[00:06] jcastro: dude, that's going to be a great success story [00:06] are you going to set them up to scale on-demand when they get high traffic and everthing? [00:17] jcastro: lol no worries [00:17] jcastro: saving OMG how? [00:18] I wonder why they didnt redirect their DNS to a splash page somewhere :P [00:18] with a 301 [00:18] OMG is going to taking a SERP beating [02:06] mhall119, jcastro - http://www.peppertop.com/greys/2010/02/ubuntufied-flying-object/ [02:14] nigelb: lol [02:15] quick! change the tag line! [03:03] woah, link to my blog at the end of that [03:19] * nigelb blinks [03:21] ha [03:32] mhall119, almost done [03:32] man [03:32] what a horrible mess [04:01] jcastro: what happnened? [04:18] he migrated hosts [04:18] and like, had no clue how to set it up [04:18] so we just deployed it on aws using juju [04:19] o_O [04:19] anyway, epic juju win [04:19] man [04:19] our wordpress and mysql charms are for chumps [04:19] we need to optimize them [05:03] jcastro: He didnt know how to do a Wordpress Install? [05:03] seriously [05:03] :( [05:04] epic so bad [05:04] jcastro: ^ yes they are [05:04] jcastro: Uhh you might want to have him Submit a rdns request :P [05:05] http://ec2-23-21-149-239.compute-1.amazonaws.com/ <-- omg [09:24] * popey wonders what he'd have to do to get jcastro to jujuify the ubuntu uk podcast website ☺ [11:56] can anyone explain to me the idea that bug 958384 is invalid because it needs discussion? I dont think the fact it needs discussion means it isnt a bug...? :/ [11:56] Launchpad bug 958384 in libreoffice "Libre office default style settings are poor." [Undecided,Invalid] https://launchpad.net/bugs/958384 [12:37] jussi: it's an opinion then [12:41] czajkowski: I dont think the fact of it is disputed by anyone, just that how to solve it needs discussion. [12:45] jussi: but the discussion is pointed to a ml [12:45] which imo is a better place than bugs tbh [13:36] yeah, but it is nice for the bug to exist so it can be fixed by a commit at some point [13:57] nigelb: do you know about boot to gecko? [13:57] yeah [13:58] you can clone gaia and run it on your browser. [13:58] probably only nightly though. [13:58] how about on my galaxy s2 [13:59] oh, you can probably run the whole thing. I don't know how far it is though. [13:59] ok [14:00] I can check with someone on monday [14:02] looking at the roadmap it might still be a bit early to jump in [14:07] jcastro: are you going to guest-post on OMG about getting them back up? [15:40] mhall119, I just blogged it on mine [15:40] and will syndicate on cloud.u.c [15:40] and probably make an OMG! specific one [16:09] jcastro: hey, how did you copy the data around? [16:09] rsync [16:09] of the mysql data files? [16:09] I'm surprised it took that long then. [16:10] oh wait, it's between a different provider and aws. Nevermind. [16:18] yeah [16:21] what happened to OMG [16:21] did I miss something? [16:21] It was down for 3 days. jcastro called on the power of juju to fix it. :D [16:21] ic [16:40] mhall119, hang with me here [16:40] mhall119, ok so he's working on the charm now [16:40] and we'll redeploy to m1.smalls [16:41] do you know if you can move a bucket from one person to another? [16:43] jcastro: I don't, sorry [16:55] jcastro: what is the juju bootstrap node for? Surely you don't need a dedicated instance just to use juju [16:55] it runs the zookeeper [16:55] but we can move that to run on the same instance as the proxy [16:55] we just can't right now until next week [16:55] as the feature hasn't landed [16:56] but once it lands you could run everything on one node if you wanted like how AlanBell would want [16:57] I would be interested to know if things are actually faster or slower on one node vs 4 nodes [16:58] one big node vs. 4 smalls? [16:58] yeah [16:58] think I would bet on a single node being faster than having a database on a separate node to the application, and without an extra proxy jump [16:59] or all on one small vs all on 4 smalls [16:59] I am reasonably confident that this will be cheaper than his colo set up [16:59] the problem for OMG being on only one (or say, one large or medium) is then we wouldn't be able to grow and shrink on demand [17:00] The biggest issue right now is apache serving up all those images [17:00] AlanBell: I'd expect better performance with