[00:56] jose: still kicking around? [01:00] lazyPower: yep! [01:00] lazyPower: I'm testing the newly-fixed charm :D [01:01] jose: right on! Did you still need to PM me? [01:01] or are you unblocked? [01:02] I think I'm good to go by now [01:02] allright cool. I didn't want to leave you hanging [01:02] thanks! [01:05] Thats how we roll jose. Charmers 4 lyfe [01:05] :P === freeflying__ is now known as freeflying [01:15] jose: good work on the queue too man, you're killin it with the quality MP's [01:15] thanks, just trying to :) [01:15] let me know if any of those are buggy, maybe a couple will be [01:16] We'll have feedback on improvements if they are :) === sputnik1_ is now known as sputnik13 [01:59] lazyPower: still around? [01:59] jose: indeed. Whats up? [01:59] I'm having some problems printing a variable [01:59] pastebin? [01:59] sure, second [02:00] lazyPower: http://bazaar.launchpad.net/~jose/charms/precise/mailman/trunk/revision/32 [02:00] those exact lines in the diff [02:00] they're printing literally "$DOMAIN" instead of the domain [02:03] jose: http://paste.ubuntu.com/7160367/ [02:04] there's an illustration of whats going on. What you're seeing is a symptom of how single quotes work in bash [02:04] they are string literal annotations. [02:04] hmm, got it! [02:04] thanks [02:04] w0ot, manually fixing it makes it work! [02:06] jose: you have single quotes surrounding those blocks, thats why you're getting the "$DOMAIN" printed. I dont understand what the 'manual fix' is - but happy its working :) [02:06] yep, testing another deployment now! === Ursinha is now known as Ursinha-afk === Ursinha-afk is now known as Ursinha === vladk|offline is now known as vladk === sputnik1_ is now known as sputnik13net [06:50] arosales: thanks for the heads up! :) [06:51] jose: oh, np. Thanks for your contributions. I see you have been busy :-) [06:51] good stuff. [06:51] well, I thought juju could use some of my time :) [06:51] I too say that issue with Bip and we were wondering why proof was having an isue with a 2 character yaml key [06:51] s/say/saw/ [06:52] Marco mentioned it was because of a yaml issue, it needs to be 3chars min [06:52] jose: well the help with improving the overally charm quality is very appreciated. [06:52] \o/ [06:53] ya we didn't find anything in the yaml spec, but I too saw when I changed ip to any char with 3 char or greater proof was happy [06:53] :P [06:53] jose: We are in the midst of a charm audit so it was nice to see your merge requests [06:54] I just need to learn how to write amulet tests and I'll give a hand with tests [06:54] maybe it can be considered for a charm school? 'Writing amulet tests for your charms' [06:56] jose, for sure a good topic we should revist. We have an initial one @ https://juju.ubuntu.com/resources/videos/ [06:56] http://www.youtube.com/watch?v=KlTeRjGv4E4 [06:57] oh, missed it as it wasn't at ubuntuonair, not subscribed to Jorge :P [06:57] anyways, I'll take a look at that one and see what can I do [06:58] we trying to make the video page complete in case someone does miss a charm school or just joins and wants to catch up. [06:58] jose: docs are also at https://juju.ubuntu.com/docs/tools-amulet.html [06:58] seems easy enough [06:59] if you pull the source for charms such as rabbitmq, haproxy, or memcached you'll see some code examples in /test of amulet based tests [07:00] jose: our of curosity were you following jcastro's charm audit email, https://lists.ubuntu.com/archives/juju/2013-December/003331.html ? [07:01] arosales: not actually, I've been mostly involved with juju this last two or three weeks [07:01] ah ok well your timing on fixing charms is perfect :-) [07:01] cool! [07:02] I'm also trying to convert most (if not all) READMEs to Markdown [07:03] +1 [07:03] and I see you are using the teamplate from charm tools which is nice [07:04] makes life easier [07:08] for sure [07:11] well, time for me to go to bed, laters! === Kyle is now known as Eliz [07:13] good night jose, good to talk to virtually here === CyberJacob|Away is now known as CyberJacob === vladk is now known as vladk|offline === vladk|offline is now known as vladk [08:50] gnuoy, https://code.launchpad.net/~openstack-charmers/charm-helpers/ssl-everywhere/+merge/209301 [08:50] looking [11:44] marcoceppi: hey! is the minecraft server charm still good? [11:52] jose: not really. it works but the version mineshaft is old [11:53] hmm, ok [11:53] I'm thinking on integrating some basic tests on my pagekite charm before it gets to the store [11:54] that charm school is really useful in that matter :) === marcoceppi is now known as marco_traveling [11:56] marco_traveling: have a safe travel! [12:00] uh, guys, I have a problem here... I am trying to write tests for pagekite but a config option would be the kite's secret [12:38] jamespage, i added testing for databases to my MP also [12:38] MPs === roadmr_afk is now known as roadmr [13:36] man, content-fetcher looks like a great idea === vladk is now known as vladk|offline [13:46] jose: read that from an external source [13:46] jose: scratch that idea [13:46] i dont know what i was thinking.. need coffee === roadmr_afk is now known as roadmr === roadmr is now known as roadmr_afk === roadmr_afk is now known as roadmr === benji_ is now known as benji [14:14] jcastro: would be good to have a mini blog post about content-fetcher I think [14:14] nod [14:17] where are charm's hook files located on a deployed node? === vladk|offline is now known as vladk [14:24] also, is this a bug or something i'm doing wrong: [14:24] tvansteenburgh@trusty-vm:~/src/charms/precise⟫ juju debug-log [14:24] ssh: connect to host localhost port 22: Connection refused [14:24] ERROR rc: 255 [14:26] and which bug list would i consult to see if it's a reported bug? [14:26] tvansteenburgh: is this in an lxc environment? debug-log doesn't work in lxc if I recall [14:26] rick_h_: ok, yeah, it is [14:26] tvansteenburgh: ok, that's a known issue then. Looking for the hook path [14:27] ok, hook path i can figure out, was just being lazy, hoping someone knew offhand [14:27] tv so checking real quick I find it in /var/lib/juju/agents/unit-juju-gui-0/charm/hooks [14:27] :) [14:27] all good [14:28] cool, thanks [14:51] ERROR cannot use 37017 as state port, already in use [14:51] How do we resolve that error with the local provider? ^^ [14:52] jcastro: ps aux | grep mongo [14:52] kill pid [14:52] man, still? [14:53] not sure if there's a different thing but it's what I know/use [14:53] normally it's something went boom [14:53] yeah [14:53] something went bada-boom [15:11] big bada-boom [15:12] <3 fifth element, just saying === Ursinha is now known as Ursinha-afk [15:26] marco_traveling: delete charm/bundle landed and removed those vms ones. [15:30] jcastro: we can garden out the old bundles with poor names now. let me know when you want to check it out [15:34] rick_h_, busy with a mongo cluster ATM, just clean em! === Ursinha-afk is now known as Ursinha [15:54] rbasak, rick_h_: quickstart 1.3.0 with the --distro-only flag just released on PyPI [15:55] frankban: thanks! Please could you update bug 1282630 to match? Then we can change the FFe and MIR bugs. [15:55] s/change/chase === sputnik1_ is now known as sputnik13net [15:57] rbasak: done [15:58] frankban: thanks! I'll chase in #ubuntu-devel. [15:58] rbasak: thank you! [16:02] frankban: is the goal to get juju-quickstart on the desktop CD, on the server CD, or both? [16:03] rbasak: I guess both, rick_h_? ^^^ [16:04] I never considered this aspect before. We'll need to check with the release team to do this at this late stage, assuming the MIR is accepted. [16:04] I'm not sure if it needs an FFe or not, and we'll miss beta2. [16:04] (and I hope there's space, etc) [16:04] jamespage: ^^ any advice here, please? [16:05] jamespage: I presume this is higher priority now that juju-core looks unlikely for main? [16:06] lazyPower: it'd still be public, yeah. any other ideas on how to do that? [16:06] otherwise, a test wouldn't be possible unless some values are set manually [16:06] rbasak, frankban: I think that's out of scope for 14.04 now [16:06] if we don't mir juju-core, we won't do the mir for quickstart [16:07] jose: reach out to the support dept/mailing list for the service and see if there are testing credentials that can be used that dont "actually" work, but will allow you to mock the login [16:07] jamespage: that confuses me a little. My understanding is that we'll want juju-quickstart more in this case? [16:07] rick_h_: ^^ [16:08] rbasak, surely juju-quickstart depends on juju-core? [16:08] jamespage: no - it installs juju-core from a PPA (by default). On discussion from the MIR, it can also install from the archive only as an option. [16:08] lazyPower: hmm, the point is not about the login, but checking that the