[00:06] FYI, there's a new charm-tools in the PPA now [00:07] includes 'unpromulgate' === imbrando1 is now known as imbrandon [00:20] imbrandon: did you get your newrelic-php charm to deploy? [00:20] imbrandon: the metadata is a bit confusing [00:20] imbrandon: no interface: juju-info .. [00:20] SpamapS: i was about to try in in just a few minutes [00:20] no i did it blind so far [00:20] untill a few ago no to deploy it [00:20] imbrandon: I think you need interface: juju-info [00:20] i thought i put it [00:21] * imbrandon looks [00:21] SpamapS: i can give you a valid key [00:21] if you wanna try it too [00:21] the one i use for brandonholtsclaw.com so the numbers dont matter if a little skewed [00:21] etc [00:22] well I'd think juju would just refuse [00:23] newrelic have a free account for low traffic sites IIRC, so you can get a key easy. [00:23] yeah that should go in the README of the charm :) [00:23] lifeless: yea for the system monitor only [00:23] lifeless: not the lang ones like php [00:24] this is a php one [00:24] imbrandon: wait, what ? [00:24] that sucks [00:24] imbrandon: copyright: This charm is not endorsed or affiliated with the Locker project and the Locker source is the respective work of its author. [00:24] yea the free only does like cpu memory and disk and like that [00:24] $pointless [00:24] SpamapS: whoop [00:24] SpamapS: sounds like an old push [00:24] SpamapS: let me make sure bzr is uptodate [00:25] bzrhater [00:25] subordinate: true [00:25] requires: [00:25] juju-info: [00:25] scope: container [00:25] provides: [00:25] logging: [00:25] interface: newrelic-php [00:25] is that not right ? [00:25] nope [00:25] at the same level as scope: container, you need 'interface: juju-info' [00:25] at least, thats how I read the docs [00:25] ok and leave the other too [00:26] so two juju-infos [00:26] ? [00:26] same level as scope [00:26] requires: [00:26] juju-info: [00:26] right but do i leave the other [00:26] interface: juju-info [00:26] ok [00:26] scope: container [00:26] kk yea thats what i was asking if i left both [00:26] * imbrandon fixes [00:27] no files dir either [00:27] lifeless: yea "system" monitoring is free for unlimited servers, "application" monitoring is now [00:27] not* [00:28] yea bzr was out dates [00:28] bholtsclaw@ares:~/Projects/local/charms/precise/newrelic$ bzr add . [00:28] adding files [00:28] adding files/548C16BF.gpg [00:28] doh [00:28] * SpamapS suspends review [00:28] imbrandon: I think you are wrong [00:28] 3 seconds and i'll have it pushed [00:28] http://newrelic.com/pricing/details [00:28] lite [00:28] free [00:29] right but thats for the newrelic-sysmon [00:29] includes basic db and app monitoring [00:29] anyhow [00:29] ahh yea [00:29] when theysay basic they mean BASIC [00:29] oops gathers more data :) [00:29] as in is it running [00:30] and is always free [00:30] but yea [00:30] * imbrandon will add a bit about that in the readme [00:30] imbrandon: you need a start/stop hook [00:30] contrary to popular belief, having an init.d installed in runlevel 2 is not enough [00:31] SpamapS: k [00:31] eventually juju will use stop/start to support migrating services, and possibly snapshotting. [00:32] SpamapS: also if you do grab a lite key while i push this, and the "deploy" it to test the charm, they count that as a depooy and will send ya a data-nerd t-shirt :_ [00:32] I am starting to think more and more that we should take the debhelper 7 approach and try to get this all down to one hook file that just calls out to various helpers [00:32] heh [00:32] 99% of charms are doing the same things [00:33] No revisions or tags to pull. [00:34] push [00:34] ed [00:34] should be rev 3 [00:34] on bzr [00:37] SpamapS: PM with key if you wanna try without hassle of signing up, just a convience [00:40] just added a line in the readme too about getting a ( free ) newrelic account [00:40] ty lifeless , i mis-understood [00:40] oh nice , I get a data nerd shirt [00:40] yup yup, i got two from them , nice shirts actually [00:41] and i got a book "The Lean Startup" hard cover they was giviing away a few weeks ago [00:41] they are always giving some kinda swag out to twitter followers [00:42] Ok I'm going to point it at my thinkup instance [00:42] kk [00:43] i'm after testing and prom etc i'm gonna deploy it to omg instead of how its done by hand now [00:43] omg is using cloudflare now too [00:43] :) [00:44] 2012-05-01 17:43:44,651 unit:newrelic-php/0: hook.output ERROR: /var/lib/juju/units/newrelic-php-0/units/newrelic-php-0/charm/hooks/install: line 13: syntax error near unexpected token `else' [00:44] crap [00:44] * imbrandon looks [00:44] forgot ; after ] [00:44] damn i was gonna say this is like the simplest chamr [00:44] lol [00:45] pushed [00:45] yeah I just haxored it in and I'm rerunning the install hook now [00:45] kk [00:45] :) [00:45] * SpamapS hugs juju resolved --retry [00:45] I think we should just alias that to 'juju retry' [00:46] I almost always want --retry anyway [00:46] 2012-05-01 17:46:09,907 unit:newrelic-php/0: hook.output ERROR: /var/lib/juju/units/newrelic-php-0/units/newrelic-php-0/charm/hooks/install: line 10: md5: command not found [00:46] ha-ha [00:46] md5sum [00:46] imbrandon: man u suck at this! ;-) [00:46] * SpamapS hides [00:46] i am so gonna make the error and info functions in juju to be "ohai = info " "ohnoe = warn" "ohpoo = error" [00:46] wow [00:46] imbrandon: misunderstood what ? [00:47] lifeless: the free leel [00:47] level [00:47] SpamapS: no md5sum on base install ? [00:47] frak [00:47] no *md5* [00:47] md5sum is there [00:47] well, it might be [00:47] not entirely sure actually [00:47] ph, frakin osx thin [00:48] there is here [00:48] not sure on linux [00:48] imbrandon: ah, so it does do more than you thought ? [00:48] bholtsclaw@ares:~/Projects/local/charms/precise/newrelic$ md5 -q files/548C16BF.gpg [00:48] be9704aa1e7001d855a60c4c05d7598b [00:48] bholtsclaw@ares:~/Projects/local/charms/precise/newrelic$ [00:48] md5sum is in coreutils [00:48] lifeless: yup, not a lot , i'd still pay,. but yea [00:49] imbrandon: realistically, that check doesn't add any security. If a bad actor could change that, it could change install [00:49] true [00:49] * imbrandon just drops it [00:49] 2012-05-01 17:49:21,939 unit:newrelic-php/0: hook.output ERROR: /var/lib/juju/units/newrelic-php-0/units/newrelic-php-0/charm/hooks/config-changed: line 12: /ect/newrelic.cfg: No such file or directory [00:49] jesus i am bad [00:49] imbrandon: I am your debugger [00:50] imbrandon: perhaps consider variables rather than repetitive hard coding as a strategy to mitigate mt. dew induced type-o impact [00:50] imbrandon: or maybe Mavis Beacon? :) [00:51] * SpamapS will stop going all alpha-dog on imbrandon now :) [00:51] all pushed [00:51] heheh [00:51] np [00:51] i deserve it most of the time :) [00:51] including now [00:51] 2012-05-01 17:51:24,404 unit:newrelic-php/0: hook.output INFO: Starting New Relic Proxy Daemon: newrelic-daemon [00:51] 2012-05-01 17:51:25,412 unit:newrelic-php/0: hook.output INFO: FAILED [00:51] no lic key [00:51] likely [00:52] I did set one, I think [00:52] perhaps not in the confusion [00:52] SpamapS: Would you like to share some examples of your ninja variables work in a charm? :D [00:52] examples accepted [00:53] A=speling [00:53] lol [00:53] echo "I suck at $A even when I'm competing in a $A bee" [00:54] Starting New Relic Proxy Daemon: newrelic-daemon OK [00:54] shaweet [00:55] imbrandon: one weakness.. [00:55] imbrandon: it doesn't restart apache [00:55] yea i thought about that [00:55] imbrandon: seems like you could just restart any that are running. [00:55] i would need tro know any way php could be installed and start apache nginx php-fom [00:55] etc [00:56] imbrandon: thats what the scope: container relationships are for [00:56] ? [00:56] it stil dont help me to know the type i need to restart [00:56] imbrandon: you can have one interface: php-app-server ... and then the