=== tasdomas is now known as tasdomas_afk [01:14] ok, so as long as im not trying to provision anything on the actual host, i should be able to run juju within an lxc guest, correct? === freeflying is now known as freeflying_away === freeflying_away is now known as freeflying === freeflying is now known as freeflying_away === freeflying_away is now known as freeflying === shang_ is now known as shang === tasdomas_afk is now known as tasdomas [05:54] why am i getting 'error: net: no such interface' when i try to deploy juju bootrap? I definitely already have lxc working on this host and a couple containers going [06:23] MACscr: is this using the juju client ? [06:24] davecheney: lol, my last name is chaney =P [06:24] close, but different [06:24] anyway, yes, its the juju client i guess [06:25] MACscr: what is the command you are using ? [06:25] i installed the juju-core, etc, on it in hopes of deploying the juju-gui within an lxc instance [06:26] 'deploy juju boostrap' [06:26] im following this tutorial (the answer) http://askubuntu.com/questions/65359/how-do-i-configure-juju-for-local-usage [06:26] MACscr: 'deploy juju bootstrap' isn't a command I know [06:28] well 'juju bootstrap' gives the same response [06:28] * davecheney reads http://askubuntu.com/a/65360 [06:28] MACscr: the instructions say 'sudo juju bootstrap' [06:28] sudo o_O [06:29] ah. [06:29] my apologies, i was using sudo, same result [06:30] sarnold: you have to use sudo for the local provider as lxc requires root to create and destroy containers [06:30] this is the only time it is required [06:30] it is a limitation of lxc [06:30] MACscr: can you do it again, pass -v and paste the output [06:30] pastebinit, or paste.ubuntu.com [06:31] http://paste.ubuntu.com/6027630/ [06:31] is it possible to run more than 1 charm on a node? [06:31] why is it trying to use that bridge when im using br0 for LXC as im not using nat, im getting dhcp from the actual router [06:32] for example to consolidate several openstack services on a single node? [06:32] kurt_: yes. using LXC [06:33] kurt_: investigate deploy-to -- I'm not sure of its current status.. [06:33] yes, I remember that was around, but I was unsure of its current status too [06:34] * kurt_ thinks its time to start RTFMing about LXC [06:35] MACscr: looks like you either don't have lxc installed, or it isn't installed correctly [06:35] you have no lxc bridge address [06:35] kurt_: I'm not sure lxc would be a good fit for deploying parts of openstack environment -- juju couldn't manage lxc containers on some hosts and maas or kvm on other hosts... [06:35] MACscr: I will log a bug for this, juju could be clearer in what has gone wrong [06:36] sarnold: --to is available in 1.12 and later [06:36] davecheney: sure i do, i just dont have one setup for nat. I currently have 2 LXC instances installed and working great [06:36] (i think, it is certainly avialable in the 1.13 devel series) [06:36] davecheney: AWESOME :D [06:36] davecheney: thanks [06:36] MACscr: all I can say is the way your LXC is setup is not how juju expects it to be [06:36] and it can't handle it [06:37] well thats obviously a mistake on its part to assume im using lxcbr0 =P [06:37] davecheney: will deploy to stay in the code? I think it was there before, then removed [06:37] is that configurable? [06:37] now added back again :D [06:38] https://bugs.launchpad.net/juju-core/+bug/1216775 [06:38] <_mup_> Bug #1216775: cmd/juju: local provider doesn't give a clear explanation when lxc is not configured correctly

