[00:15] hazmat: yeah I think going to the current series makes the most sense [00:43] <_mup_> Bug #984484 was filed: subordinate charms should be able to open ports < https://launchpad.net/bugs/984484 > [01:06] * SpamapS feels so lonely.. oh so ronery [01:06] oh wow, my aws bill last month was only $0.08 , nice [01:07] i think they might have messed up tho lol [01:09] SpamapS: http://aws-s3.assets-online.com/pixeldrop/fun/macpro-apple-online-store.png [01:09] haha [01:09] SpamapS: and what ya think about a apache mod_speedy only on 443 and nginx only on 80 , hrm :) [01:09] mine is hovering around $80 - $100 these days [01:09] I should probably destroy-env soon.. have had 4 m1.smalls running since Friday [01:10] yea normaly mines just under 100 [01:10] imbrandon: I love that idea. I think I can make that work just with the subordinate [01:10] imbrandon: I was thinking that I'd change it to just proxy to the local port 80. :) [01:11] SpamapS: cool cool, wont work already thought about that [01:11] heh [01:11] spdy needss end to end, cant even have nginx reverse 443 to it [01:11] so no one second cache love for spdy [01:11] untill nginx gains support [01:12] been reading up on it the last hour or so [01:12] but thats my extent lol [01:14] imbrandon: I was wondering about that [01:14] but yea, i am thinking a major reworking of the omg nginx stuff might be in my pipeline soon, as in i've already begun to forumlate it in my head and on my dev server, some of the hastily things we did , while is running awesom right now, i;ve been tuning and refining diff ideas to make it sooo much better [01:15] but it will be between now and uds and likjely i'll grab a hallway session with you and marcoceppi and whomever else and go over it more in person [01:15] then MAYBE implment it after [01:15] imbrandon: if we can make it generic enough, it should be a good 'nginx' charm :) [01:15] right thats my ultimate goal [01:16] is to have a "full stack" but all be little sub charms where it makes sense [01:16] that way too parts can be droped out like the php and rails dropped in [01:16] and still share the good bits [01:18] but i think its gonna take me the next two weeks of iterations and stuff to have it be as compartmenalized as i'd like and that means it will be a good time to "present" a semi if not fully working "stack" at uds betqween us that have intrest in such things [01:18] and published "public" after that ( like blogged about etc, i'm sure it will be in the charm store much sooner ) [01:19] imbrandon: I don't actually see how mod_proxy can't be used [01:19] * SpamapS is trying it right now [01:19] least thats the .plan i've been working from the last week [01:19] SpamapS: let me find the link,on the spdy forum a official dev said it needed end to end to work right [01:19] one sec [01:21] here is where i got that info, it may be old or wrong [01:21] https://groups.google.com/forum/?fromgroups#!topic/mod-spdy-discuss/XCQG4w0plaE [01:21] the orig question is deleted now [01:21] it was there earlier [01:21] but the important part the response is still there [01:22] the orig question was about using nginx to proxy back to 443 spdy and then serve on 80 [01:22] or something [01:22] or no, the orig is there, i just had it collapsed [01:22] SpamapS: ^^ [01:23] am i reading that wrong ,or are they just wrong, or its old [01:23] from feb 4 [01:23] imbrandon: I don't think thats definitive [01:23] perfect, i would LOVE to be able to do it [01:24] infact my "ultimate" setup i run on "my" sites even though i purport nginx is ... [01:24] apache based zend server on 8000 , zend php on 127.0.0.1:9000 fpm and then nginx reverse proxy 1 second cae to localhost 8000 [01:25] but thats because i actually pay the money to have zend server and zend studio [01:26] although zend server community edition might be nice too in a genericized charm [01:26] why apache zend on 8000? [01:26] arbitraty how i set it up years ago and just keep doing it that way [01:26] no diff between 8080 or whatever [01:27] no I mean why apache? [01:27] just cant or shouldent use 9000 , as fcgi and fpm use it default and dont use 10080-10088 cuz zend server uses it for the "gui" [01:27] oh zend server is built on apache [01:27] fpm seems to scale better than mod_php at this point. [01:27] its all one "package" setup [01:28] yea i use fpm [01:28] with apache [01:28] :) [01:28] ohhhhh i noticed too aparently tcp fpm WAY out preforms unix:/ socket fpm [01:28] i need to run some benchmarks to confirm but i've