=== 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 === tasdomas_afk is now known as tasdomas === defunctzombie is now known as defunctzombie_zz [05:18] kurt_: If you're using GUI, you'll want to use cloud:precise-updates/grizzly [05:34] macroceppi: thank you. is there anything I need to do w/r to the other components or repositories and such? [05:37] marcoceppi: That failed too. :( http://pastebin.com/Lht2LULb [05:38] kurt_: and you've tried cloud:precise-grizzly ? [05:39] kurt_: the openstack charms are mad complicated to deploy [05:39] I have a deployer file I used during a demo that successfully sets up everything but swift [05:39] if you want to use it for referrence [05:40] With the demo I was able to log in to horizon and push images to glance, but networking was broken. It was enough for my demo but might help you sort out configuration details, etc [05:41] kurt_: http://paste.ubuntu.com/6001802/ [05:41] I'm traveling ATM, but I have a goal next week to get OpenStack deployed using MAAS + Juju and document the process (as long with document Juju + MAAS) [05:44] kurt_: are you deploying openstack on MAAS? [05:46] * marcoceppi boards a plane [05:47] marcoceppi: yes [05:47] with juju-gui [05:50] kurt_: using juju-gui shouldn't cause a problem here. It's method of deployment is near identical to that of the CLI. Do you get different results using the CLI? [05:50] I haven't tried via command line [05:51] kurt_: the issue I think is with the proper configuration of the charms, not nessisarily with juju-gui [05:51] kurt_: that deployer file I gave you is a YAML representation of a near working openstack deployment I did earlier [05:51] each service has the configuration options I used under it, it might be a good starting point for you [05:52] Ok [05:53] why didn't you have to enter the vip parameter for keystone? [05:53] it was considered mandatory under the gui [05:53] anyways I'm going off on tangents [05:56] marcoceppi: thanks. I'll try that [05:58] marcoceppi: and btw - yes I tried cloud:precise-grizzly in the gui === jam1 is now known as jam === defunctzombie_zz is now known as defunctzombie === freeflying is now known as freeflying_away === CyberJacob|Away is now known as CyberJacob === CyberJacob is now known as CyberJacob|Away === CyberJacob|Away is now known as CyberJacob === CyberJacob is now known as CyberJacob|Away === freeflying_away is now known as freeflying [09:20] jrwren: I was wrong it was cpu-power=0 === defunctzombie is now known as defunctzombie_zz === tasdomas is now known as tasdomas_afk === jrwren_ is now known as jrwren [14:34] I've just deployed the tomcat6 charm on a private openstack instance and port 8080 hasn't been added to the security group rules. Is this a known problem ? === freeflying is now known as freeflying_away === mthaddon is now known as mthaddon` [15:28] how could I expose the juju bootstrap server since it doesn't get assigned a service? [15:38] weblife: are you referring to your bootstrap server or your root node 0 bootstrapped node? [15:39] weblife: either way, just ensure at least one interface is on your public facing network and you are golden. [15:46] kurt_: do you mean in the etc/networks/interfaces file? [15:47] yes [15:48] weblife: are you using MAAS? [15:48] AWS ;x [15:50] Ah, yeah, you'll need to ensure whatever interfaces you want exposed are publicly routable. In MAAS, you need to use NAT to route from an external interface to an internal interface. At least that's what worked for me. === CyberJacob|Away is now known as CyberJacob [16:12] weblife: the docs are wrong [16:12] juju doesn't support instance-type [16:13] jcastro: ?? Oh lol. [16:14] jcastro: thumper and marcoceppi helped me out on that. [16:14] you need to specify it with cpu-cores and memory [16:15] jcastro: I found specifying cpu-power=0 alone works also [16:15] oooh [16:15] nice trick, I didn't know that one [16:15] can you paste me your entire line? I can put it in the docs as an example [16:16] jcastro: juju bootstrap --constraints "cpu-power=0" [16:20] ok I added that as an example [16:20] it'll be in the next doc rerun. [16:20] evilnickveitch: ^^^ [16:22] kurt_: do you understand AWS? This would mean I need to creat an Elastic IP and associate