[06:38] kurt_: s/--deploy-to/--to/ [06:39] MACscr: it is not configurable, it is a hard requrement at this time [06:39] ahh [06:39] kurt_: the name of the optoin has changed [06:39] it used to be a jitsu(?) plugin [06:39] but we intergrated it into the cli and it's an option on the deploy subcommand [06:39] but jitsu has been depreciated, right? [06:40] kurt_: yes, it's dead, like the parrot [06:40] lol [06:40] I'm currently on .7. do I have these options? [06:41] kurt_: no, you should upgrade to 1.12 [06:41] full warning: there is no upgrade path from 0.7 to 1.12 [06:41] but there wasn't one from 0.6 to 0.7 either [06:41] so we haven't really made it much worse in that respect [06:41] lol [06:42] there is an upgrade path once youre on juju 1.x [06:42] but you still have to bite the bullet and move away from your old pyjuju environments [06:42] davecheney: would I basically be starting from scratch? [06:42] davecheney: no upgrade path? I had the impression it mostly just worked? [06:43] ie. have to destroy the environment [06:44] kurt_: yes, just like 0.6 -> 0.7, you will need to make a fresh environment [06:44] * kurt_ winces [06:45] sarnold: 1.x and up has an upgrade path [06:45] there is no upgrade path from 0.x to 1.x [06:45] sorry [06:46] well, I guess if it allows me to deploy multiple charms and thus services on fewer nodes, its probably worth it [06:47] is 1.12 stable? [06:47] kurt_: yes [06:47] that is why it lives in the stable ppa :) [06:48] davecheney: does lxcbr0 have to to nat? if not, can it use dhcp? [06:48] ok, I'm not totally familiar with bzr / ppa and all that yet - learning [06:48] i guess i could rename things [06:48] or do i maybe need to create a second network [06:49] MACscr: lxcbr0 is a layer 2 bridge mode interface [06:50] well technically its just the name of an interface. I was never aware it had to be used in a certain fashion [06:51] MACscr: i'm not really sure how you have configured your lxc networking [06:51] so pretty much i cant use juju with lxc unless i want to use nat? which is obviously pretty useless without doing some sort of port forwarding [06:51] juju pretty much expects to be able to control it [06:51] and that the setup that arrives out of the box is the setup that is present when you try to use the local provider [06:52] MACscr: i don't know if this is related to nat [06:52] it's probably simply that you've change the default LXC config from what juju is expecting [06:52] the local provider is still classed as experimental [06:52] or (insert word != production) [06:52] davecheney: out of the box, all of the lxc guests are not accessible outside the host because they are using nat [06:53] MACscr: using the bridge mode interface we expect they DHCP from the same source as the host === defunctzombie_zz is now known as defunctzombie [06:55] davecheney: i apologize for my ignorance, but both my host and my lxc instances are all getting their ip addresses from the same dhcp server on my router [07:03] MACscr: s'ok [07:03] no need to be sorry [07:03] lxc is super complex [07:04] :) [07:04] and has loads of knobs to twiddle [07:04] davecheney: not really, it can be complex, but it doesnt have to be in its most basic form =P [07:16] davecheney: so from the sounds of it, there is no way to use juju with lxc and still have the lxc instances accessible on the LAN? [07:19] MACscr: i don't think that is correc [07:19] that is exactly how the juju lxc provider expects things to work [07:19] we bridge the lxc containers onto whatever your default ethernet interface is [07:19] so they get DHCP addresses [07:19] i think [07:19] let me check [07:19] thats not now lxc works by default [07:19] er, how [07:20] MACscr: you probalby know more about it than me [07:20] well im only a few days into lxc [07:20] MACscr: i'm waiting for confirmation [07:20] you may find it easier to ask this question in #juju-dev [07:32] i installed juju but when i used it it gave me this error: error: No environments configured. Please edit: /root/.juju/environments.yaml it seems it's a bug . what should i do? [07:32] tokern3: did you mean to run juju deploy as root? [07:33] sarnold: yes [07:35] sarnold: http://fpaste.org/34727/13775025/ [07:38] tokern3: did you configure your environments.yaml properly? [07:38] no i just installed it today sarnold. i did nothing on it [07:38] http://fpaste.org/34729/75027261/ sarnold [07:39] tokern3: ah. you must tell juju how to contact your cloud, with which credentials, for it to be useful. :) [07:39] tokern3: really, please don't run juju as root [07:39] tokern3: there are links for configuring juju for aws, hp cloud, and openstack: https://juju.ubuntu.com/docs/ [07:39] there are only two