespagembruzek, look for 'key' and 'auth'13:34
mbruzekjamespage, Running the tests now to get that.13:35
jamespagembruzek, where does stdout go? I could not see the print statement anywhere13:35
mbruzekIt depends upon how you run them.  If you are running juju test 20_relations_test.py  you need the -v flag13:37
jamespageah - right I see13:37
mbruzekI run them differently so the systems are still up after the test is done.13:37
mbruzekjuju test  <test name> both bootstraps and destroys environment13:38
mbruzekjamespage, http://pastebin.ubuntu.com/7119833/13:42
jamespagembruzek, amqp stuff looks good - however the ceph cluster isnot bootstrapped so it won't send the keys13:43
mbruzekAhh I see now, thank you for pointing that out.13:44
lazyPowerdosaboy: are you around?13:44
dosaboylazyPower: otp but yes go ahead13:51
lazyPowerdosaboy: i noticed the MP we have reviewed is getting quite a bit of back and forth. touching base with you so we can find a middle fo the road resolution without the delay.13:52
bloodearnesthowdy. Is there a way for a charm to respond when a floating ip is added?13:55
bloodearnestlike, does it automatically fire any realtion-changed hooks with a public-address?13:55
lazyPowerbloodearnest: floating ip on which provider? elasticip from AWS?13:55
mbruzekbloodearnest, How would the charm know when that changes?13:55
lazyPowermbruzek: good point!13:55
bloodearnestmbruzek: indeed, I didn't think so (though that would be ideal from the charms pov, no?)13:56
bloodearnestlazyPower: openstack13:56
mbruzekIf something changes on the relation, a *-relation-changed gets fired yes, but what would trigger that13:56
bloodearnestlazyPower: but, interested in the general13:56
lazyPowerbloodearnest: since juju doesn't spend cycles monitoring the ip table of the machine, i dont think thats in the cards13:56
mbruzekbloodearnest, I supposed one could write a daemon that checks the ip, and if somthing changes trigger a relation-changed event or something of that nature13:57
bloodearnestmbruzek: right, and that would now include a public-address automatically? Or would you need to provide it explicitly?13:57
bloodearnestmbruzek: ack, was thinking along those lines13:57
bloodearnestor set the ip manually as a config (ugh)13:57
mbruzekbloodearnest, I would have to check, but I believe that the  private address is often part of the relation-changed information.13:57
lazyPower^13:58
mbruzekSo if you could somehow detect that the ip changed you could fire the relation changed juju event and it would get sent to the related charms.13:58
marcoceppibloodearnest: you could write an out of band check13:59
marcoceppibut there is no mechanism in juju that i know of13:59
bloodearnestmarcoceppi: yeah, I thought so. Just wanted to check if I was missing something.14:00
bloodearnestmaybe when juju grows a network abstraction, perhaps14:00
marcoceppibloodearnest: its a compellimg case for a unit-changed hook14:01
bloodearnestmarcoceppi: WANT14:01
bloodearnestthat could do volume changes too. nice.14:01
marcoceppibloodearnest: you shoild file a bug about ot on core. afaik juju has no way of tracking unit changes, inclidimg address, but given your use case it would make sense14:02
* marcoceppi is typing from a phone :\\14:02
dosaboylazyPower: sorry lots of calls :(14:03
=== marcoceppi is now known as marcoceppi-mobil
dosaboylazyPower: so can you remind me which MP this is?14:03
lazyPowerdosaboy: perfectly ok. ping me when you're off the phone?14:03
dosaboymysql per chance?14:03
lazyPowerindeed14:03
bloodearnestmarcoceppi-mobil: ack, will do14:03
bloodearnestis there a hack to trigger a hook context from a cron job in 1.16 (i.e. what juju-run does in 1.17)?14:04
marcoceppi-mobilbloodearnest: not without a lot of work14:05
bloodearnestmarcoceppi-mobil: right. 1.18 is due like RealSoonNow™, so maybe I can convince IS to use that in prod :)14:08
marcoceppi-mobilbloodearnest: that's whati would do14:08
zchanderHI, is anyone familiar with preseed settings for MaaS? I want to partition a disk while commissioning14:14
bloodearnestmarcoceppi-mobil: https://bugs.launchpad.net/juju-core/+bug/129467314:19
_mup_Bug #1294673: Support hook for handling unit changes (like public IP address) <juju-core:New> <https://launchpad.net/bugs/1294673>14:19
