_mup_Bug #939932 was filed: Should be able to set --repository from system environment <juju:New> < https://launchpad.net/bugs/939932 >00:01
hazmatSpamapS, why does my repo need to be in ~/.juju00:02
SpamapShazmat: I mean your pointer to repository should be able to be specified from a config file somewhere in ~/.juju00:03
hazmatSpamapS, ah.. yeah.. that sounds good00:03
SpamapShazmat: I'm not entirely sure environments.yaml is the right place00:03
SpamapShazmat: one thought I had.. environments should actually be usable as repositories00:04
SpamapSlong term anyway :)00:04
* hazmat notes its beer o clock around here00:18
SpamapShazmat: yageshegedra00:19
_mup_Bug #939944 was filed: Default Charm namespace should be overridable by environment variables. <juju:In Progress by clint-fewbar> < https://launchpad.net/bugs/939944 >00:19
* SpamapS is getting the hang of lbox.. ;)00:19
locke105is there a way to use juju without an amazon S3-like store? looking for a way to deploy to an openstack setup but I don't want setup a Swift module if i dont have to04:39
_mup_juju/enhanced-relation-support r9 committed by jim.baker@canonical.com04:51
ejatcan someone review lp:~fenris/charms/oneiric/symfony/trunk07:53
ejatim having a problem when i run the install script manuall on ec2 .. its works ..07:54
ejatbut when i use juju deploy ...07:54
ejatit can be view / test on ec207:54
ejat!ping SpamapS & hazmat07:56
ejat!ping jcastro07:56
m_3ejat: hey, so take a look at the logs on the service unit08:55
m_3ejat: they're found in /var/lib/juju/units/<unit>/charm.log08:56
m_3that should help debug this a bit08:56
ejatm_3: ok thanks ..09:01
m_3ejat: I'm spinning it up now, but I created a new Bug #94014009:03
_mup_Bug #940140: Charm needed: Symfony <Juju Charms Collection:New> < https://launchpad.net/bugs/940140 >09:03
m_3please grab it and attach your branch... it pops up on our review queue if you also add the tag: 'new-charm'09:04
ejatok thanks m_309:05
ejatill try to update the bugs accordingly09:05
m_3np, it's just a good place to keep track of conversation about it... other than irc logs :)09:10
m_3ejat: it just came up to a 'started' state for me09:24
m_3ejat: what problems are you seeing?09:25
ejatyeah .. show started .. but when i open the public dns .. its show nothing ..09:28
ejatrather than i manually deploy using the script ..09:29
m_3are you "exposing" the service after deploying?  i.e., 'juju expose symfony'09:30
m_3(it's firewall stuff)09:30
ejatm_3: yeah .. did it also ...09:31
ejat# Make it publicly visible, once the symfony service is exposed09:31
ejatopen-port 80/tcp09:31
ejatinside the install09:31
m_3there's an additional command that runs from your client09:31
m_3juju bootstrap09:31
m_3juju deploy ...09:32
m_3juju expose symfony09:32
m_3it's designed to be a controlled/gated network exposure for a whole stack of services09:32
ejatyeah .. only done that 3 command ..09:33
m_3ejat: hmmm... if juju status shows that the service unit is 'started' and lists 'open-ports:', then I'm not sure what's going on09:35
m_3I'm getting a 'It works!' from apache and a directory listing when I hit http://<ec2-addr>/symfony09:36
m_3right, that's showing the pretty 'project created' page:  http://ec2-107-22-14-139.compute-1.amazonaws.com/symfony/web/09:38
m_3maybe it's been upgraded and is a bit stale?  you might try recycling your environment if nothing else is running09:40
ejatyour instance on precise ?09:43
ejator oneiric ?09:43
ejatm_3: if u see that .. mean my charm work :)09:44
ejator i should change my     default-series: oneiric09:45
ejatin my juju environments09:45
m_3ah, sorry... I was assuming you were using oneiric instances because your branch had charms/oneiric/symfony09:46
m_3ejat: yes, it looks like the install part of the charm works fine09:46
m_3I'd recommend changing to 'default-series: oneiric' in your ~/.juju/environments.yaml file09:47
ejati already put that in my environment09:48
ejatcan u try http://ec2-46-137-225-47.ap-southeast-1.compute.amazonaws.com/symfony/web09:50
* ejat confuse .. y u getting the pretty page .. but i am not :(09:50
ejatm_3: can u view it ?09:52
m_3ejat: I get a php file09:53
m_3but it's not viewable09:53
ejatyeah .. thats what i get too ..09:53
m_3this is on an oneiric instance?09:53