cases where you need to use sudo, and they are using the local/lxc provider [07:40] bed time :) have fun! [07:40] sarnold: wait a min [07:40] tokern3: i don't think juju deploy is doing to do what you think there [07:40] i wanted to install jenkins [07:40] http://webcache.googleusercontent.com/search?q=cache:34CJESdYOe8J:https://wiki.jenkins-ci.org/display/JENKINS/Installing%2BJenkins%2Bon%2BUbuntu+&cd=2&hl=en&ct=clnk&gl=us [07:40] i'm doing every command in this page . [07:41] is the cloud part necessary ? [07:41] tokern3: but -where- did you want to install jenkins to? your amazon cloud? your hp cloud? your openstack cloud? some lxc instances? [07:41] tokern3: the cloud is the whole point of juju. :) [07:41] tokern3: ok, you need to configure juju first [07:41] to tell it where to deploy jenkins too [07:41] https://juju.ubuntu.com/docs/getting-started.html#install [07:41] ^ start at configure [07:44] tokern3: if you just want to install software on one machine for production use, and you're _not_ using any cloud computing infrastructure, maybe you're _really_ looking for "apt-get install jenkins". [07:45] tokern3: I suggest reading this page before going much further: http://blog.labix.org/2013/06/25/the-heart-of-juju [07:45] you might be in the wrong place depending upon what you want to get done. :) [07:45] good luck, and have fun. :D [07:46] sarnold: i installed jenkins : apt-get install and det port8080 for it but i didn't set juju at all. my goal is integrate autotest into jenkins. thank you for help === defunctzombie is now known as defunctzombie_zz === defunctzombie_zz is now known as defunctzombie === defunctzombie is now known as defunctzombie_zz === gary_poster|away is now known as gary_poster [13:34] jcastro: if you guys want to hit up some node and ruby folks as well as java/etc. check out adding codemash.org to your list of conferences to hit up [13:36] * jcastro nods [13:36] rick_h: do you typically go? [13:37] jcastro: I've not gone the last two eyars, but went the 3 before that [13:37] man I am 90% sure we're sprinting that week [13:37] jcastro: a little bit of a outside the normal crowd [13:37] which I think is good tbh [13:38] want to submit a talk? :) [13:38] jcastro: heh, was thinking about it. I tried to bring some python love there, but it's a small chunk of people. JS/Ruby would be larger communities. [13:38] will think on it, but wanted to put it on your radar [13:39] yeah [13:39] maybe marco or mims can join you [13:41] jrwren: do you recall numbers from there? I know they stopped growing to keep from outgrowing the space, but can't recall how many it was. [13:45] rick_h: what? [13:45] jrwren: how big is codemash? 600ish people? [13:45] oh the kalahri expanded. last year was 1200 [13:45] likely to be 1200 this year too [13:45] ok, I wanted to think over 1k but seemed too big and knew that was around OLF. Thought I had confused them. [13:46] you gonna submit juju talks? [13:46] jrwren: you should do one! [13:46] jrwren: I'm thinking about it. I want to charm up bookie now that lxc containers work I can debug the charm nicely [13:46] our node charm should be finished by then [13:46] or do the rack charm, it's ninja [13:46] jrwren: but I know there's a bit node/ruby/windows community so might be cool to hit up off the beaten path [13:47] Do a node app .... on Azure. [13:47] jcastro: ok, i'll submit one. [13:47] rick_h: yes. there is a small python group there too. [13:48] jrwren: yea, I've tried to rep python there, but not made it the last couple of years [13:48] IMO juju transcends platform, well, it doesn't help windows. [13:48] rick_h: i'll try to rep python more there. [13:48] jrwren: yep, was thinking it'd be good to get some visiblity in some circles that don't read us that often [13:49] I can get a box of shirts, and swag, etc. [13:49] let me add it to my events target sheet [13:50] jrwren: are you entering the charm contest? [13:50] jcastro: not afaik. [13:50] jcastro: if I use juju much, it will be at work on proprietary codes. public charms will be of no use. [13:51] charms can deploy anything [13:51] other than that, I