=== marcoceppi-mobil is now known as marcoceppi
=== marcoceppi is now known as marco-traveling
Fishy_juju ssh only goes to service, not machine?  So if I kill a service (but the machine is still up), no juju way to ssh in?15:19
lazyPowerFishy_: just give it the machine #15:23
lazyPowereg: to ssh to machine 1 - juju ssh 115:24
bbcmicrocomputerany idea when the OpenStack charms are going to appear for Trusty?15:25
marco-travelingbbcmicrocomputer: ask jamespage15:26
marco-travelingafaik they already work on trusty, just need to promulgate the as such15:27
bbcmicrocomputermarco-traveling: k, cool15:27
Fishy_juju ssh 1 Permission denied (publickey,password). ERROR exit status 25515:27
Fishy_(whereas juju ssh wordpress is fine)15:28
mgzFishy_: which provider?15:28
Fishy_local15:28
mgzlocal you can only use the units15:29
marco-travelingFishy_: is this local provider?15:29
mgzFishy_: bug 123349715:29
Fishy_blah ;)15:29
Fishy_that bug is about a password protected ssh key, mine is wide open15:31
Fishy_maybe same root cause though15:31
mgzyeah, should be another bug too somewhere15:32
=== cmagina-away is now known as cmagina
bloodearnestmarco-traveling: so, my bug is duplicate of https://bugs.launchpad.net/juju-core/+bug/121557915:42
_mup_Bug #1215579: Address changes should be propagated to relations <addressability> <juju-core:Triaged> <https://launchpad.net/bugs/1215579>15:42
jamespagebbcmicrocomputer, probably at or around the time trusty releases15:48
jamespagebbcmicrocomputer, if you want to test the branchs are all under ~openstack-charmers15:48
jamespagebbcmicrocomputer, /icehouse suffix15:49
bbcmicrocomputerjamespage: k, thanks15:50
jamespagebbcmicrocomputer, oh and just to be clear - the precise charms will be the trusty charms15:51
jamespageso don't look for 'trusty' in the branch path15:51
bbcmicrocomputerjamespage: k :)15:51
Fishy_whats it mean if I see 3 charms with the same name in the charm store, but none under reccomended.. and none work from a juju deploy command line?16:23
Fishy_some kind of beta charm that hasn't been approved yet?16:23
Fishy_i.e.   Location: cs:~canonical-ci/precise/gerrit-5916:24
Fishy_deploy did finally work, ignore me16:24
Fishy_need full path i guess16:24
rick_h_Fishy_: so charms can be customized for various users16:27
rick_h_Fishy_: so you'll often see copies of a charm that a user is working on or with16:27
rick_h_Fishy_: in this case, you found the charm used by ~canonical-ci16:28
Fishy_the source link didn't work for these though.. is there a normal procoedure to take one of these and grab the source?16:28
Fishy_so I can tweak for my own needs16:28
Fishy_https://bazaar.launchpad.net/~canonical-ci/charms/precise/gerrit/trunk/files16:29
Fishy_no worko16:29
rick_h_Fishy_: there's a thing in charm tools called 'charm get'16:29
rick_h_where you get give it a cs: url and get the downloaded charm16:29
Fishy_nice16:29
rick_h_Fishy_: oh hmm, that's not public. I wonder why that got pulled in16:30
Fishy_the entire charm isnt supposed to be or just the source url16:34
rick_h_Fishy_: I'm looking into it.16:34
rick_h_Fishy_: you've hit some edge case and I'll need to look into how it got this way16:34
rick_h_it's not typical16:34
Fishy_apparently I am good at edge cases with juju16:34
rick_h_stick around, we like finding those. :)16:35
Fishy_if I move my entie infastructure over to juju land, I will be here a few months...16:35
lazyPowerwe like the sounds of that Fishy_16:36
Fishy_next up is try to write my own charm for an internal app16:36
Fishy_if it's not too bad, then I just need to do it 50 more times16:36
lazyPowerFishy_: we also have a testing tool for validating your charms are doing what you expect them to16:37
lazyPowerAmulet to the rescue!16:37
Fishy_What testing tool?16:37
lazyPowerFishy_: https://juju.ubuntu.com/docs/tools-amulet.html16:37
lazyPowerif you need test examples, mbruzek wrote an excellent amulet test in the memcached charm16:38
lazyPowerhe's also got some solid work in rabbit-mq16:38
Fishy_nice16:38
Fishy_is it insane to try and use "Subordinate services" as kind of mix in behavior?16:39
Fishy_like say I have a home-rolled backup service I write...  then have 10 charms, all using my backup service16:39
