=== zyga is now known as zyga-afk [02:25] * SpamapS is getting even closer to generic monitoring now.. :-D [02:26] I wonder if there is an open source real user monitoring package [02:29] yes [02:29] the one newrelic uses is open source too [02:29] they are all based on the same lib from a google employee [04:37] SpamapS: there is also jiffy and mod_pagespeed, probably the most used though is the one google analytics and newrelic use called epsiodes http://stevesouders.com/episodes2/ [06:15] I have deployed maas on a server and trying to deploy new servers from this one, but all my new instances have instance-state: unknown [06:15] any ideas of where and what to look at? [07:57] all i wanted was a pepsi, just one pepsi, and she wouldent give it to me ... [08:47] bbcmicrocomputer, morning [08:47] moring all! [08:48] jamespage: morning [08:48] bbcmicrocomputer, fancy having a stab at the solrcloud charm? Apache just alpha'ed Lucene and Solr 4 [08:48] and we already have a nice zookeeper charm.... [08:49] jamespage: sure, am already working on the existing Solr charm to add dynamic core/db hookups [08:49] jamespage: assign the ticket to me :) [09:24] bbcmicrocomputer, bug assigned - have fun! [09:24] looks interesting [09:24] jamespage: cool, thanks [09:26] jamespage: yeah it's pretty cool being able to distribute Solr.. although they still don't do distributed idf, so you have to be careful how you distribute your documents amongst servers [09:28] bbcmicrocomputer, have you looked at solandra at all? [09:28] jamespage: only briefly [09:29] bbcmicrocomputer, I did a while ago [09:29] but not recently [09:29] it looking interesting [09:32] jamespage: yeah that is quite cool using the Cassandra backend [09:32] bbcmicrocomputer, the last solr deployment I did had a multi-master configuration across data centres [09:33] jamespage: that is pretty impressive! [09:33] we ended up using reliable message queues to ensure consistency between active site application and both masters [09:33] but it was complex [09:34] I think solrcloud and solandra both solve that issue more effectively.. [09:34] jamespage: yeah, sounds like it would [09:52] does Juju support use specific instance when deploy a charm [10:21] freeflying, specific instance or specific instance type? [10:22] jamespage: instance type I meant, sorry [10:22] freeflying, in which case the answer is yes [10:22] * jamespage digs 'contraints' documentation out [10:23] freeflying, take a look here - https://juju.ubuntu.com/docs/constraints.html [10:23] jamespage: saying I want deply hadoop, can I have master deployed with one type, and slave with another type instance? [10:24] freeflying, yep - take a look at http://markmims.com/cloud/2012/06/04/juju-at-scale.html [10:24] this is some testing we did with hadoop + juju at scale [10:24] the master is deployed on m1.large and the slaves on m1.mediums [10:25] jamespage: yup, just get there, very nice, thanks [10:25] freeflying, note that we have a few 'hadoop' charms in the charm store ATM [10:25] use 'hadoop' rather than any of the others - there is a bug outstanding to get that resolved. [10:25] freeflying, and pester me if you have any issue - I wrote it :-) [10:26] jamespage: very neat :) [10:27] freeflying, I need to write an in-depth blog post on using the hadoop charm - the README has quite a bit of info [10:28] http://jujucharms.com/charms/precise/hadoop [10:28] it has lots of deployment topologies... [10:28] jamespage: at first glance, its a complex charm :) [10:29] freeflying, its certainly more complex than some [10:29] it uses a late binding technique to decide what role a service is playing as its related to other deployments of the hadoop service [10:30] jamespage: but for many devops, they don;t need dig into the charm itself, just use it, get their workload fullfilled [10:30] freeflying, +1 to that [10:30] jamespage: thats the magic of juju :) [10:30] absolutely [10:31] freeflying: wow, long time since I've seen your face round , how ya been :) [10:31] heya jamespage [10:31] imbrandon, hey - just looking at you nginx charm [10:31] imbrandon: doing pretty well, how about you [10:32] jamespage: ahh its ummmm, functional , still many rough edges but i wanted to get others eyes on it [10:32] :) [10:32] freeflying: good good, more juju than kde these days :) [10:33] imbrandon: yep, some of the irc logs tells that obviously [10:34] * imbrandon is actualy writing a proof-of-concept jitsu replacemnt/arch rewrite , heheh dunno how its gonna fly tho [10:34] imbrandon: it sounds cool [10:35] gonna run it by SpamapS and maybe jamespage this morning and see what they think [10:35] :) [11:18] mmmm lp:~imbrandon/+junk/pepsi [11:19] fooooood time , afk [11:38] imbrandon, what is the container scoped website relation for in the nginx charm? === zyga is now known as zyga-afk === zyga-afk is now known as zyga [12:44] Is there a way I can get a postgresql service to give me a db & creds without a charm that relates to it? [12:44] Just wondering if I can spin one up to poke at while I'm figuring some things out === salgado is now known as salgado-brb === salgado-brb is now known as salgado [14:16] jamespage, there's a juju-jitsu deploy-to [14:17] oh.. [14:18] jamespage, solandra is dead, its folded into datastax enteprise along with brisk [14:27] hazmat, thought it might be === zyga is now known as zyga-food [15:00] jml: you want to run the relation hook without an actual relation? [15:00] SpamapS: pretty much, yes. [15:00] jml: you could deploy a subordinate to the pgsql box .. pgclient or something like that [15:00] Hmm. [15:00] jml: seems a bit silly tho.. why not just deploy a dummy charm and simulate w/ debug-hooks? [15:01] SpamapS: "just" [15:01] charm create dummy [15:01] add relation [15:01] deploy, debug-hooks [15:01] SpamapS: I ended up installing postgresql server on my laptop [15:02] SpamapS: there's a bunch of stuff on my laptop that would be tedious to move over to a new juju instance [15:05] jml: bind mount in the LXC might help.. right? [15:08] SpamapS: yeah. [15:10] SpamapS: although my original motivation was convenience (hey, juju makes it easy for me to set up a throw-away postgres db, right?). as is, it's turned out to be less convenient. [15:10] SpamapS: which isn't such a big deal. it's probably just not the right tool for the job. [15:14] why did it turn out to be less convenient? [15:15] jml: was it just less convenient because you couldn't type 'sudo su postgres' to become the admin? [15:16] 'sagi -y postgresql; sudo -u postgres createuser -s -d jml; sudo -u postgres createdb -O jml udd' [15:16] jml: because 'juju ssh postgres/0 -t sudo su - postgres' is about the same :) [15:16] jml: indeed, that is easier if you don't mind having it native and such [15:17] SpamapS: right. I don't mind having it native. I'd prefer not to, but I can live with it. I certainly prefer it to having the *client* machine being another instance, which involves all sorts of tedious copying of source and data files. [15:17] or worse, remote editing [15:18] right forget the client being a juju box.. [15:19] jml: I think just deploying, and running the same sudo stuff via juju ssh is fine === salgado is now known as salgado-lunch [16:39] jml, you can invoke a relation hook in a non relation hook context, but outside of any hook context? [16:40] oh.. you want the data === fenris is now known as Guest37369 === zyga-food is now known as zyga-afk === salgado-lunch is now known as salgado === almaisan-away is now known as al-maisan === al-maisan is now known as almaisan-away === Leseb_ is now known as Leseb [22:43] hrm.. config-get seems to have no way to print out an empty string properly [22:43] if I specify a single value that is an empty string it should just print the string of bytes, not ''