feel all the cool charms are written. [13:51] even blobs! [13:51] wordpress and discourse charms exist. The world is complete. [13:51] lol [13:51] when do CFPs open for codemash? [13:52] I think they are open. [13:52] I was actually whining to people to stop demoing wordpress [13:52] it's like man, show something complicated [13:52] the point is that it is not supposed to be complicated. [13:52] yeah so if you wanna do node that charm should be rewritten by then [13:53] jrwren: well, installing PHP stuff is relatively simple to do vs say .... hadoop [13:53] hadoop? never heard of it. wait. is that some jvm thing? :) [13:53] discourse is complex IMO. [13:54] yeah [13:54] it's not in the store yet though [13:54] soon [13:54] i looked at that charm. its interesting. it uses tcmalloc in ruby. nice complex stuff === marcoceppi__ is now known as marcoceppi === freeflying is now known as freeflying_away [16:40] jcastro: you were asking me if I had installed a later version of juju? [16:40] I am going to install 1.12 [16:40] yeah [16:40] ok [16:40] starting over [16:40] that's the latest stable. [16:40] so no recovery after the vmware crash? [16:40] just boom? [16:40] what do you need me to test since I'm about to destroy everythign [16:40] it's all good [16:41] I still have some learning to do w/r network architecture [16:41] w/r openstack [16:41] I got everything successfully installed incl dashboard [16:41] i just couldn't get the networks to work or a vm to spin up within openstack [16:42] I believe later versions of juju will give me more flexibility in my deployment strategy [16:42] allow me to deploy more than one charm/service on a node [16:46] yeah [16:46] we do --to now so you can specify a machine [16:46] that's really nice [16:46] will that be supported within the gui? [16:46] currently a WIP in the gui [16:46] the feature itself is newish so there's some skew there [16:46] eh..huh? [16:47] WIP? [16:47] work in progress, sorry! [16:47] lol [16:47] http://www.jorgecastro.org/2013/07/31/deploying-wordpress-to-the-cloud-with-juju/ [16:47] has --to examples [16:47] nice [16:48] while I have your attention. do you have a reference network architecture visio or document? I have seen jamepage's excellent write up, but it stops short with defining the actual network architecture used. [16:48] I mean for openstack [16:49] he references ensuring services are on particular interfaces without illustrating the reference [16:50] I found some good folsom-based examples that would likely work on the openstack site [16:50] hmmm [16:50] but it would be nice to match up the work he did in his reference architecture [16:51] for HA [16:51] that's a jamespage or adam_g question I think [16:52] yeah, I think they must be super busy or I've been pestering them too much for info :D [16:52] maybe there's something upstream on a doc page [16:52] SpamapS: ^^^ any idea? [16:52] jcastro: your daily reminder that it's WordPress [16:52] indeed [16:53] <3 [16:55] If you guys have ideas or manage to speak with Jamespage, I would love to get my hands on that info [16:55] sorry wha? [16:56] kurt_: oh you want a recommendation for the physical network to deploy Neutron on top of? [16:56] kurt_: http://docs.openstack.org/grizzly/openstack-network/admin/content/ has some things I think [16:56] SpamapS: that would be awesome [16:56] is that what jamespage used as a model for his HA openstack doc? [16:57] and yes I have been looking at that. [16:57] kurt_: http://docs.openstack.org/trunk/openstack-ops/content/ may too [16:57] my complexity comes from: I am on vmware and I am using MAAS [16:57] http://docs.openstack.org/trunk/openstack-ops/content/network_design.html [16:58] ah nuts, I don't think --to works on MAAS yet [16:58] oh seriously?? [16:58] NOOOO :) [16:59] jcastro: why wouldn't it? it worked with jitsu deploy-to [16:59] kurt_, jcastro: --to worked OK for me [16:59] it requires provider support? [16:59] SpamapS: it's maas provider specific [16:59] with MAAS [16:59] there is jamespage! [16:59] I mean, I don't know that anybody has ever assembled the 28 actual machines to do a full HA