seen it multi places now [01:29] what? [01:29] yea seems bas ackwards [01:29] that doesn't make much sense at all [01:30] so like i said i need to see with my own eyes, but the word is php unix sockets dont scales as nice as php fpm on 127.0.0.1:9000 tcp [01:30] i think omg might be a good test for that theory at some point [01:31] it has the traffic to get "real" numbers and shouldent hinder the site too bad if one is not as good as the other [01:31] but thast another "when i have time to test correctly " type thing [01:33] but yea i use apache because , well 1 if yu do it right apache CAN be nearly as nice as nginx, and i get all the .htaccess reqrite out of the box etc , and 2 probably the bigger reason is its what is supported and bundled with zend server and the GUI tightly intergrates with it even though you CAN use other servers and zend php its not the "package" deal [01:33] but nginx makes a killer reverse proxy on the same box, and even servs some vhosts directly [01:34] with the 1 sec cache. so personaly on my server i get the best of both worlds so to speak [01:35] ok proxy does not work [01:35] definitively :-P [01:36] [Wed Apr 18 01:33:30 2012] [warn] proxy: No protocol handler was valid for the URL /favicon.ico. If you are using a DSO version of mod_proxy, make sure the proxy submodules are included in the configuration using LoadModule. [01:36] SpamapS: enable mod_proxy_http too ? [01:37] oo perhaps I forgot that [01:37] heya elmo :) [01:38] imbrandon: ey [01:38] oh ok that *was it* [01:38] elmo: works! [01:38] eh, typing hard [01:38] imbrandon: hey even [01:38] https://ec2-107-22-22-168.compute-1.amazonaws.com/ [01:38] SpamapS: cool [01:38] :) [01:38] of course, statusnet is broken cause I forgot to configure it [01:39] nice [01:39] can you pipe it to 80 ?> [01:39] i'd love to be able to do spdy over 80 [01:40] i only have ssl cert for one domain, and i dont even use that domain nor have the ssl cert installed atm [01:40] lol [01:40] imbrandon: well no apparently SPDY implies HTTPS [01:40] https://wwwbox.co is the only cert i have [01:41] SpamapS: yea i know, its cuz google is an ass about it, no tech reason it cant work on 80 AND there is more initial latency on https eventhough spdy makes up for it later [01:41] spdy on 80 would be even faster :) [01:43] imbrandon: I think google is thinking about your privacy :) [01:43] or even about oppressive governments [01:43] self signed works fine [01:44] yea but if i'm gonna do it i dont want the user to have popups or warnings [01:44] so i'd buy one for brandonholtsclaw.com [01:44] if i get a referal code i can pick a godaddy.com cert up for like 16$ a year on sale [01:44] :) [01:44] anyway, I'll have to wrap it up later tonight [01:45] ok cool, ping me when your working on it, i'm very intrested and doing related stuffs [01:45] i'll be on late tonight [01:45] I think we can make it work, and awesome, by simply having the primary charm feedback where it stores static files and then the mod_spdy can serve stuff directly... eventually we can go with a 'worker' mpm apache and it should be able to at least try to keep up with nginx's crazy speed [01:46] imbrandon: lp:~clint-febar/charms/precise/mod-spdy/trunk [01:46] yea and i want a little help with my first sub too, when ya get back, i gpt the bones of it done and it might be ready for the store [01:46] but not sure 100% and no way to test for a few hrs [01:46] imbrandon: just have to add a ProxyPass and ProxyPassReverse to the top of files/all-ssl and it works [01:46] kk [01:46] i'll try thaat here in a few moments [01:46] imbrandon: oh, and 'a2enmod proxy proxy_http' [01:46] yea i have those already [01:47] anyway, out [01:47] l8tr [01:48] yea apache with only BARE min modules installed can come very close to nginx, nginx still beats it in the very very heavy traffic places, but i'm talking 100+ req a sec or more [01:49] untill that point it can keep up great if configured right [01:50] and even more with varnish etc, but i'm loving nginx and kinda dropped varnish and its vcl's from my toolbag in favor of learning nginx better to do the same job and more [01:51] nginx with a 2gb ramdisk ( in a beefy server with like 8+ gb ram ) and pointing it fcgi_cache or proxy_cache to the ramdisk, wow wee [01:51] speed out the yang, it actually becomes a cpu bottleneck [01:52] before anything else [01:54] SpamapS: imbrandon: upstream nginx on twitter says "May" for nginx/spdy. [01:54] just now getting into my devops experince where until now faster