it with my instance right [16:22] jcastro, ack, am actually working on that page now [16:23] jcastro: good deal [16:33] weblife: I've not played with AWS extensively. Sorry mate. [16:37] weblife: what do you mean by expose? and why would you do that? I was able to ssh to my aws node 0 without anything special. [16:57] jrwren: I am playing around with using juju on a single node. I know it defeats the purpose, but juju also allows you to avoid that you can ssh into e need to understand ec2 cli commands. One line and you have a instance that you can ssh into and setup from there. [16:58] thats where my cursur went, lolo [16:58] bootstrap gets me exactly what you just said. [16:58] * jrwren goes and tries. === mthaddon` is now known as mthaddon [16:58] jrwren: I am playing around with using juju on a single node. I know it defeats the purpose, but juju also allows you to avoid the need to understand ec2 cli commands. One line and you have a instance that you can ssh into and setup from there. [16:59] Yeah but you cant connect though the public address it seems [16:59] i found using the boto python api to be a good way of avoiding the ec2 cli tools :) [16:59] i don't understand? of course you can connect through the public address. [16:59] that is the only way that you can connnect unless you are on another ec2 node in teh same security group [17:00] ec2-54-218-135-114.us-west-2.compute.amazonaws.com Running a node.js program [17:00] weblife: perhaps you need to deploy an ssh charm or an 'ubuntu' charm or something similar, so that you can then use 'juju expose ' and then use the public ip? [17:00] that is why i asked "what do you mean by expose?" [17:01] a node.js program doesn't get exposed. do you mean a node.js server using something like socketio? [17:01] a server, yes sorry [17:02] specifically a port 80 web server? [17:02] yep [17:02] then, what sarnold said. :) [17:03] im going to try that [17:15] It's kinda funny deploying Ubuntu charm but I see the purpose. [17:25] that is the only way AFAICT [17:36] jrwren: darn you. marking me decrypt you terminology. haha http://en.wiktionary.org/wiki/AFAICT [17:36] thank you again [17:36] i always assume LKML abreviations are allowed :) [17:40] either your saying inux kernel mailing list or lord keep me strong or your f-ing with me. haha [17:40] linux kernel mailing list [17:41] or all of the above [17:41] :) === defunctzombie_zz is now known as defunctzombie [17:48] oh wow. AWS gave me close to $200 in credits sweet I can run t1.micro instance for 14.88 a month compared to the $44 something for a m1.small. plus the 750 hours of free t1.micro. [17:57] I'm trying to solve juju-gui installation problem for nova-compute with cloud:precise-grizzly. I'm looking at the hooks and running in to 403 download error as shown here http://pastebin.ubuntu.com/6003723/ [17:57] Anyone know how to solve this? [18:01] is there a proxy in the way? [18:01] no sir [18:01] there is NAT [18:01] can you wget the file from behind the nat? [18:02] wget http://ubuntu-cloud.archive.canonical.com/ubuntu/dists/precise-updates/grizzly/main/binary-i386/Packages [18:03] jcastro: (I assume this is Jorge?) That works on the node itself without problems [18:03] yeah it's me [18:03] :) [18:03] hrmph [18:03] jamespage: or adam_g ^^^ any ideas? [18:03] I am actually trying to follow Scott's tutorial on deploying openstack with juju-gui on your blog mate [18:04] oh lol, yay good info [18:04] oh you mean the video [18:04] I've ran in to a few problems along the way, but working through them [18:04] yup [18:04] oh hold one [18:04] I have something that might help [18:04] great [18:04] FYI - this is on vmware, but that shouldn't matter [18:05] i thought juju didn't support vmware [18:05] http://pastebin.ubuntu.com/6003753/ [18:05] kurt_: I can also see about just getting you the bundle [18:05] jrwren: doing this on my own with great success [18:05] so you can just deploy it all in one go [18:05] that would be awesome! [18:06] kurt_: do you have juju code to support vmware "environments" ? [18:06] hey adam_g, do we have a bundle like that handy? [18:06] jrwren: rem…do I need that? So far things have been going ok [18:07] I think he's deploying the nodes for openstack on