deploy without --to ;) [16:59] SpamapS, :-0 [16:59] it's a high priority bug that the core team is well aware of. :) [16:59] jcastro, is that a new bug in 1.13.2? [17:00] * kurt_ thinks RATS! [17:00] I'm sure it was working OK with I tried it the other day with 1.12.0 [17:00] jamespage: Oh really? [17:00] jcastro, I'm pretty certain yes [17:00] oh good then! [17:00] well, I can test it for you guys :) [17:01] kurt_, what was your question re network architecture? [17:01] yeah let me know [17:01] * SpamapS stuffs a sock in his mouth and backs away slowly [17:01] jcastro: --to is provider agnostic iirc [17:01] SpamapS: just trying to reel you back in sir! [17:01] jamespage: I was looking at your HA openstack reference architecture [17:01] guide, etc [17:01] kurt_, ah - my epic [17:01] can you please provide a supporting network map? :D [17:02] I have run in to all kinds of issue with the quantum stuff and being able to allocate a network/floating IP etc [17:02] and even spin up a vm [17:03] jamespage: Aha, nevermind, I am thinking of the --contraints not working on maas, not --to [17:03] jcastro: last I talked with dfc about it. --to is really dumb and just does whatever you tell it to [17:03] kurt_, so the charms implement this - http://docs.openstack.org/grizzly/openstack-network/admin/content/use_cases_tenant_router.html [17:03] and I think its my own lack of understanding holding me back and having the underlying network not really set up correctly [17:03] marcoceppi: yeah I was thinking of constraints [17:03] jcastro: ahh yeah. e need more constraints for Maas [17:03] kurt_, each tenant must create/have created for them a private network and a route that connects it to the outside world [17:03] that ends up on the quantum-gateway unit btw [17:04] then you can float ip addresses on your instances for public access. [17:04] kurt_, the quantum-gateway charm current requires a dedicated network port connected to the 'public' network to support this feature [17:04] public = internet network right? [17:04] public can be another connection to the private network in test deployments [17:04] kurt_, yep [17:05] kurt_, look - I'm time limited right now - hopefully that points you in the right direction [17:05] Ok, I have been using NAT on my MAAS region controller - I think I am maybe breaking routing with that [17:05] take a look at the quantum-tenant-net and quantum-ext-net commands that get installed on the nova-cloud-controller node [17:05] they have help :-) [17:05] jamespage: yes, its a good start. [17:06] jamespage: maybe I can put together a visio with my stuff and email to you? you could give it a quick once-over? [17:06] kurt_: make it nice so we can steal it and put it in the documentation! [17:06] lol [17:07] as long as you credit me as a contributor :D [17:07] always dude! [17:07] maybe I'll earn that t-shirt then :) [17:07] I was going to bring one for ya anyway [17:07] jcastro: what version of juju do you want me to test? [17:08] 1.12? [17:08] yes please [17:08] sure [17:08] 1.13.2 is the new dev [17:08] and I am interested in you trying that too but -stable ftw for now [17:09] ok, give me some time today. I need to thread in my "normal" work [17:09] jcastro: do you work with Maarten Ectors? [17:10] yeah [17:10] ok, he reached out to me for some feedback [17:10] I'm guessing that's fairly normal for mailing list contributors [17:12] I think it depends on what you posted [17:13] you posted that grizzly question? [17:15] juju/pkgs for the latest charm-tools right? [17:19] web-brandon: ATM yes. though that may change soon [17:25] jcastro: I think it was... [17:29] If I want to use juju with aws, does the S3 mandatory? [17:30] marcoceppi: are you moving it to its own PPA? I noticed one but it had some failed builds. === tasdomas is now known as tasdomas_afk [17:46] web-brandon: it'll eventually be going in the juju/stable ppa [17:46] X-warrior: yes [17:46] web-brandon: with a separate daily builds ppa [17:49] is it possible to choose the aws machine type? for example [17:49] micro instance? [17:51] I found it, --constraint ty [17:54] X-warrior: you'll