was better ALWAYS but a few of the sites ( like grammy.com ) challanged that notion for me lately , where slighly slower config but able to handle a more sustained load is atually better, where as even other high traffic sites i've gneer'd like pets.com never hit that mark [01:55] imbrandon: I would say a good way to get ready is add an upstream/ppa flag to the soon-to-exist-should-already-be-in-progress nginx subordinate [01:55] jcastro: yup thats the idea, working on doing that tonight with some help/overlap from SpamapS [01:55] so when it's released you just flip the one bit [01:55] May is relatively eons from now anyway [01:55] yup yup, i like that idea [01:56] yea but it will get pushed sooner, once more ppl pickup on the new spdy release [01:56] more will have intrest and work on it more even upstream as things pickup [01:56] so it will come out sooner than whats there now [01:57] I think it's a cool use case [01:57] its probably gauged on current level of dev contributions timeline, some hotrod will come in and get it 80% of the way and a core dev will be semi forced to fix the otyher 20% early :) [01:57] ( nginx that is ) [01:58] thats how it seems in my head atm :) who konws i'm often wrong [01:58] lol [01:59] "ok this new thing is out, I want to play with it and stuff, but I don't want to mess with all this stuff. Oh, I'll just redeploy on staging with this new charm and try it." [02:01] imbrandon: if this would have been nginx support you know we would have stayed up all night, heh [02:01] but yea i fully intend to do that, my intentions are to "own" the nginx and some other related charms for the long haul *as much as is done in ubnutu, we all contrib cross and like i'm sure SpamapS work and exprince with the stack and marcoceppi with his other php stuff as well as many others , but you konw my meaning , not "hoarding" owning it, but more like the guy that takes the time to research ever little but and test and code and etc , not j [02:01] hahaha well with it reverse proxy it can be [02:02] hehe thats what me anc clint was just talking about and i think i am gonna be up all night here working on it LOL [02:02] its psudo supoort but the result is the same [02:02] :) [02:03] AND we'll be the first, not only in ubuntu but the greater internets from what i can tell searching since you posted about it today [02:03] yeah [02:03] btw did you get my feed on there ? [02:04] yeah [02:04] yup, i am, just checked, plenty of my aderol perscription left, i'm pulling an al nighter and gonna get some rough cuts pushed [02:04] lol [02:05] imbrandon: hey so something SpamapS and I mentioned over the phone was perhaps doing the apache subordinate thing for charms that right now are using the built in node webserver, etc. [02:05] like the more developer-y charms [02:05] subway just uses the built in node thing afaict, etc. [02:06] I think subway is interesting because it's a 2 way "application" that needs lower latency, and built in ssl for your chat would be a nice win anyway [02:09] imbrandon: we have no issues with apache for LP, and we're past 100rps [02:16] SpamapS: your statusnet instance is still running [02:19] lifeless: yea but youve taken the time to configure it right, i'm talking about joe blow [02:19] and i am not saying apache falls on its face there, [02:19] only that nginx starts to pull ahead alot more [02:19] at that point [02:20] jcastro: yea thats very very common [02:20] re: the subway thing with apache [02:21] infact its the princapal generally behind like php fpm that really runs on port 9000 serving php and python wsgi now they are very very diffrent but the general princeapal is the same [02:22] so sticking a apache/nginx infront of subway and putting it to 80, then letting nodejs direct connections too is common [02:22] well maybe not for subway but that general type of app [02:22] nodejs is good at longpoll and other stuff http servers arent really made for [02:24] but they are still usefull combine, and thats where ppl like us come in, its easy to do all this crap stand alone, but make it all work togather and in the best way is the hard part, things like juju and chef distribute that more to more devops but thats recient, mostly learning what we know untill now has been a black magic and handed down from mentor to mentee like blacksmithing and over years sometimes working togather [02:26] that knowhow is becoming a commodity now, so for ppl like me to stay relevant in 5 more years i need to be on the tech end that makes the