* ejat wondering y u can get it :)09:53
ejatyups .09:53
ejatjust now u launch it on what instance?09:54
m_3try to 'juju ssh symfony/0'09:54
m_3and then 'sudo apache2 restart'09:54
m_3see if that fixes it09:54
m_3I've only run it on oneiric09:54
ejat0 or 1 ?09:55
m_3sorry, that should be 'sudo service apache2 restart'09:55
m_3whichever unit you've got started09:55
ejatyeah its works09:55
ejatafter restart09:55
ejatcan u retry09:55
m_3I bet you need an extra restart after you make the symlink for sf09:56
ejatowh okie ..09:56
m_3yes, I see the pretty 'project created' page09:56
ejatbut need someone to review the code .. make it more clean ..09:58
m_3sure... we can put it through the usual review process.  When you're ready just add the 'new-charm' tag to that bug10:00
m_3and I or somebody in the group will review the branch10:01
m_3not right now though :)... back to sleep for me10:02
ejatthanks so much m_310:03
m_3glad to help10:04
ejatsee ya soon .. or maybe in uds-q .. :)10:06
* ejat also feel \\0/ can contribute some charm as state within the uds-p milestone .. 10:07
ejattrying to submit other charm if possible .. gtg to0 ...10:08
=== grapz is now known as grapz_afk
=== grapz_afk is now known as grapz
jamespagehazmat, m_3: I'd really like to refactor the hadoop charm into mapreduce and hdfs charms - any objections?  I'll leave the current hadoop-master/hadoop-slave as is12:28
hazmat james_w sounds good to me12:28
hazmatjamespage, +112:29
* hazmat checkouts ironfan12:29
jamespagehazmat, it would mean that the hbase-master charm could be multi-master much more easily12:29
=== grapz is now known as grapz_afk
=== grapz_afk is now known as grapz
SpamapSjamespage: that sounds great! :)12:34
* jamespage is running mapreduce jobs on his pandaboard this morning :-)12:35
=== grapz is now known as grapz_afk
hazmatjamespage, nice!  i've been meaning to touch base someone about juju on arm12:38
hazmatjamespage, which jvm?12:38
jamespagehazmat: the only one for armhf (openjdk)12:39
jamespagenow that the thumb2 port has landed as default in openjdk-6 its actually OK12:39
jamespageand passes the TCK12:39
=== grapz_afk is now known as grapz
m_3jamespage: splitting hadoop makes sense to me15:23
m_3jamespage: I also like Clint's approach (ceph and mysql) to a single charm that can be deployed into different roles15:28
m_3jamespage: either way, I'm not particularly fond of the names "-master" and "-slave"15:29
m_3professionally, that is :)15:29
jamespagem_3: remind me how that works again?15:41
m_3sure... so consider the simple one... mysql15:43
m_3juju deploy ... local:mysql masterdb15:45
m_3juju deploy ... local:mysql slavedb15:45
m_3juju add-relation masterdb:master slavedb:slave15:45
m_3probably better named something like 'replicaset1' or something... but I'm cutting/pasting15:45
jamespageah - I see15:49
* jamespage goes to refactor his charms15:49
m_3yeah, it's really elegant15:50
m_3but shhhhhh... we don't want it to go to his head :)15:50
m_3jcastro: review queue length back down to 1 (I'll catch saltstack next week)16:39
jcastrom_3: was rick's thing hard or easy?16:39
m_3jcastro: pretty straightforward... just some basic updates to the rails charm16:45
jcastrohave you had a chance to look at summit yet?16:45
m_3I'll put the rails one up for review soon... wanted to do some cleanup before showing it to rails peeps at conferences16:46
m_3summit's next on the list today16:46
* jcastro nods16:46
m_3the good news is we actually have a django charm16:46
jcastrohey alright!16:46
m_3like the rails/node charms, this one will deploy a django app straight from a bzr repo16:46
jcastrothat's basically awesome16:47
m_3so hopefully it'll go quick16:47
jcastroit gives you a good story for those language conferences16:47
m_3haven't really checked the state of the charm though16:47
m_3might need some updating / completion... it was never put up for inclusion into the store for some reason16:47
m_3written by... Michael Nelson16:48
m_3looks like great work... just don't know if he finished or not16:48
m_3I'll find out today/tomorrow16:48
gary_posterm_3, thanks for review!  We'll get to the fixes16:53
m_3gary_poster: thanks for the charm!  please do branch lp:charm-tools too when y'all get a chance... y'all have great stuff for a charm-helpers-py!16:55