lazyPowernot at all. I wrote a subordinate for papertrail-app (a PAAS log aggregator service)16:40
lazyPowerbecause I didn't want to pay/manage my own logging infrastructure16:40
lazyPowerI've used it in 3 production deployments thus far, managing a total of 52 hosts.16:41
bloodearnestFishy_: that's the idea, there is a backup subordinate already around somewhere, but not in the store yet I think16:41
lazyPoweri throw those metrics in there because it makes me feel special.16:41
Fishy_haha16:41
Fishy_just need to see how it works mixing in to stuff people already wrote.. I see how it would be easy for my own charms to all put data in a common place so a backup subordinate could grab it..   harder to see how my charm would work on say a wordpress charm vs a mysql charm vs a postgresql charm16:43
Fishy_without devolving into a   if postgress:  <>  else if mysql <>16:43
bloodearnestFishy_: the backup subordinate has a relation that the primary charm can implement to tell it what to back up16:45
Fishy_ok so then i would need to write my own version of postgresql that supports fishy-backup16:45
Fishy_which seems reasonable16:45
Fishy_except how to deal with upstream postgresql charm changes16:46
Fishy_part of what is nice about using a precanned charm is to get the upstream fixes16:46
bloodearnestFishy_: heh, I would like to use a charm called fishy-backup :)16:51
bloodearnestFishy_: fork and pull changes periodically works pretty well, especially in your case where you would probably not alter the upstream functionality, but add new stuff16:52
=== timrc is now known as timrc-afk
Fishy_ya that's true16:54
Fishy_would at least be an easy merge16:54
lazyPoweryeah we catch most of the schenanigans during peer review16:58
bloodearnestFishy_: we do that for all charms, and deploy from a local repo. We don't deploy from the charmstore at all16:58
=== timrc-afk is now known as timrc
Fishy_and juju download is the start of it?  or do you go and find the actual svn/git repo17:01
Fishy_assume juju download is just a snapshot and not a repo?17:01
=== vladk is now known as vladk|away
lazyPowerFishy_: in what context do you mean? charm get? that fetches teh current snapshot from wherever the parent branch of the charm is located.17:02
Fishy_i dont think charm get exists17:04
Fishy_in stable juju17:04
rick_h_Fishy_: it's in charm-tools package. You have to install that on its own17:07
Fishy_genius17:10
marco-travelingbbcmicrocomputer: i still think unit-changed would be a sweet hook17:11
=== DrabMakyo is now known as ShinyMakyo
Fishy_wow people still use bzr17:19
=== timrc is now known as timrc-afk
bbcmicrocomputermarco-traveling: what happened to charm-helper-sh?17:24
=== vladk|away is now known as vladk
=== vladk is now known as vladk|offline
=== cmagina is now known as cmagina-away
=== cmagina-away is now known as cmagina
=== timrc-afk is now known as timrc
=== cmagina is now known as cmagina-away
=== cmagina-away is now known as cmagina
marco-travelingbbcmicrocomputer: it was depricated17:55
marco-travelingbbcmicrocomputer: only available on precise and in a PPA, but probably won't be going to trusty17:56
bbcmicrocomputermarco-traveling: ah ok, thanks17:57
marco-travelingbbcmicrocomputer: if you need the ppa, I think ~charmers owns it17:57
* marco-traveling goes and looks17:57
bbcmicrocomputermarco-traveling: ah no worries, it was just my curiosity asking17:57
marco-travelingbbcmicrocomputer: really, we're pushing charm-helpers (v2) instead, which will have something soon for charms in shell17:58
bbcmicrocomputermarco-traveling: ah nice!17:58
=== vladk|offline is now known as vladk
=== Ursinha is now known as Ursinha-afk
* alexisb has just deployed her first service in the cloud using juju18:20
sarnoldwoo :)18:20
alexisbvery good documentation did not take long to figure out, cool stuff18:20
sarnoldisn't that fun? :)18:21
sarnold"ooh I just made a VM spin up a few hundred miles away and now it is doing my bidding!"18:21
alexisbit *was* fun :)18:21
=== Ursinha-afk is now known as Ursinha
marco-travelingalexisb: glad you had a good experience! Feel free to ask here if you have any questions :D18:41
sarnoldwallyworld: your connection is bouncing18:43