top of vmware [18:07] not using vmware directly with juju [18:07] nope - correct [18:07] kurt_: i don't know. i'm trying to figure out what you are doing, becuase I might like to do it too :) [18:07] I'm using vmware for MAAS, then deploying on top of that [18:07] so far it's working pretty well [18:07] I see, so MAAS to get openstack on vmware, then juju to that openstack. [18:08] yes, but also using juju-gui [18:08] kurt_: do those notes I pasted help? [18:08] I hope to have a working openstack model in the end and write it up [18:08] looking... [18:09] I can try that Jorge. Hey - one more question for you…keystone required a VIP, which it shouldn't [18:09] it should be able to deploy without that [18:09] kurt_, that 404 is strange. the archive seems to be working fine from where i sit [18:10] yeah works for me too [18:10] yes, me too adam_g [18:10] 403 [18:10] kurt_, VIP is only required if you plan on multi-unit clustering. the VIP becomes the highly-available API endpoint [18:10] as I said, it works even from the node, but not from the hook - and also when I do apt-get update [18:10] kurt_, are you hitting a proxy somewhere? [18:10] adam_g: yes I agree - but via the gui, it won't deploy without it [18:11] kurt_, huh? [18:11] it should [18:11] ->should<_ :) [18:11] I wonder if it works through the API [18:11] errr, through the CLI I mean [18:11] there's been a few cases where the GUI won't deploy something but the CLI will [18:11] I did not try that yet - marcoceppi asked me to to do the same thing [18:12] I am willing to be 5 internet dollars the CLI will work fine [18:12] Will items deployed via CLI pop up in GUI? [18:12] yep [18:12] COOL - nice feature [18:12] yeah it's nice to demo to people [18:12] have one guy driving in the background, etc. [18:12] does the GUI take into account whether a config option is required/optional? [18:13] lol, nice feature for demos yes [18:13] it was a config issue last time the GUI failed but CLI worked, I bet it's the same thing this time [18:13] rick_h: ^^^^ [18:13] * jcastro can't wait for the CLI to talk to the API natively so we can not have problems like this anymore. [18:14] how far are we off for that? :) [18:14] not within the next few weeks, heh [18:14] but maybe we can just CLI this one to get you on to the next step [18:15] yes - are we talking keystone now or nova-compute [18:15] and then when the GUI guys are around I can file a bug and they'll fix it. Last time it was a 24 hour turnaround [18:15] sweet [18:15] can we make that keystone VIP field optional? [18:15] that's an adam_g question [18:16] kurt_, in any case, the VIP config setting isn't used until you add an optional relation to an hacluster service. you can just put in any IP for the time being [18:16] ah ok - maybe just some documentation then around that adam_g? [18:17] <- the layman get's confused :) [18:17] maybe in the README for the charm? [18:17] that would be awesome [18:17] kurt_, maybe. i'm not familiar with the GUI. not sure if its a bug or a feature of the GUI's config handling [18:18] I can file the bug, which charm does it exhibit this in? keystone? [18:18] yeah, as jcastro said, it will become a non issue when gui talks directly to api [18:18] man dudes, we need to sort https://help.ubuntu.com/community/UbuntuCloudInfrastructure [18:18] yes, keystone [18:18] the GUI talks to the API directly, it's the CLI that doesn't [18:18] m_3, jcastro: who maintains jenkins.qa.ubuntu.com? she's dead. [18:18] oh :) [18:18] sorry [18:19] bac: actually, dunno... canonical IS I'd imagine [18:19] m_3: thx [18:19] bac: looks like there's discussion of it ongoing atm [18:20] m_3: great [18:20] kurt_: how do you like the GUI so far? [18:20] Love it [18:20] its a nice piece of work and definitely the right directly [18:20] direction IMHO [18:21] If I can get openstack working with it, I will be so excited. [18:22] Once I get openstack working, I'm going to look in to converting some of the apps I work with in to charms. [18:23] there's a contest dude [18:23] I would love to go back to the company I work with and show them how easy it -could be- to deploy services. [18:23] up to 10k in prizes! [18:23] https://wiki.ubuntu.com/ServerTeam/OpenStackHA will be interesting to