tunnel works... that would need some valid credentials [16:08] jamespage: the underlying idea for juju-quickstart is that it's cross-distro multiplatform etc - eg. from PyPI. [16:09] jamespage: there's some talk around this subject in the MIR bug. [16:09] rbasak, yes - but not for the distro packages which should IMHO always install from the distro [16:09] jose: no idea bud, i thought you were doing unit tests, not amulet tests [16:09] yep, I started those :) [16:09] theory states you could have a config option for them, that the user would have to input, but they would fail consistently in CI [16:09] jamespage: please could you comment on the MIR bug in this? I did raise this point, so the option to only install from the distro was added. But I didn't ask for any particular default. [16:10] jamespage: I think it makes sense for the user to at least be able to ask for a PPA installation as an option, even if it is decided that from distro is default. [16:10] jamespage: in this case, juju-quickstart is still useful even without juju-core in main. [16:10] jose: let me ping marco about this when you get back, as I have a PAAS charm in the store that needs amulet tests too, that require sensitive credentials. [16:10] s/you/he/ [16:11] rbasak, agree that having ppa as an option is fine [16:11] lazyPower: awesome, thanks! [16:11] rbasak, but I'm not sure I see the value of including this in main without juju-core [16:11] 'use this fully supported tool to install a tool that's not fully supported' [16:12] jamespage: I think the goal is to have it on the CD, and that was the driver for main inclusion, rather than support. [16:12] rick_h_, frankban: ^^ [16:12] rbasak, I see juju-core and quickstart tied tbh [16:12] jamespage: I guess there's a mismatch of expecations here, so we should talk soon. [16:13] jamespage: I understand what you're saying - it's just that rick_h_ seemed quite keen on having juju-quickstart on the CD anyway, even without juju-core in main. [16:13] rbasak, jamespage: the goal is to gice to users/developers the ability to go from zero to a fully deployed environment (including the GUI and bundles support) in a single command [16:13] give even [16:13] frankban: you can more or less get that from universe though. Two commands - apt-get install juju-quickstart, and juju-quickstart. Right? [16:14] frankban: so I think what jamespage is asking is what is the additional benefit of this being in main/on the CD? [16:14] rbasak, frankban: I don't see this as any different to the objections being made on the juju-core MIR re tools coming from simplestreams [16:14] (if juju-core is not) [16:14] jamespage: we agree that PPA is OK to be an option, so what's left is what should be the default. [16:14] Oh, I see. [16:15] With juju-quickstart in main and juju-core in universe, there's a dependency from main to universe, even if it's not expressly stated in metadata. [16:15] I hadn't considered that. [16:16] jamespage, rbasak: IC, could you please add a comment on that regard on the bug? [16:16] rbasak, you got it [16:16] frankban, will do [16:16] thanks [16:23] jamespage: rbasak sorry catching up, was in a meeting [16:23] jamespage: rbasak the goal for quickstart in 14.04 is not on any cd. [16:23] jamespage: rbasak It's a post 14.04 goal, so we want to keep it in mind [16:24] jamespage: rbasak I'd say first priority would be desktop cd vs server cd [16:24] jamespage: rbasak but I'm assuming on that one. I think it fits that use case more and it's part of our argument on why it's ok to use a ppa for getting juju [16:25] jamespage: rbasak and yes, the goal of quickstart is that something small is in main that users can use to get juju-core and juju-local even though it's not in main [16:25] rick_h_, tbh I think that breaks the rules [16:25] jamespage: rbasak so with core not in main it's even more useful to end juju users to have that available to them [16:25] jamespage: only sort of, though. Eg. "pip install " is permitted in main. [16:25] (I assume it is) [16:26] jamespage: understand, that's what I'm trying to learn/get a grip on. I'm not sure on the rules. We've got a vision for a tool for end users to be able to get started with juju as easy as possible and we're working on what we can do to deliver that [16:26] As is wget, to take it to the extreme. I'd like to identify some kind of underlying principle to this kind of thing. [16:27] jamespage: rbasak the more steps we can automate and move away from the user from a fresh ubuntu install the happier everyone gets as it eases juju adoption, which is quickstart's #1 goal [16:28] The actual rule is from here: http://people.canonical.com/~cjwatson/ubuntu-policy/policy.html/ch-archive.html#s-main [16:28] looking [16:28] "In addition, the packages in main...must not require a package outside of main for compilation or execution (thus, the package must not declare a "Depends", "Recommends", or "Build-Depends" relationship on a non-main package)" [16:28] Now, what's the spirit of that, as well as the wording? [16:28] rick_h_, rbasak: I think my argument is about how this adds value by being in main IF a) we don't want to get on the ISO and b) juju-core is not in main (think Recommends juju-core>juju-quickstart) [16:29] juju-quickstart in main is useless without _some_ external source (eg. universe or PPA). Perhaps that's a factor. [16:29] jamespage: so if we can get it on the cd then awesome, but we set expectations it might take more time to make that happen. [16:29] OTOH, so is "pip install". [16:29] jamespage: so if we can get it then let's assume the value add is there [16:30] jamespage: the end value is that "user installs ubuntu, wants to try out juju, juju-quickstart will get juju, configure a lxc environment, bootstrap, and bring up the gui in one single built in command [16:30] jamespage: I don't hold Recommends juju-core>juju-quickstart as highly valuable. We should have it, but I imagine that docs could tell people to install juju-quickstart and go from there. [16:30] rick_h_, I'm not comfortable proposing this for the ISO unless we also have juju-core in main [16:31] jamespage: ok, so you'd suggest we get quickstart into main as step #1 with juju in universe. Next release we work on getting juju into main and quickstart on the iso? [16:31] jamespage: I'm fine staging it out, but want to make sure we konw what we need to do to make the end vision happen [16:31] rick_h_, no [16:32] rick_h_, I think the policy that rbasak highlighted above applies even if the relationship is not declared in the package dependencies [16:32] juju-quickstart depends on either universe or a PPA to function [16:32] main can only depend on main [16:33] jamespage: ok, so we are in universe while juju is in univsere. When they get into main we can then also go for main & on the iso at that time? [16:33] rick_h_, which is why I tie quickstart and juju-core together for main inclusion [16:33] rick_h_, sure [16:34] jamespage: what about stuff on the iso? Normally all dependencies are included, but they wouldn't be in this case. [16:34] jamespage: ok then. I can tell that story fine. I just need to know that we're on the right path per what we're supposed to be getting to [16:34] jamespage: do you think the same bundling rule should apply to the iso? ie. require juju-core to be on it if juju-quickstart will be there? [16:34] rbasak, I'm not sure tbh [16:35] jamespage: I understand your point of view, but I'm not completely convinced about your position myself. I can see both sides. It's very fuzzy. [16:35] It comes down to how to interpret the rule, and this is definitely a grey area. [16:35] rbasak, good job its not actually up to us then :-) [16:36] jamespage: perhaps we should open this up for wider discussion? ubuntu-devel ML? [16:36] rbasak, I think the bug is fine === vladk is now known as vladk|offline [16:40] jamespage: rbasak ok, so 1) we'll have quickstart and juju-core in universe and quickstart's --distro-only flag will be ok for 14.04 [16:40] jamespage: rbasak 2) Discussion is required to think on how/if quickstart can be in the iso once it and juju-core get into main in a subsequent release [16:44] * rick_h_ goes for lunch afk [16:49] lazyPower: btw, the charm is ready for review whenever you want [17:05] jose: Thanks. We're a bit slammed at the moment with stuff, we'll try to get the community charms out asap. [17:05] thanks for the heads up [17:05] no worries [17:05] and no hurries too === vladk|offline is now known as vladk [17:19] hmm, is there anyway to assign a (fake?) public ip to a unit using local provider? From juju's pov, I mean [17:20] some thing that will be returned from unit-get public-address === vladk is now known as vladk|offline [17:58] rick_h_: ta! === marco_traveling is now known as marcoceppi [18:11] hazmat, any idea why -deployer