alexisbmarco-traveling, thank you and will do18:53
=== vladk is now known as vladk|offline
=== vladk|offline is now known as vladk
joseguys, if I want my hook to run a python script which does not go to background, how should I do it?19:15
=== Ursinha is now known as Ursinha-afk
=== Ursinha-afk is now known as Ursinha
=== vladk is now known as vladk|offline
=== CyberJacob|Away is now known as CyberJacob
marco-travelingjose: do you want the hook to block or not on the script? If the former, then just run it, if the latter write an upstart script20:00
joseuh oh, I think not, I'll check how to write an upstart script then20:00
josemarco-traveling: that phpbb charm I was writing, I'll need to stop as the feature we're looking for to get the DB info in place is not there yet20:02
josebut the base is there, maybe in the future I can re-do it20:02
marco-travelingjose: upstart is pretty simple if you need any help #upstart or ping me20:03
josecool, thanks!20:03
josemarco-traveling: would a charm be approved if the user has to go and manually get the database values and set them by itself?20:07
evdoes the charm hook context not pull in /etc/environment?20:07
marco-travelingjose: yeah, that won't fly20:09
marco-travelingev: probably/maybe not?20:09
evboooo20:10
webbrandonany juju contest going on?20:25
marco-travelingwebbrandon: not at the moment20:34
webbrandon:( `20:36
marco-travelingwebbrandon: there's an Ubuntu Touch app development competition going on20:55
evmarco-traveling, others: what's the preferred way of setting some environment data for all charms in a deployment? Say I'm at a very security conscious company and they have a http proxy for all traffic. I cannot just set the juju http-proxy config option because juju doesn't also have a no_proxy setting for putting the IPs of swift, glance, etc in.20:58
marco-travelingev: uh, that's a good question20:58
marco-travelingev: so, there's set-environment, but that's really only for the juju deployed environment20:59
marco-travelingnot like, environment variables20:59
marco-travelingor, maybe that command does do that20:59
marco-travelingev: let me check20:59
marco-travelingev: yeah, not, it's just to modify the juju environment, not environment variables21:01
marco-travelingev: so, my first though it like a proxy subordinate, that can talk you can set filter based rules on21:02
marco-travelingerr, not subordinate, service21:02
marco-travelingso set http proxy to this deployed service, then update it's configuration with filtering rules on which ips to send to the actual proxy and which shouldn't be21:03
marco-travelingbut that sounds messy21:03
evmarco-traveling: I have the ability to bake the image21:04
evbut I need to know where to stick http_proxy https_proxy, socks_proxy, no_proxy, etc so juju respects them21:04
evas when I put them in /etc/environment in the baked image, it didn't find it's way into the environment for $charm/hooks/hooks.py21:05
evoh21:05
marco-travelingev: I'd ask around in #juju-dev21:05
evokay, thank you21:05
marco-travelingThere might be a place that you can place these environment variables within juju that the hooks will pick it up21:05
=== cmagina is now known as cmagina-away
Fishy_install is ran by user root, but start is run by user ubuntu?21:36
Fishy_is that the convention21:36
Fishy_or anything except install is run by ubuntu21:36
peterklipfelhi all, I made an oops, and now I have a mess. I accidentally spun up too many lxcs, my computer exploded, so I killed all the juju processes and deleted the lxcs. When I run juju status, I see a bunch of non-existent containers. The processes are still around, but they won't die (I assume there's a daemon respawning them). How do I reset everything?21:36
Fishy_sudo juju destroy-environment21:36
peterklipfelFishy_: that doesn't work. I get "ERROR failed to rename container directory"21:37
Fishy_next time dont delete stuff by hand ;) messed somethin up21:38
Fishy_have not seen that error though21:41
Fishy_(yet)21:41
Fishy_for juju could just blow away ~/.juju directory21:42
lazyPowerpeterklipfel: use --force21:42
Fishy_but lxc has stuff elsewhere21:42
lazyPowerpeterklipfel: juju destroy-environment local --force21:42