* m_3 too many "y'all"s in one place... I guess I'm feeling very Texan atm16:56
m_3my fav is still "y'all're".... which is actually more characters than "you are"... which defeats the purpose of a contraction... harumph16:57
SpamapSm_3: but it rolls of the tongue so nicely... like y'all're just a rollin in thu mud on a sundee17:08
jamespagem_3: that works really nicely17:23
jamespagethanks for the tip!17:23
jamespagem_3, http://bazaar.launchpad.net/~james-page/charms/precise/hadoop-hdfs/trunk/view/head:/README17:24
m_3jamespage: awesome... that looks great17:25
jamespagem_3: certainly reduces the amount of code replication across the two charm approach (most of it is the same)17:25
m_3we can add additional namenodes too... not sure what to call that relation tho :)17:25
* SpamapS is all for charm consolidation whenever possible17:26
m_3yeah, buildbot ended up using cross-charm links to reduce code-repl17:26
SpamapSI've been thinking about a streamlined approach to configuring the various roles.. I wonder if we could add a --role=slave and that would pull $charmroot/roles/slave.yaml as the default values to override the usual defaults..17:26
m_3consolidation would work well for them17:27
SpamapSThis happens with ceph really badly.. you don't know if its a single node ceph tester, or a 100 node ceph buildout, until the 2nd unit is added...17:27
m_3SpamapS: including sample deploy scripts into the readme works well for that imo17:27
SpamapSm_3: yeah, thats the way right now17:27
SpamapSI'm getting way too far ahead of myself with the role thing. :)17:28
m_3they had a $CHARM_DIR/samples/ directory with various config scenarios17:28
* m_3 makes bruce lee fighting sounds17:29
m_3jamespage: a namenode-cluster gets complicated with the relations though... I'd hvae to do that one on the whiteboard :)17:30
jamespagem_3: yeah - in 1.0.0 it is a single point of failure TBH17:31
m_3w/ a lot more coffee17:31
jamespageits possible with DRBD and stuff17:31
m_3I thought they added that in 20517:31
jamespagem_3: nah17:32
m_3or was that for the 0.23 line17:32
SpamapSI'm more and more convinced that some charms need a config item that is "wait for relationX" so that you can just skip configuration until relationX is established, so that you don't get into the situation we see with nova where it stores in sqlite, then mysql17:32
m_3ah, gotcha... my bad17:32
jamespageSpamapS, +1 to that17:32
SpamapSI need to return to the ceph charms so I can demonstrate that..17:32
SpamapSactually using that approach would probably allow easy mysql ring replication for the mysql charms too17:33
m_3SpamapS: or some way to know that the relation is actually completed before then doing a 'juju set --config ' from an outside script17:33
SpamapSm_3: hm, I think thats another approach, but one that will be more complicated and not yield better results17:34
jamespageother than using a full config management solution like puppet does anyone have a nice way of determining whether a restart of a service is required?17:34
jamespageI though about checksumming the config files and only restarting if stuff had changed.17:34
m_3jamespage: hmmm... no clue17:38
m_3that might be useful as an upstart primitive though.... service xxx status17:39
SpamapSjamespage: I do that in mysql17:42
jamespageSpamapS, I'll take a look - just trying to avoid blind restarts17:43
jamespage'something might have changed'17:43
SpamapS        if oldhash != tmd5.digest():17:43
SpamapS            os.rename('/etc/mysql/my.cnf','/etc/mysql/my.cnf.%s' % md5.hexdigest())17:43
SpamapS            os.rename(t.name, '/etc/mysql/my.cnf')17:43
SpamapSjamespage: with mysql is pretty important to only restart when critical configs are changed. ;)17:44
jamespagenow to refactor the hbase-* charms into a single charm17:48
_mup_Bug #940492 was filed: Charm metadata must support subordinates and container relations <juju:New> < https://launchpad.net/bugs/940492 >17:51
_mup_Bug #940498 was filed: Juju must provide the implict juju-info relation <juju:New> < https://launchpad.net/bugs/940498 >17:56
_mup_juju/enhanced-relation-support r10 committed by jim.baker@canonical.com21:59
_mup_More work on spec21:59

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