things that make it irrevant , and then do it all again , it goes in 3 to 5 year cycles like this since i've been active online in the mid 90's, likely sicne the 50's and early sixtys with the adevent of unix with SysV+ [02:27] * imbrandon gets back to charming [04:14] jcastro: yes I know, but now it shows the statusnet error on 443 ;) [04:18] imbrandon: FYI, Ryan Dahl created node.js specifically *not* to need any frontend proxy for low latency needing apps [04:19] SpamapS: yea i just ment it was common to do for those type things [04:19] not really nodejs specific [04:19] like, his whole point was that you can write an app that will be highly coherent between requests.. "sessions suck" [04:22] Heh... looks like statusnet is just plain broken [04:22] yea but in type i'm including google dev appserver and python and perl and non_mod_php [04:22] all the kinda non standard plain http on 80 apps [04:22] ;) [04:22] nodejs just kinda got lumped in there [04:22] we really need to wrap up "real" automated tests.. these hooks pass install/config-changed but they'r enot really error checking [04:22] well some node apps [04:31] ugh [04:31] we need a 'switch-charm' command [04:31] can't improve any cs: charms in place.. :-P [04:32] ? [04:33] hey ok so if i ssh into a juju instance , can i manually file off relation get and stuff ? [04:34] to se the actual output instead of charm upgrading and echoing it or somehing to see the output in a log [04:34] imbrandon: you can but you need to know the relation id [04:34] can i get that from status , or ? [04:35] $JUJU_RELATION_ID in the charm [04:35] and relation-ids command [04:35] k [04:35] you also need to set JUJU_SOCKET [04:35] yea thats the error i got was something about the socket [04:35] k [04:35] is it in /tmp or something [04:36] I foret [04:36] I forget even [04:36] kk [04:37] probably /var somewhere [04:56] mmmm [05:28] Ok, this definitely *feels* faster in chromium than in firefox https://ec2-23-21-39-39.compute-1.amazonaws.com/main/login [05:28] firefox does spdy too [05:28] imbrandon: but its not turned on [05:28] ahh ok [05:30] statusnet is seriously a very weird webapp [05:30] you can't have a nickname of 'SpamapS', only 'spamaps' [05:30] doh [05:30] registration not allowed [05:32] also always builds absolute links [05:32] I hate apps that do that [05:32] ha [05:32] yea me too [05:32] wordpres kinda does, it atleaste uses whats in the db [05:32] so it the db is changed all change [05:32] still shitty tho [05:33] I don't think it does that for page references though [05:33] it will redirect you.. [05:33] but I can work around that [05:33] this.. this is building the page with which is *stupid* [05:33] "/foo" would be fine [05:33] Resource interpreted as Font but transferred with MIME type application/octet-stream: "http://ec2-23-21-39-39.compute-1.amazonaws.com/theme/neo/fonts/lato-italic-webfont.woff". [05:33] and make it work proerly [05:34] imbrandon: yeah, its not an optimized app [05:34] yea and a href="//amazon.com/blah" would be even better if they NEED absolute lionks [05:35] putting http{,s}: on links is dumb just for this reason :) [05:36] for like off site resources. the only time i do it is if the site dont have https too, then i try to find a nother link or just build whatever it is myself [05:36] lol [05:37] imbrandon: right but in this case, they're just pulling in HTTP_HOST and building the whole link [05:37] thats just lazy [05:37] but i havent found one in a while that dont except ga.js and it does just with a diff name, due to a ie6 bug, but if you dont care about ie6 then its all good ( hint: i dont, not even on commercial gigs ) IE8+ only [05:37] SpamapS: my martha plugin grabs http host [05:38] and still works [05:38] but there's no reason to put the host in there [05:38] or even a leading / usually [05:38] well the leading slash i can see [05:38] templates can be used [05:39] and that nessesitates it incase subdir etc [05:39] .. works fine :) [05:39] sure if you want to put logic in the template to tell what dir your in [05:39] or just use root relitive links and all are happy :) [05:40] * SpamapS tries subway now [05:40] ( css too , like when using less, you never know where the final product will be ) [05:41] imbrandon: I usually use image_path('foo.jpg') which does in fact figure out where the request was made and build a relative link. [05:41] sides, ../ transversal in php is code smell for an audit imho :) [05:41] imbrandon: thats how symfony did it anyway :) [05:41] imbrandon: thats not in code, that is going to be emitted in the html. [05:41] and browsers are happy to use it [05:41] yea its similar in drupal and zend too but the end result thats rendered is a root relitve link [05:42] lazy [05:42] smart imho [05:42] less headache [05:42] can't sub-host though :( [05:42] less going back to fix little shit [05:42] sure ya can [05:42] anyway, whats a good charm that exists now that has tons of on screen assets? [05:42] why not ? [05:42] ThinkUp maybe? [05:42] hrm [05:42] omg-wp ? hahaha j/k [05:43] never used thinkup [05:43] sucks down your twitter feeds and facebook and G+ and puts it on one site [05:43] oh nice [05:44] nagios , hrm that really dont have any assets [05:45] stackmobile is the only other one maybe, never seen its gui tho so i duno [05:48] hi all [05:49] ello [05:49] imbrandon: http://ec2-23-22-28-42.compute-1.amazonaws.com/ [05:49] simple GUI [05:50] nice tho [05:50] could use a bit of spruce but alot better than most [05:51] i hate "flat" buttons like that, they dont have to go all out css3 gradient crazy but that is one of my pet peives. hell leave it to the ui toolkit if your gonna make it flat :) [05:51] hell i bitch alot [05:53] ohh and they use bootstrap on their own site, smart guys and gals [05:53] :) [05:53] and h5bp, wow, i'm inpressed for a floss web app [05:53] :) [05:55] h5bp ? [05:59] html5boilerplate [05:59] ah [05:59] sorry was reading some of the other code [05:59] <-- ui ignorant by choice [05:59] its best practices from tons of experts , like paul irish etc [06:00] there is a huge community arround it, and they put a TON TON TON of thought into every single byte in the boilerplate [06:00] everything is there for a reason and in a certain order for a rewason etc etc [06:00] and what makes it so cool is its all 1 run by build scripts etc, no runtime or language deps, eg rails php python etc all can use it [06:01] and 2 they explain WHY everything is why it is [06:01] and CONSTANTLY update it [06:01] like many times a day [06:02] they hang out here on freenode in #html5, and here read a tad bit of this in their issue queue to see how much thought goes into each little part, and this is a tame one [06:02] https://github.com/h5bp/html5-boilerplate/issues/378 [06:02] i have hella respect for those guys [06:04] its nice to work with stuff that is clearly maintained out of a sense of duty :) [06:04] yea [06:06] ok so i can forgive the button for all the other goodness this app has :) [06:06] hehe [06:10] SpamapS: know if nginx can directly serve content from the configs ? like i used to use a snipit that would actualy serv the /robots.txt from an apache vhost config if one wasent on the filesystem [06:10] hrm [06:10] probably have to look it up /me goes to do that [06:11] hard to call a winner with this one https://ec2-23-22-28-42.compute-1.amazonaws.com/ [06:12] ? [06:12] but... mod-spdy to the rescue, no mods t thinkup to get it SSL wrapped [06:12] imbrandon: SPDY doesn't help much with such a tiny well designed site. :) [06:12] hehe right [06:12] well it would if it say had 1000000 little images like flickr [06:13] but yea [06:13] tiny images i tend to base64 encode and put them in a data uri anyhow , like the bullets for ul's etc only exception is font icons i use [06:14] that way its cached with the css [06:14] Oh, mediagoblin would be a good one [06:14] and only one http req total [06:14] I don't even know if my old mediagoblin charm will work [06:14] heh [06:14] wordpress has tons of images [06:14] if you toss a theme in it [06:14] yeah but I hatezorz our default wordpress charm ;) [06:15] like generic wp + a gaudy theme [06:15] hehe [06:15] hey do drupal [06:15] and it will probably just redirect me to http:// [06:15] it needs prom anyhow [06:15] and will be a good test [06:15] I'll be reviewing it on Friday [06:15] or maybe tomorrow [06:15] thats cool but it does have lots of little images [06:15] need some solid apps for charm school on Thursday [06:16] yea you'll be like the 5th hehe , seems like everyone looks at it once and never returns hahahahahahahhaha [06:16] imbrandon, not me :D [06:16] but yea, i've installed it a few times now on micros, its works, still lots of room for imporvment :) [06:16] koolhead17: heh [06:17] koolhead17: not sure i've met you :) Hi 0/ [06:17] hi imbrandon :) [06:18] imbrandon: so maybe call it "thering" [06:18] while you review it, the phone rings [06:18] :) [06:20] its ok i'm in no hurry , just got it done at a bad time, everyone heading to os