you [18:24] kurt_: you've picked the right time to try it. We use it but we've lacked feedback from other users as to what to fix/make better. [18:24] I already had that bookmarked :D [18:24] kurt_: what command did you use to set the VIP thing? I can put it in as an example [18:24] I just added it manually [18:24] in to the gui [18:25] oh, so just in the field then [18:25] I used an unused IP address on the MAAS network. [18:25] yes [18:25] I think adam_g said its only needed once you start deploying HA services [18:25] so it can be anything [18:25] yeah I'm just trying to figure out what to put in the charm README [18:25] "VIP is only required if you plan on multi-unit clusterming. The VIP becomes a highly-available API endpoint. If you don't need it you can define an unused IP on your network." [18:25] how's that? [18:26] jcastro, you should be able to deploy without setting it [18:26] that's fine. he actually said it could be *any* IP address [18:27] I'm deploying with precise, so maybe you guys fixed it [18:27] precise is the target, so that should work, I'll go ahead and file a gui bug [18:28] jcastro: there was also a "best practices" deployment guide too [18:28] yeah so we want to do a documentation sprint soon [18:29] consolidating all these openstack docs would be a nice goal [18:29] I'm willing to give you guys feedback as I go along or assist if you need testing of things (time permitting of course) [18:30] absolutely! [18:30] I have shirts to send your team too! [18:30] my platform is vmware fusion on mac os x. [18:30] *laughs* - my team of one [18:30] just as important as a team of one million! [18:30] cheers [18:31] http://blog.xtremeghost.com/2012/11/lets-shard-something.html [18:31] FYI too ^^ [18:33] I'll check that out. Juan probably lives somewhere near me. [18:33] we hit the coast regularly, we should hang out [18:33] definitely! [18:33] I'm in the bay area [18:34] if you like beer, I find you the right places. [18:34] i can rather [18:35] * jcastro nods [18:35] burn brighter at h o t m a i l [18:35] just let me know! [18:41] https://bugs.launchpad.net/juju-gui/+bug/1214087 [18:41] <_mup_> Bug #1214087: GUI fails to deploy keystone [18:41] blam! [18:41] jcastro, should that also fail with the mock environment on jujucharms.com? [18:41] kurt_: if you have any other notes or pain points, just lmk and we'll fix them as we see 'em [18:42] Ok, is this channel best way? [18:42] sure, or mail juju@lists.ubuntu.com [18:42] or jorge@ubuntu.com [18:42] whatever works for you [18:42] very good. thanks Jorge :) [18:43] kurt_: did you have any problems getting maas up and running? [18:43] that's the one I'm real curious about [18:43] MAAS was fine. I have to manually start the nodes is the only pain. I am working with the libvirt devs to get libvirt working under mac osx so maybe I can have auto starting nodes too [18:44] when the nodes don't manually start, it does mean some fiddling with restarting and redeploying the nodes [18:44] Oh - and clock sync is a major pain - I have to go fiddle some clock sync bits [18:45] what is key is that maas has a direct connection to the internet [18:45] I used NAT on my primary clustered node to get around that [18:45] * jcastro nods [18:46] are you using the mac juju client or the ubuntu one in a vm? [18:46] ah…I have two interfaces on my primary clustered node [18:46] ubuntu one in VM, all VMs [18:46] running with Fusion on mac osx [18:47] I manually created a vmnet3 interface with the internal MAAS network [18:47] so maas could handle dhcp/dns [18:48] when you're all done you should write this up so other people can set that up [18:48] that's pretty slick [18:48] yes, I am very happy with the set up [18:48] thanks - I plan to [18:48] as long as I can get all the way through the openstack stuff :D [18:49] Jorge - what part of the world are you in if I may ask? [18:49] Ann Arbor, MI [18:49] hey so, did the deploy keep going or is it stuck somewhere else? [18:49] I haven't had a chance to check yet [18:49] deploy got stuck on repositories…haven't fixed that yet [18:50] ah [18:50] the 403 [18:50] apt-get update gets same errors [18:50] yes [18:50] maybe its a problem with the precise image in maas [18:52] yeah we might need to