would bail if you modify a charm locally? [18:15] jcastro, see flags [18:16] jcastro, its a safety valve with a cli flag [18:16] jcastro, it should default to allowing local mods [18:17] jcastro, there's a -L flag to disallow local mods afaicr [18:23] hazmat is there an opposite to the -L flag? [18:23] We want to allow local mods [18:25] mbruzek, jcastro there's a bug that this is non obvious .. but try with and with out the -L flag ... one should work/toggle the behavior [18:25] yeah I'll have to circle back [18:25] the mysql charm is broken with master-slave for us anyway [18:37] marcoceppi as a heads up, we have gotten our storage proposal charm branches in order, reviewed and set to Fix Committed for block-storage-broker and storage subordinate charms. === vladk|offline is now known as vladk [19:29] hi guys [19:29] im new to juju [19:29] how do I start [19:38] trying to figure out how this all works, so I start a vm and install ubuntu server on it then install juju and then use juju to deploy a service I want on that specific server? === Ursinha is now known as Ursinha-afk === Ursinha-afk is now known as Ursinha [20:12] Valduare: hi, i'm new here too [20:12] hi [20:13] https://juju.ubuntu.com/docs/ is a great place to start [20:14] the machine you install juju on does not get services deployed to it. it's used to send commands to the environment that juju is controlling [20:17] ah [20:18] currently I have a freenas box serving storage through iscsi to an esxi host box that runs 14 vm’s [20:24] Valduare: using the manual provider you could use juju to power your workloads on that esxi host [20:24] create the vms manually, register them in juju, then juju deploy --to [20:25] was just reading this http://insights.ubuntu.com/resources/tutorials/interested-in-maas-and-juju-heres-how-to-try-it-in-a-vm/ [20:25] looks like i could possibly make this all virtualized? [20:26] IMHO, remove the complexity of having maas as your provisioner and just use the manual provider [20:26] while maas is cool, its accomplishing the same thing youa re doing with ESXI, by registering the hosts yourself to juju [20:27] gives an interface for rebooting vm’s and such too though? [20:27] actually, juju doesn't do machine management like that. Its pretty focused on service orchestration, so it sits a layer above configuration management. [20:28] in terms of the tech stack. [20:28] I mean maas would [20:29] I'm not nearly as familiar with maas as I am with juju. Perhaps someone else with a bit more experience would be able to answer that. [20:29] sarnold: ^ [20:31] so on the juju subject, it lets you specify what machine to provision such and such service [20:31] Valduare: the intention behind maas is to turn bare metal into a cloud, so theres a domain controller to handle the provisioning via cloudinit so you get bare systems with an image, similar to what you would get by booting the ubuntu AMI's on aws. [20:31] ie mysql [20:31] Correct, you pick a service from teh charm store, or charm up your existing app, and define what the relationships require to operate, such as a database name, host, user/pass combo, and juju handles setting up the services [20:32] eg: mediawiki has a sql dependency. You deploy mediawiki, teh charm does nothing, deploy mysql and relate it to mediawiki and whammo, you've got a functioning mediawiki installation being managed by juju [20:32] ready to scale up/down, and be configured via the juju settings. [20:32] and if you're using the gui, you have a nice network map to visualize your deployments [20:33] so since I just have the one host right now [20:33] I should deploy a juju bootstrap vm [20:33] on there and then start migrating all my services over to vm’s provisioned by juju [20:35] if you use the manual provider, the provisioning would be manual. [20:35] juju would handle the configuration mangement and orchestration [20:35] ie: how a service should be deployed, and how relationships are handled with the service. === vladk is now known as vladk|offline [20:37] does it launch a new vm for each service? [20:37] ie mysql in its own ubuntu vm [20:39] It depends on your provider. Thats why i've been specific with terminology about the manual provider [20:39] as it implies, all provisioning is done manually with the manual provider [20:39] if you're using a maas provider, it pulls from the allocation pool in maas. if you're using a cloud provider