MySQL on a separate node [17:00] and getting that to S3 asap [17:01] mhall119: depends how chatty the database and the application are I guess [17:02] the main thing to do is avoid using the database at all and use wp-supercache or something [17:03] if you are hitting the db, you want it fast and close, so multiple network requests to get stuff doesn't sound like a good plan [17:04] I think Wordpress defaults to MyISAM, which is pretty darn fast [17:04] plus MySQL does query caching [17:05] jcastro: Elastic Load Balancing costs $0.025/hr, and supposedly does auto-scaling too, which might work better for them than haproxy [17:06] yeah ryan pointed that out to me [17:07] you might have to make custom OMG images though, for that to work [17:07] it would be awesome if Amazon supported juju charms for this stuff [17:07] yeah plus, I'd like to give him some options to not be AWS specific [17:08] we have to be AWS specific on the storage for now [17:08] but it'd be nice to say ... in a year he finds a nicer price on rackspace or a local ISP for instances [17:08] and just redeploys there. [17:08] or bare metal if he wants to colo again [17:08] true, being able to quickly redeploy somewhere else is a nice option to have [17:09] though not all will support S3, so that's something to consider in your wordpress-omg charm [17:09] I would love to build servers using juju as a standard practice, but it just chews up too many instances [17:10] one more week [17:10] one more week [17:10] i've been aching for it too [17:10] an extra instance is a deal breaker. [17:10] especially since you also can't say "juju I want an m1.small for you, an x1.large for wordpress, and an m1.medium for mysql" yet. [17:10] but next week we get all of that. [17:11] unfortunate timing [17:11] I want to say, here's an m1 server. Install wordpres, mysql, and apache on it. [17:11] Not spam 4 servers for that. [17:12] exactly [17:12] right [17:12] *spawn [17:12] We don't do that until next week [17:12] one week! [17:12] then one day when that isn't enough, split it out smoothly [17:12] I'll take that as win. [17:12] right. [17:12] believe me no one is aware of this pain more than juju folks [17:12] I will check this out when precise releases. [17:13] I expect all of this would be fixed then. [17:13] indeed [17:13] one more week? [17:13] if we can do a juju based openerp install on one instance, that can be scaled up to lots of instances then that would be a big win [17:13] finally a tool that's not in ruby. [17:15] AlanBell, actually just last week someone submitted openerp [17:15] mhall119, yeah, the feature is called "constraints", it's all reviewed and ready to land. [17:17] so they did, old version though [17:17] no worries [17:17] we can easily make that pull trunk [17:17] in fact that's what marco's working on for the wordpress charm [17:17] it stupidly pulls the archive version [17:17] which is basically shipping the old one that doesn't upgrade ootb [17:18] boo... [17:18] openerp 6.1 has slightly different install routine as the webserver now runs in-process [17:19] would you be comfortable reviewing the charm? [17:20] yeah, or I might hand that over to TheOpenSourcerer [17:22] nod [17:23] I don't even know why we should have wordpress in the acrchive. [17:24] the juju charm store isn't active yet, so it's not obvious to people [17:24] that with charms [17:24] you don't have to care about the archive [17:24] we'll have fresh wordpress for the life of 12.04 [17:26] mhall119, how typical [17:26] the comments on facebook are talking about unity. [17:27] nigelb: I said the same thing :) [17:27] jcastro: obviusly. facebook crowd isn't known for maturity :) [17:28] marcoceppi: It is a bigger security hole :P [17:31] I totally don't get Facebooko [17:42] o/ [17:50] ok let's do math [17:50] ~16gb of traffic so far since last night [17:50] so, 12 hours worth [17:51] wtf, 16GB in 12 hours? [17:51] no wonder you want his images moved to s3. [17:51] it's only a cent cheaper [17:51] still [17:51] at 11 cents, it's $100 a month [17:52] that's $316 a month [17:53] how much did he used to pay? [17:53] lol [17:53] 200 pounds [17:53] which right now translates to [17:53] $316 a month [17:53] ha! [17:53] hahaha [17:53] that's sunday traffic [17:53] let's assume he's busier [17:53] and double the traffic [17:54] it's $416 [17:54] ah bummer [17:55] the price only goes down to .09 a gig when he tops 10TB. [17:55] ok am I reading this wrong [17:56] the data out of S3 is the exact same price as EC2, 12 cents [17:56] that can't be right [18:22] gosh [18:23] I bet rackspace is cheaper. [18:24] this is what most of our hardware is http://www.hetzner.de/en/hosting/produkte_rootserver/ex5 [18:24] ha, hetzner :) [18:25] we slice those up using KVM into quite a few VMs, about 10 per host [18:26] Manually? [18:26] yeah, we wanted to use the eucalyptus stuff but that required an additional host as a manager [18:27] nice [18:27] so it is just KVM, and I use the virt-manager thing to poke at them and spin up new ones etc [18:29] would love to do it in a more buzzword compliant manner, but cloud stuff is massively more expensive [18:29] heh, yeah. [18:32] What makes AWS better than say rackspace? [18:33] AWS has more features [18:33] Also been around longer [18:33] and more datacenters [18:33] it is from a company that real people have heard of too [18:34] lol [18:34] When did AWS come out? [18:35] 2002. [18:35] I know that there is no direct price comparison because the sizes are different, so I'd have to do math [18:37] I've also never gotten along with anyone from Rackspace [18:37] lol [18:37] I have a few good friends who are with Rackspace.. They all started with Slicehost.. but are now at RS [18:37] soren is a nice guy AFAIK. [18:37] to me, when looking at AWS to setup, it seems overwealming [18:38] but then he no longer works at rackspace. [18:38] cjohnston: It's dead simple actually. [18:38] cjohnston: juju, duh :) [18:38] lol [18:38] juju is brand new [18:39] cjohnston: the button clicks are not too complicated. [18:39] but for me, I run everything on a 1gig cloud server at RS.. $0.06/hr... I don't need multiple isntances or anything like that.. I run multiple blogs on it, I get IP addresses as needed.. [18:39] RS is *too* simple for me to run anyhting on it professionally. [18:40] I guess I like the fact that its nice and simple [18:40] I'm talking about DBs that are 20 GB in size. etc. [18:40] Each provider has it' use case [18:40] For personal use, I'd probably trust rackspace than amazon. [18:41] It is cheaper. [18:41] I'm just trying to figure out, because I *have* to learn to use amazon, should I learn it and move my stuff over.. [18:42] RS seems to have more DCs.. they have 8 [18:43] a little less spread out.. but still comparable [18:43] huh? [18:43] does RS let you select which data center to use? [18:44] hrm.. I guess for cloud servers they are still in DFW [18:45] among traditional hosts Rackspace (at least in the US) is one of the more dominant ones [18:45] if anyone can compete with AWS it's them [18:45] but they needed openstack for that [18:45] YEah [18:45] Not yet there. [18:45] I waiting for them to have openstack. [18:45] they have a cloud thing [18:45] but it's not openstack [18:45] not as great as amazon yet. [18:46] not as many features. [18:46] It's pretty much a monopoly [18:46] yeah, sadly. [18:46] however HP's Public cloud is in beta [18:46] as openstack matures everyone will just start selling instances [18:46] the same way they sold VPSes and shells [18:46] oooh. that should be interesting [18:46] is the HP thing openstack? [18:46] It is [18:47] on ubuntu [18:47] \o/ [18:47] if openstack had an equivalent to s3 stuff that was solid [18:47] like a storage solution [18:47] we could then have juju just generically support storage [18:47] and the charms wouldn't need to be s3 specific [18:48] there's still stuff that can do something like that [18:48] then you could just move to whatever hosting provider you want [18:48] glusterfs? [18:48] gluster and ceph [18:48] I expect storage to be the big topic this year in openstack [18:54] jcastro: I am getting a tomcat message from omgubuntu.co.uk [18:55] DNS is out of date for you then [18:55] ah [18:55] try privacy mode in your browser [18:55] and try again [18:55] you might be caching the old 301 redirect [19:01] jcastro: you are kidding, right? [19:02] jcastro: opnestack has swift for file storage.. It's what rackspace uses in production called "cloudfiles" [19:02] predates openstack. [19:02] Daviey, oh sweet so I can just point juju to it and get the exact same thing I can with s3? [19:02] jcastro: that is the