charms that implement it can feed back things like what service to restart [00:56] oh yea [00:56] that way [00:57] yea that was the "comming soon" thing we mentioned [00:57] btw you should be able to refresh the page once or twice and it start feeding data [00:58] its almost realtime [01:00] I don't see any data yet [01:00] but, it appears to be working basically [01:01] make sure your on the apps tab [01:01] and not the system one [01:01] at the top left [01:01] imbrandon: thats pretty impressive [01:01] :) the sysmon one will fill in the system tab [01:01] etc [01:02] then ruby and python [01:02] et al [01:02] imbrandon: I wonder if we shouldn't just make a "newrelic" subordinate and have each different app type installable via relation or config [01:02] well i thought about that for the apps but figured it would be configin [01:03] confusing* [01:03] the reason I like doing it that way is that you can have one place for your key.. one place to upgrade [01:03] oh also, I'd recommend calling the config option just 'key' [01:03] stil all uses /etc/newrelic.cfg [01:03] or whatever [01:03] kk [01:04] no reason to prefix it when it is already namespaced into the charm [01:04] k [01:04] so in theory [01:04] then as long as i do them all the same [01:04] if you installed the python one now [01:04] you would not need to mess with the key [01:04] since i put it in the if/else [01:05] on config-change only [01:05] right [01:05] so kinda the same just not said expliositly [01:05] bah [01:05] Interesting, it seems like it held off feeding back the massive spike that my siege created until I let the load drop [01:05] its a little behind [01:06] not 1000% realtime [01:06] but almost [01:06] damn tho [01:06] drilling down now [01:06] We'd have been able to fire 3 people at my last place w/ this [01:06] :) [01:06] not because they wouldn't have work to do [01:06] yea dude it rocks, and thats just the default settings [01:06] but we'd finally be able to prove that their code was causing all the suck [01:07] if you get into the api you can set some treally cooll stuff [01:07] https://newrelic.com/docs/php/the-php-api [01:08] wow [01:08] what a PoS thinkup is [01:08] https://newrelic.com/docs/php/php-agent-phpini-settings#inivar-tt-top100 [01:08] every hit to index.php does show tables like '?' [01:09] heh nice [01:09] this app wouldn't even work on mysql 5.0 .. it takes a table lock (briefly) to do that [01:09] good thing for mysql query cache [01:09] lol [01:09] bunch of lazy load insanity here [01:10] newrelic.framework = "" [01:10] Scope: PERDIR [01:10] Type: String [01:10] New Relic tries hard to automatically detect the frameworks it supports, but it can get it wrong if you are using experimental new versions or you have customized the framework. This setting can be used to manually set the framework if the auto-detection fails. This must be one of "cakephp", "codeigniter", "drupal", "joomla", "kohana", "magento", "mediawiki", "symfony", "wordpress", "yii", "zend" or "no_framework" to force no framework even if one was detec [01:10] ^^ very nice [01:10] imbrandon: the mysql QCache is also a problem.. though they've finally got it working ok by splitting it up into multiple mutexes [01:11] SpamapS: yea an ppl often make it huge thinking it will help [01:11] and iit burries them [01:13] yeah, I went through the thrashing up and down for a while at my last place before I realized turning it *OFF* actually made things better :) [01:13] and yea in like a day or so you'll get an email from $sales saying "we sent ya a tshirt, enjoy" but then takes like 2 weeks to show :) [01:13] just fyi [01:13] I didn't give them an address yet [01:13] but thats cool [01:13] ahh ok [01:13] I'll gladly wear it, this is cool stuff [01:13] :) [01:14] SET time_zone = '?' [01:14] HAHAHA [01:14] theres one that just runs into concurrency problems [01:14] 1,483 ms [01:14] wow [01:15] imbrandon: here's one thought. What about just restarting apache2 and php5-fpm .. those