like aws, hpcloud, or azure it performs the api calls to provision a new vm for you, so therefore yes it does it for you. [20:40] so with the manual provider i’d specify I want mysql on such and such vm [20:40] correct [20:41] and if I were using openstack? [20:41] does it separate all services out to their own vm [20:42] openstack communicates to nova to spin up your machines for you. its got a provisioning api [20:42] sorry, juju communicates to nova to spin up your machines. [20:42] i mistyped. [20:44] there are a ton of layers to do all this lol [20:45] It can be, however juju can spin up an openstack cluster in 25 minutes, i think we've benchmarked even shorter times [20:45] and hten, you're free to create a juju environment on top of that openstack cluster that juju deployed. Using juju to deploy juju. Juju ception? === vladk|offline is now known as vladk === vladk is now known as vladk|offline [20:48] haha, jujuception :) [20:48] lol [20:49] how long would that take on amd e-350 processors :P [20:50] I haven't run any benchmarks on that, so good question [20:50] i actually dont have experience with those processors either [20:52] amd version of atom processor basically [20:53] ah, well you wont be breaking any earth shattering speed records [20:53] but I would think it would do fine [21:08] so juju sets up a full openstack system with storage and everything? [21:20] hmm im not seeing a charm for nginx? [21:20] Valduare: see bug #994699 [21:21] <_mup_> Bug #994699: Charm Needed: Nginx [21:22] hmm [21:24] Valduare: thats been a very back and forth topic in the past, if we want to charm up nginx as a service and deploy content as subordinates or just include nginx in your service dpeloyment [21:24] thus far, including nginx as part of the service deployment has been the winner, but there are some new charms coming in that set a web host with subordinate based content deployments. [21:25] however, nothing thats landed in the store yet [21:26] what kind of things are you using charms to deploy? [21:26] im searching through the charms now and … there’s not much? [21:26] Valduare: what are you expecting to see thats not there? we have 158 services to date of charms promulgated [21:28] things like gitlab, canvas-lms, openerp, nginx-php-fpm-mysql, [21:28] thare are give or take ~ 70 more in personal namespaces that are not in the store, as they dont follow the store guidelines, or are not ready for public consumption. [21:28] Valduare: gitlab is in there, openerp is in there [21:28] just searched for gitlab 0 came up [21:28] ah now it shows heh [21:28] i've actuallly got a prototype gitlab charm [21:28] er, gitlab-ci charm [21:29] its not ready for promulgation by any stretch, but ready for people to stress test and provide feedback [21:30] and there's an openrp bundle sitting int he review q waiting to be analized and acked for the store so its drag and drop deployment. [21:30] whats the bundle do differently than the openerp charms already there? [21:32] the bundle is a deployment schema that has all the services required to utilize openERP in a client/server model [21:32] where the backend is split from the front end, ready to be scaled individually depending on your needs [21:32] Valduare: https://juju.ubuntu.com/docs/charms-bundles.html [21:40] and it’d install it all on one vm? [21:42] however many VMs are necessary.. [21:43] what specifies the necessity [21:44] Gators#1 [21:44] gah. and now you all no the password to my vm [21:44] s/no/know/ [21:45] if i had a nickel for every time... [21:45] if you hadnt mentioned…. we’d all think you were rooting for a highschool basketball team or something... [21:48] well, not high school: http://espn.go.com/mens-college-basketball/team/_/id/57/florida-gators [22:01] lazyPower: I was just reading about juju some more :P they were talking about wordpress juju add-unit wordpress does juju have some sort of load balancing system in place? [22:01] you can deploy haproxy infront of it [22:03] this is stuff i’ve never gotten into before.. :) [22:03] .. and it all seems too good to be true when you just juju add-unit -n 10 wordpress or whatever AND IT JUST HAPPENS [22:03] is there a way I could setup a remote server and haproxy between the two locations so if one goes down ie internet outage it’d serve from the other location>? [22:03] scary :) [22:08] Valduare: actually, with manual provider, you can do cross