plan! [19:02] right, the plan [19:03] but doesn't help me right now. [19:03] (there is also a ceph object storage component, but not experimented with it myself) [19:03] yeah, for now his charm will be S3-specific [19:03] jcastro: Canonsistack doesn't yet provide S3. [19:03] but the idea of being able to redeploy on openstack is awesome, I will actually try it too [19:04] * AlanBell ponders a performance faceoff with a multi-tier juju rig vs an all in one on a VM setup [19:04] Daviey, oh so what you're telling me is that limitation is our deployment specific? And that everything is fine in upstream? [19:04] "fine" as in, will be working [19:05] jcastro: Well, swift is largely regarded as the most stable component [19:05] man, I am delighted to be so wrong. [19:05] Daviey, someone should put you in charge of stuff [19:05] there is a shim which proxies to s3 in nova.. it's not /well/ tested, compared to the rest of it. [19:06] jcastro: hah [19:07] man I am kicking myself for the xlarges [19:07] Isn't he already incharge of stuff? :D [19:07] marcoceppi, it was getting late and I was getting inpatient [19:07] should have texted Clint way earlier [19:07] also, seeing Daviey on IRC on a weekend. [19:07] I'm shocked beyond belief. [19:08] $60 dollar mistake, that's like a nice steak. [19:12] popey, clearing the cache do it? === webjadmin_ is now known as jalcine [19:22] man there's tons of things we can optimize here [19:22] these pages are heavy [19:22] and the images are like, being resized in html [19:22] http://cdn.omgubuntu.co.uk/.../Screen-Shot-2012-03-12-at-17.03.47-50... is resized in HTML or CSS from 500x295 to 150x88. Serving a scaled image could save 183.2KiB (91% reduction). [19:22] a whole bunch of these right on the homepage [19:23] I bet we can easily slim down the amount of things he needs to serve on top of what we're doing [19:25] jcastro: He needs to smush his images and use image polish to dynamically resize on the fly [19:25] I think WP does this in the newish version [19:26] the new charm will just use upstream wordpress so he can update on his own schedule [19:26] jcastro: there is a wordpress Yahoo SmushIt plugin... it will resmush all his images [19:26] as for Image Polish only Cloudflare offers this atm [19:26] * jcastro notes that [19:29] cjohnston: should I be seeing http://summit.ubuntu.com/uds-q/track/community/ [19:29] jcastro: I would be interested in a funkload report of a default (unoptimised) wordpress install on a single small instance compared to a 3 instance setup [19:30] czajkowski: no [19:34] nigelb: Are you kidding? [19:34] nigelb: I'm on most weekeneds [19:35] Daviey: since your around and all.. want to help me with an error that I came across with summit ;-) [19:35] hehe [19:36] cjohnston: oh? [19:36] Daviey: http://summit.chrisjohnston.org <-- click login [19:37] a google search gave me 3 results, which didnt help. :-/ [19:38] cjohnston: Are you missing a *kwargs? [19:38] i don't think I changed anything that would effect login... [19:39] it works locally too [19:40] cjohnston: is production summit running 1.3 now? [19:40] yes [19:42] https://code.launchpad.net/~chrisjohnston/summit/action-item-links and https://code.launchpad.net/~chrisjohnston/summit/new-meeting-stuff plus a little css is what was merged in [19:43] cjohnston: Hah, https://bugs.launchpad.net/software-center-agent/+bug/873244 suggests a missing kwarg's aswell [19:43] Launchpad bug 873244 in software-center-agent "custom openid_render_failure has wrong signature" [Undecided,Fix released] [19:45] hrm [20:07] Daviey: I've looked back through the code, and unless I'm missing something, I don't see it. :-/ [20:22] cjohnston: I suspect it's that you are lacking relevant data on your devel site [20:22] Perhaps don't have an admin user, a UDS object created, or something like that. [20:22] the devel site was working prior to the merge as well [20:25] Oh, NFI then [20:25] sorry :) [20:25] ok [20:25] ty [20:25] cjohnston: Have you reverted the merge? [20:26] i just did.. it has an issue with the undoing the migration [20:26] who merged a migration that doesn't allow a reverse? : [20:26] :) [20:28] it has an inconsistant migration history [20:33] Daviey: NFI is now my standard response to summit :D [20:49] :) === huats_ is now known as huats === jalcine is now known as jalcine_