two would cover 99.9% of the cases where a daemon is serving PHP [01:15] good call [01:15] untill something better [01:15] is done [01:15] * imbrandon adds it [01:16] is the service apache or aoache2 ( /me dont have it installed ) [01:16] apache2 [01:16] k [01:16] and use the service command [01:16] that way it works even if we move it to upstart [01:16] yea i do [01:16] actually, /etc/init.d/apache2 works that way too.. so.. :-P [01:16] cuz php5-fpm will restart its self that way [01:16] but, service > /etc/init.d/oldcrap [01:17] yea i like upstart works liek god [01:17] god == the bomb tho [01:18] imbrandon: ok, so thats my review. Change the name of the config option to 'key' and restart the services. Everything else can stay as a TODO for the future. Well done. :) [01:19] ty ty and i'm pushing herer in a 30 seconds [01:19] also what do you think about calling it newrelic [01:20] nah, its the php one , there is another that will be just newrelic [01:20] Ok [01:20] this installs the newrelic.so [01:20] sysmon will be just newrelic [01:20] We need to grow 'replaces' and 'provides' like capabilities for charms so when we consolidate those or rename something users can get it on an upgrade-charm [01:20] as i needs absolutely nothing else [01:20] heh, 'juju dist-upgrade' [01:20] lol [01:20] * SpamapS is not sure he likes the implications there [01:21] Ok time to go [01:21] imbrandon: will promulgate later tonight [01:21] brew upgrade all [01:21] family is home [01:21] SpamapS: ty ty [01:21] kk [01:25] SpamapS: ok pushed, for when you do get back round [01:25] * imbrandon grabs some late dinner [01:29] is there a no-cache option to forcibly reload a charm from the repository? [01:29] when doing a deployment ^ [01:30] hrm not sure, i think it just looks at the revision [01:31] ok. too bad, that would be handy [02:06] SpamapS: how far behind is the go client, as in is it useable _at all_ now and not upto feature parady or ... === ahs3` is now known as ahs3 [02:47] SpamapS: zomg sexy .... love; why cant this be unity .... me want .... https://plus.google.com/u/0/100530892038948253747/posts [03:40] imbrandon: as I understand it, a week or two ago the go client was able to spin up ec2 instances [03:40] imbrandon: and some details about how the bootstrap will differ from the python client are being worked out [03:40] imbrandon: so, its still quite far behind, but the foundations are laid [03:40] imbrandon: FYI, there is a --upgrade option to deploy that satisfies what nathwill was asking earlier [03:43] ahh [03:44] SpamapS: kk, i was thinkning about compiling the go here local and trying to kick the tires a bit , nothing serious really just see what works and dont [03:51] I tried that about 2 months ago and there was nothing, but I'm told there is at least something to play with now [03:51] imbrandon: you can also start up your own charm store ;) [03:59] oh ?>\ [03:59] with go or ? [03:59] just in general ? [04:17] imbrandon: the charm store backend is written in go [04:17] and is in lp:juju/go/store [04:28] imbrandon: still have that bit about Locker in the copyright [04:28] oh wait no [04:28] it says newrelic now, never mind :) [04:29] :) [04:29] imbrandon: so, the provides .. [04:29] that needs to go [04:29] unless you're going to add a hook, or explain who will require: that [04:29] ... um i wantthat later [04:29] is it going to be purely for private-address/public-address ? [04:30] probably [04:30] and unit names [04:31] imbrandon: so you don't actually have to do anything on the newrelic side of things? [04:31] and the ability to have other svc change its configs [04:31] huh ? [04:32] imbrandon: I'm asking if its possible w/o hooks [04:32] if what is [04:32] your not clear at all [04:32] imbrandon: that relation [04:32] i'm still very lost [04:32] You have a relation [04:32] with no hooks [04:32] yes [04:32] I'm questioning the value of such a thing [04:33] its for later so names and other config bits can easily be set [04:33] from peers [04:33] There is an *implied* juju-info relation in all charms. https://juju.ubuntu.com/docs/implicit-relations.html [04:33] imbrandon: until you have actual hooks, just use that. [04:33] right but it wont always be juju-info that was a hack till i get php-app in all others [04:34] no then i got to go back and change it yet again [04:34] why [04:34] step back, this is the provides, not the requires [04:34] yes , it proviodes a php logging interfae [04:34] it does [04:35] so, the only way an interface: can be defined is by an implementation (currently). I'm hesitant to have an interface w/o an implementation. [04:35] Oh, does it like, listen on a port or something? [04:35] what does it hurt, and it helps by making things very clear [04:35] yes [04:35] the daemon does [04:35] this is uncharted territory [04:36] so tell me to F off all you want. :) [04:36] right [04:36] I'll take it well [04:36] heheh well i'm looking at it more like i'll probably use it so good to have it from the get go and it dont hurt [04:36] So at this point, interface: newrelic-php , listens on a given port on the private-address. [04:36] instead of doing like juju-info and then changing later [04:37] right, well right now its 127.0.0.1/unix/sockt [04:37] *doh* [04:37] but i will have it just use one daemon and a tcp port [04:37] when ther is more than one [04:37] i'm still missing what why its bad to have [04:38] and i see it as good even witout a implmntation as it makes it very clear [04:38] even in the status [04:38] It still sounds like its a broken interface at this point. [04:38] no documentation for it.. [04:38] no partner implementation [04:38] Not saying "NO" [04:38] just asking for a good reason to keep it [04:39] and i'm asking for a good reason to ditch it [04:39] when it sounds like you'll have to change things in the charm to make it actually work anyway [04:39] cause its broken [04:39] no it works fine [04:39] in telling me that via the meta info its a logging php infterface [04:39] there's nothing that can relate to it [04:40] that cant be the *only* reason for the meta info otherise why have it [04:40] and just go on whatever interfaces ahve implentations [04:40] It is *definitely* the only reason [04:40] then just enumerate the implmented interfaces [04:40] and ditch all the meta info [04:40] Otherwise somebody goes off and writes themselves a thing that wants to log to newrelic-php [04:40] and then they get really annoyed that it won't work [04:40] and they have no example to follow from the author [04:41] if if exposes no info for them to do so [04:41] then they wont [04:41] and i linked the docs in the readme [04:41] So, it doesn't say 'will-provide' logging, it says 'provides' logging. And yet, it listens on no ports. [04:41] logging dosent have to mean it provieds to all [04:41] or on a port [04:42] it does to the app [04:42] and only the app [04:42] but it still does [04:42] and its still an interface [04:42] thats scope: container then [04:42] its in container [04:42] Yeah, so thats not relatable over the network [04:42] so add scope: container [04:42] omg [04:42] yes it is and no [04:42] it is not setup that way [04:43] there is not a config opetion [04:43] so it is not there [04:43] for public use [04:43] but it IS there [04:43] Well since you can't show me how its setup, because you haven't implemented it, I stand by my -1, if for no other reaosn than primal fear of the unknown. :) [04:43] look in the only config [04:44] thats not -1 as in, NACK for cs: [04:44] just, -1 [04:44] you're good to go for cs: [04:44] ... >.< [04:44] though I'm going to open a bug in the charm as a reminder to fix it or drop it. :) [04:44] ok then i'm truely missing something [04:45] and i'll mark as wontfix unless you can tell me what the hell you mean [04:45] relations define *network* service relationships. There is no logging network interface exposed by the installed/configured software. [04:45] so it is effectively broken [04:45]