need to set cpu-power=0 mem=128 cpu-cores=0 with constraints in order to get micro with juju. we really don't recommend micros though [17:55] marcoceppi: oh really? ok, thanks for the information [18:11] So I'm trying to bootstrap to sa-east-1 but I'm getting this error: "error: provider storage is not writable"... when I try to us-east-1 for example it created the machine. S3 says there is no region specification so I think it is ok. The only difference I could find is, ec2 from sa-east-1 has 2 machines (not created with juju) and us-east-1 has no machine. What could be wrong? [18:15] X-warrior: juju currently operates entirely within one availability zone in one region. there's no region-spanning intelligence.. [18:15] sarnold: Yeap, but does it need a clean region? [18:16] X-warrior: no idea, but seems unlikely [18:17] So I have no idea what is going on, I can bootstrap to us-east-1 easily... but when I try to bootstrap to sa-east-1 I get that error [18:17] :S [18:51] sarnold: I found the problem, the problem was the s3 name. Just updated it and it worked. :D [18:52] X-warrior: woo! :) [18:52] Is bootstrap a valid instance to deploy services? [18:52] for example, juju deploy postgresql --to 0 === defunctzombie_zz is now known as defunctzombie [18:56] X-warrior: yes, you can deploy --to the bootstrap node [18:57] so I already run juju bootstrap, but when I try to to use deploy --to 0, it returns me 'error: no instance found', checking the amazon ec2 I see the juju-amazon machine [19:01] <_mup_> Bug #1217011 was filed: error connecting to rapi ws from chrome dev channel (30)

[19:19] 2013-08-26 19:18:04 ERROR juju supercommand.go:235 command failed: cannot start bootstrap instance: cannot set up groups: cannot revoke security group: Source group ID missing. (MissingParameter) error: cannot start bootstrap instance: cannot set up groups: cannot revoke security group: Source group ID missing. (MissingParameter) [20:40] Hello, I'm trying to deploy postgresql with the support for permanent storage, after changing the config file in this way http://paste.ubuntu.com/6030130/ I get this error, which is not very clear to me: http://paste.ubuntu.com/6030125/ [20:40] any suggestion? [20:47] vds: did you get an error and then retry perhaps? [20:47] sidnei, nope [20:50] vds: uhm, try debug-hooks and then retry --resolved, see what 'config-get --format=json' returns once in the hook context. [20:50] * vds tries [21:09] anyone know when this will trickle down into juj-core, preferably the stable ppa?: https://bugs.launchpad.net/juju-core/+bug/1210328 [21:09] <_mup_> Bug #1210328: cloudinit: switch apt-add-repository to use ppa:juju/stable [21:16] kentb: it's in the devel ppa atm, stable will only be updated when 1.14 is released. [21:17] ok. thank you sidnei [21:49] sidnei, config-get is not installed on postgresql/0 [21:49] ! [21:50] vds: are you running from debug-hooks in the context of a hook? [21:50] vds: if you run in the default shell when debug-hook start up it won't work [21:50] vds: you need the 2nd shell that pops up after you do resolved --retry [21:52] sidnei, where should I run resolved --retry? [21:52] vds: from where you ran juju debug-hooks (which I assume is your local machine), on a separate terminal window of course [21:56] sidnei, config-get in the context of the hook returns {} [21:57] vds: cool. i wonder if it's a bug? the code expects it to return the default value that's set in config.yaml apparently, and afaict that has always worked. [21:58] thumper: ^ [21:58] vds: which version of juju you're using? [21:58] * thumper looks up [21:59] sidnei, thumper 1.13.2-raring-amd64 [21:59] vds: which provider (shouldn't matter just gathering info) [21:59] vds: and which hook were you debugging? [21:59] sidnei, thumper could it be the config change I did? I'm not sure that's the way to pass a dict http://paste.ubuntu.com/6030130/ that's just what I changed [22:00] thedac, openstack [22:00] thumper, openstack [22:00] thumper, from the debug log it looks like install [22:00] vds: that looks like it should work === freeflying_away is now known as freeflying