datacenter orchestration like that [22:09] and haproxy already takes care of the load balancing in the event of an outage [22:09] but your latency will be a factor in that style of deployment [22:10] tvansteenburgh: whats your IP again? [22:10] ;) [22:10] lol [22:11] i sitll dont quite understand the logic behind the ha stuff [22:12] things like a php process pegging out the cpu I could understand but internet outages that’d knock your ha proxy out too === hatch__ is now known as hatch [22:15] Valduare: normally something like haproxy would live in front of a dozen or two dozen machines in the same rack.. [22:15] Valduare: if you lose networking to the datacenter, so be it ;) hopefully your application has been designed to failover to another availability zone [22:15] where would that be designed into? :P [22:16] im only familar with bare metal hosting lol [22:18] Valduare: it could be done at DNS level, you could do round-robin dns with haproxy IPs, one per data center.. [22:18] Valduare: or you could use IP Anycast if you're rich and powerful enough :) [22:19] sarnold: I couldn't have put it any better myself. *hat tip* [22:19] * sarnold bows [22:19] lazyPower: thanks :) did I overlook another option? [22:19] sarnold: i'm giving a talk at CMU next friday and I need to get this abstract over to the orchestrator. Would you mind giving it a proof for me? [22:19] it feels like there's got to be at least one more option. hehe. [22:19] lazyPower: sure [22:20] sarnold: there are others, but round robin/latency based dns was my poison of choice. [22:20] it worked at the entry point of the stack, vs complex configuration over convention [22:20] and im a rails guy, so i'll always pick convention over configuration, any day of the week [22:21] lazyPower: yeah, load-balanced DNS has worked awesome for OFTC for years; the servers periodically report their user load and above a certain number the server is pulled, below a certain number it's re-added, and when machines are going to be serviced, they can get taken out entirely two weeks ahead of time to let people gracefully age off the thing [22:22] but "works for an irc network" vs "works for netflix" are two different things ;) hehe [22:24] right now I got a djbdns vm running on my one mini itx server host :P [22:24] godaddy points to that for the name server of all my domains [22:24] djb <3 [22:25] coudlnt figure out bind for the life of me [22:27] bind is cool, until it stops being cool [22:27] then its a giant tree made of hate [22:28] djbdns has been pretty darn good to me so far [22:29] given all the wonderful alternatives, you'd have to pay me quite a lot to get me to use bind :) heh [22:29] so I basically need to get a dns outside of my “data center” eh? [22:30] heh I'm surprised you were able to convince godaddy to let you use only a single dns server :) [22:30] I thought 'two servers in glue' was the requirement [22:31] thats easy enough of masquerade [22:31] s/of/to [22:31] hahah [22:31] evil :) [22:35] you just put the same ip in both slots on godaddy === CyberJacob is now known as CyberJacob|Away [23:23] sarnold: so with the wordpress example, and haproxy, how does it sync data [23:24] Valduare: I've got no idea how you'd reasonably scale wordpress to multiple availability zones / datacenters ... [23:24] http://www.jorgecastro.org/2013/07/31/deploying-wordpress-to-the-cloud-with-juju/ [23:24] Valduare: mysql data dumps on cold backups would be my first attempt if I didn't find something better from an expert :) [23:29] valduare read up on MySQL replication with master slave relations [23:30] with manual provider you deploy to multip!e azs or dcs and apply the theory we have already covered [23:31] how would all the /var/www/sitedir sync? [23:31] how do you deploy to multiple servers [23:32] rsync from whatever host is used for authoring? [23:40] and whatever host used for authoring is whatever host the haproxy connects you to? [23:41] bear with me :P lol [23:41] Valduare: It would be your laptop or something? :) I dunno how these things actually get -used- :) [23:42] oh think I see whaty ou mean [23:42] so from workstation rsync to all the hosts at same time? [23:43] Valduare: ah! check out the "Handling wp-content" section here: http://manage.jujucharms.com/charms/precise/wordpress [23:44] of course that makes more sense than rsync. haha. [23:56] ok === mhall119 is now known as mhall119|offline