conf [06:20] and such [06:20] * koolhead17 rushes 4 owrk [06:20] *work [06:22] alright, sleep time [06:23] gnight [06:23] hrm ok i really need to finish this charm so i can get back to researching this spdy stuff :) [07:38] SpamapS: Do you know why I would be getting lots of merge notifications from Juju's LP :P [07:42] <_mup_> Bug #984640 was filed: Unsatisfied constraints are not reported back to the user < https://launchpad.net/bugs/984640 > [08:53] * koolhead17 assumes SpamapS is sleeping :P [10:57] somehow my ~/.juju folder is 128GB [10:57] marco@marco-g72:~/.juju$ du -sh [10:57] 123G . [11:02] marcoceppi, could it be cached charms? [11:02] marcoceppi, but, yeah, that would be a lot of charms... [11:02] fwereade_: it looks like local was still bootstrapped, and machine-agent was 120+ gb [11:02] marcoceppi, ha, phew [11:02] but this laptop has been restarted several times, I didn't think local lxc containers survived restarts [11:02] marcoceppi, still seems like a lot [11:03] marcoceppi, the containers survive, but they don't restart [11:03] well, this would have been a bootstrap from March 29th [11:03] So, three weeks worth of machine-agent.log [11:07] marcoceppi, ouch, I wonder if we already have a bug for that [11:11] I mean, what's the bug there? logrotate should be rotating machine-agent log? [11:19] marcoceppi, I think so, partly; but do you have a lot of zookeeper spam in there as well? [11:19] marcoceppi, we should probably be reducing that as well ;) [11:19] fwereade_: I torched the file, since I was at 100% disk space [11:20] marcoceppi, heh, can't blame you ;) [11:20] From what I remember it was a lot of zk [11:21] marcoceppi, yeah, looking at mine, there's quite a lot [11:22] marcoceppi, and, hmm, quite a lot of it is the machine agent (which does restart) whining that it can't find zookeeper (which doesn't :/) [12:44] SpamapS: https://cloud.torproject.org/ is only for EC2 though === carif_ is now known as carif === al-maisan is now known as almaisan-away [14:55] SpamapS: wouldent it be good to use this for the replacement of s3 to find the meta data, scroll down to the first use case ( not example ) it looks like what we;re wanting to accompish almost identicaly if we can abstract it http://docs.amazonwebservices.com/AWSEC2/latest/UserGuide/AESDG-chapter-instancedata.html#AMI-launch-index-examples [14:55] hazmat: ^^ ( re: dropping need for s3 for the instance info metadata ) [15:01] imbrandon, launch indexes? [15:02] thats for disambiguating metadata when launching multiple instances in a single api call [15:02] for the s3 usages its not contextually relevant [15:12] no it sets the data at launch [15:13] and can be called later via api to figure out what one is what [15:13] see theuse case for the 4 mysql servers and one needed to be the "master" [15:13] hazmat: ^^ e.g our bootstrap [15:15] she set "store-size=123PB backup-every=5min | replicate-every=1min | replicate-every=2min | replicate-every=10min | replicate-every=20min" and each server got one that chunk of meta data between each pipe as well as the details of the instance it ended up belonging to [15:16] i think it sets it via the "tags" , and hpc has those too, but not sure if its openstack or just a similar feature [15:16] * SpamapS reads [15:16] GET http://169.254.169.254/latest/user-data user_data.split('|')[ami_launch_index] [15:17] SpamapS: very bottom is the gist, the top is fluf about stuff we wont use i think [15:17] imbrandon: thats not actually helpful no. You need *clients* to be able to find it [15:17] not the instance itself [15:17] yea the client run it [15:18] or you me the conductor ? [15:18] the client can't get to 169.254.169.254 :) [15:18] ahh ok i was thinking clients as in instances [15:18] every time you type 'juju status' or 'juju deploy' .. there's a method that has to go *find* the ZK node [15:18] it can be run for any of them tho so it would just need to pick a random one to connect to [15:18] in ec2, the way that is done is by asking the S3 control bucket where it is. [15:19] imbrandon: I think we can do it with groups [15:19] kk i fond that looking for groups docs [15:19] node 0 is always in juju-$envname-0 [15:19] and in fact, is always *alone* in that group [15:19] wow, ok yea [15:19] lets just do that then and have it store the data local [15:20] when node 0 becomes HA .. we can just make a group for ZK nodes [15:20] thats super easy it seems [15:20] yea [15:20] hrm yea thats almost too easy [15:20] got to be a catch :) [15:20] heh [15:20]