peterklipfelok. Actually, I was almost there. I just had to go in and manually recreate the directories (empty) and then run destroy-environment until it succeeded21:42
peterklipfelthe --force flag made juju complain. It doesn't recognize it21:43
lazyPowerpeterklipfel: are you on the 1.16 series of juju?21:43
peterklipfellazyPower: yes21:43
lazyPowerah ok, thats a recent convention21:43
lazyPowerhang on there are instructions for cleaning up a broken local install21:43
lazyPower1 moment while i fetch them21:43
lazyPowerpeterklipfel: http://askubuntu.com/questions/403618/how-do-i-clean-up-a-machine-after-using-the-local-provider21:44
peterklipfellazyPower: thanks!21:45
lazyPowerpeterklipfel: np, let me know if you run into any issues. I've gotten my hands dirty quite a few times with LXC  misbehaving21:45
Fishy_how do I tell the juju that I have something running on port 8022:10
Fishy_so it shows up on status22:10
Fishy_and gets exposed22:10
Fishy_provides:   website: http22:11
lazyPowerFishy_: open-port exposed22:11
lazyPowerer22:11
lazyPowerwow22:11
Fishy_?22:11
lazyPowerFishy_: open-port 80 in your hooks.22:11
Fishy_start hook?22:11
Fishy_or install22:11
lazyPowerthat works.22:11
Fishy_i guess install22:11
lazyPowerI typically put them in config-changed if i have an ssl option so i can switch the ports on the fly22:11
=== cmagina-away is now known as cmagina
lazyPoweror if there is a port setting, i can react on it. but if its always going to be port 80, put it wherever it makes sense.22:12
marco-travelingFishy_: you typically want to open-port when the service is ready to be used22:12
Fishy_open-port : command not found22:12
Fishy_I am in bash land22:12
lazyPowerFishy_: it has to be executed within hook-running context22:12
marco-travelingFishy_: so if after installation its ready, install hook. If you have to start the service, then start hook, if it requires a database, then the db-relation-changed hook22:12
Fishy_o okay22:13
Fishy_not ready till start happens22:13
marco-travelingFishy_: then open-port 80 in the start hook should suffice22:13
Fishy_config changed is good idea when I tie it to the config22:13
Fishy_but thats step 1122:14
marco-travelingFishy_: yeah, like lazyPower said, if you can configure the port then it should /definitely/ be in the config-changed hook instead22:14
marco-travelingbut as you mentioned, you can iterate on that later22:14
Fishy_is the juju master node a single point of failure in this whole operation?22:14
marco-travelingFishy_: it is at the moment, we have HA landing for the bootstrap node in the next week or two22:14
marco-travelingFishy_: and it's a point of failure in terms of controlling the environment, if the bootstrap node goes down the other services will continue to run22:15
marco-travelingyou just wont' be able to orchestrate anymore (deploy, relate, expose, destroy, etc)22:15
Fishy_and if i brought it back up, it wouldnt re-discover what was done22:16
Fishy_so potentially lost track of what my machines are doing22:17
Fishy_cool open-port works22:19
marco-travelingFishy_: well, if you rebooted the node22:19
marco-travelingFishy_: it would pick up where it left off22:19
marco-travelinglike, if there was a networking outage or something similar22:19
marco-travelingbut if it was deleted, then there's really no way to recover from that at the moment22:20
marco-travelingwith HA you'll be able to deploy and scale your bootstrap node, so if you wanted HA you could have three bootstrap nodes, if one goes down the other two will continue serving22:20
Fishy_ya that would be good long term22:21
Fishy_for sure22:21
Fishy_master - slave ?22:21
Fishy_or can i run commands from any of them22:21
marco-travelingFishy_: well, you don't run commands from the bootstrap nodes22:24
marco-travelingFishy_: the bootstrap nodes expose an API server that the CLI and GUI, etc talk to22:24
Fishy_ya but can you talk to all 3 of the HA22:29
Fishy_or just 122:29
marco-travelingFishy_: all three, but juju will just select one22:45
bodie_do I really need to use the gophers/go PPA to get lbox or can I just go get launchpad.net/lbox?23:23
marco-travelingbodie_: the latter is better23:40
=== cmagina is now known as cmagina-away

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