post to the list for that one [18:52] kurt_, are you hitting an apt proxy somewhere on the MAAS node? [18:52] i suspect you might be, and its denying access to the cloud archive [18:52] adam_g: no proxy, though it does go through NAT [18:53] doesn't MAAS configure an apt-proxy and provision new nodes to use it, by default? [18:53] yeah but how does it work when he does it directly on the node but not from the hook? [18:53] if it was a proxy that wouldn't work either right? [18:53] oh, didnt know it worked in some cases. [18:54] 'apt-get update' fails in the hook, but succeeds manually from the same node? [18:54] does it help that I get the same results when I do apt-get update from the node? [18:54] no [18:54] both fail [18:54] ok [18:54] oh, I am confused, I thought you said it worked on the node, my bad [18:54] apt-get update exhibits same 403 errors [18:55] http://pastebin.ubuntu.com/6003921/ [18:56] apparently the wget failed too, but then resolved? [18:56] kurt_, sudo grep "Acquire::HTTP::Proxy" /etc/apt/* -R [18:56] ahh - bad command line [18:57] wget - http://pastebin.ubuntu.com/6003924/ [18:58] adam_g: nope http://pastebin.ubuntu.com/6003928/ [18:59] same results on maas clustered node FYI [19:00] m_3: pavel would like to sync up on the rack charm [19:00] m_3: he's back from holiday [19:07] kurt_, what about: sudo grep ProxyAutoDetect /etc/apt/* -R [19:07] kurt_: did the keystone VIP thing return an error or did it just not work? [19:08] in the gui I mean [19:08] adam_g: same results - nothing [19:08] kurt_, hmm [19:09] jcastro: I believe I saw the error in the debug-log [19:09] let me look through archives [19:09] kurt_: if you have it that would help [19:12] 2013-08-18 14:16:11,995 unit:keystone/3: unit.hook.api INFO: FATAL ERROR: ERROR: Config option has no paramter: vip [19:15] jcastro: is that sufficient or would you like a more complete log dump in pastebin? [19:17] pastebin wouldn't hurt, just in case [19:18] http://pastebin.ubuntu.com/6003993/ [19:20] jcastro: yeah, he pinged me earlier then disconnected [19:20] available whenever [19:22] m_3: can you just coordinate with him via mail? [19:24] jcastro: yeah === jcastro changed the topic of #juju to: Share your infrastructure, win a prize: https://juju.ubuntu.com/charm-championship/ || OSX users: We're in homebrew! || Review Calendar: http://goo.gl/uK9HD || Review Queue: http://manage.jujucharms.com/review-queue || http://jujucharms.com || Reviewer: ~charmers" [19:25] we're in homebrew! [19:27] whats the link for charms school now? Or they only on your youtube jcastro? [19:28] whoohoo! [19:28] jcastro: is that `brew install go && go get ...` or just `brew install juju`? [19:29] m_3: you get my response email? [19:29] kurt_, what juju version are you using? [19:30] juju 0.7 [19:30] weblife: not sure... lemme look [19:30] weblife: https://juju.ubuntu.com/resources/videos/ [19:30] has a link to all the charm school videos [19:30] m_3: should be brew install juju [19:31] jrwren: do you use homebrew? [19:31] yes [19:31] wanna give it a shot? [19:31] totally sweet that he wrote that. [19:31] I tried and said "how the hell do you do this with go?" [19:31] does not work [19:31] he must not have tested it. [19:31] wait a sec... that is my failed attempt [19:33] weblife: yes, now I did :) [19:33] * m_3 needs to relax the bug email filters a bit [19:34] m_3: just wanted to be sure I responded from my blackberry. Wasn't sure how launchpad would treat it without the encryption tool. [19:34] jcastro: works great. [19:34] weblife: would using express as an example app have enough there to go end-to-end? [19:34] i.e., to curl the server and verify that all's good? [19:34] jrwren: _awesome_ [19:34] very fast to fetch and build and install go 1.1.2 and juju 1.12 [19:34] yeah [19:34] I left your manual instructions there too [19:34] weblife: or would it need an actual application on top? [19:38] interesting, amazon.com goes down but ec2 seems fine [19:38] m_3: it wouldn't demonstrate the mongo config but you would see a page with header "Express" and content "Welcome to express" . [19:39] ah, ok, well maybe us-east-1 is having some problems [19:39] weblife: perfect... great idea [19:39] mongo isn't necessary for a "default" install [19:39] m_3: possible issues are it launches on port 3000 and creates the server file app.js not server.js. But should make the app vs server name convention not matter anyways. [19:39] but something coming up on the webpage is [19:40] weblife: as long as somebody can just `juju deploy node-app` without config... then `juju expose node-app` and hit the page [19:40] m_3: its pasicaly a apache "It Works!!!" page [19:40] basically [19:40] weblife: we could default it to 80 instead of 3000 [19:41] weblife: but whatever makes sense for default behavior [19:41] the 3000 default is a necessary thing to do in the world of services sharing containers [19:41] not so much here [19:42] at this point we should just do what people would expect [19:42] jcastro: I was about to say kill them, but I guess not everyone uses homebrew. [19:42] weblife: oh, and app.js -vs- server.js is fine [19:42] is the sharsum a good enough check for verification of the upstream? [19:42] weblife: that's somewhat of a mess... perhaps should be a config parameter itself [19:43] not a bad idea [19:43] weblife: ok defaulting it to app.js... that was totally just _me_ picking config out of thin air [19:43] weblife: also we can change 'config/config.js' to just 'config.js' if that's more common... dunno [19:44] no I think thats good. Shold always seperate modules in their own folder [19:44] I'm fine erring on the side of an established framework like express [19:46] jrwren: thought about killing them, but I think it's handy for people to build it themselves too [19:47] m_3: I'm going to get going on this today. (After I go though those charm school videos) Kinda want to make sure I am clear on charm caveats [19:47] weblife: awesome [19:47] * m_3 hopes you don't fall asleep during those _absolutely riveting_ videos :) [19:48] jcastro: thank you [19:48] m_3: probably will once or twice. But I have all day to myself. [19:49] sitting on 3 cups of joe right now though [19:49] :) === tasdomas_afk is now known as tasdomas [19:53] jcastro: is charismatic on screen; haha [19:54] are you close to Ohio? I'm speaking about Juju there next month! [20:04] California [20:56] adam_g: any more thoughts on the 403 error? [20:57] kurt_, no idea, other than a proxy somewhere 403'ing the request. [20:58] hrm…trying to think about how to trace this and figure this out === marcoceppi__ is now known as marcoceppi [21:29] kurt_: howdy! so you are having issues again with the deployment? [21:29] what issues are you seeing now? [21:30] getting 403s roaksoax [21:30] I'm trying to sort it myself [21:30] kurt_: when exactly? [21:30] kurt_: when do you get those 403s? apt-update? [21:30] 2013-08-19 10:50:33,321 unit:nova-compute/7: hook.output INFO: Failed to fetch http://ubuntu-cloud.archive.canonical.com/ubuntu/dists/precise-updates/grizzly/main/binary-i386/Packages 403 Forbidden [21:30] yes [21:30] in both the nova-compute fetch and on apt-get update [21:31] kurt_: try adding ubuntu-cloud.archive.canonical.com to /etc/squid-deb-proxy/mirror-dstdomain.acl.d/99-maas [21:31] and then restart squid-deb-proxy [21:31] (on the MAAS node) [21:31] kurt_: yes, on the MAAS node ^^ [21:31] I'm not running squid - is that a part of the MAAS infall? [21:31] kurt_: yes, it is part of MAAS install [21:32] interesting, ok :) [21:34] nice, that fixed apt-get update :) [21:35] cool :) [21:35] watching the debug log with baited breath... [21:36] AND WE'VE GOT NOVA-COMPUTE :D [21:37] wahoo! [21:37] that was driving me nuts! thanks roaksoax [21:37] kurt_, i knew it was a proxy! [21:37] kurt_: np :) [21:37] we need that documented or fixed :) [21:38] that config needs to be added to the cloud-init? [21:38] kurt_: yeah please file a bug against maas saying "cloud archive is not allowed by default" for maas please :) [21:38] kurt_: that file I made you modify comes from maas rather than cloud-init [21:38] adam_g: yes - thank you for your help too. I was looking at my firewall [21:39] roaksoax: file you made me? :) [21:39] kurt_: the file i made you modify comes from maas yes [21:39] ahhhh, gotcha === tasdomas is now known as tasdomas_afk [21:40] so let me ask this - what other repositories should make their way to this file? [21:40] or hosts rather I guess [21:40]