lifeless | https://juju.ubuntu.com/docs/faq.html#is-it-possible-to-deploy-multiple-services-per-machine might need updates | 01:31 |
---|---|---|
lifeless | links at the bottom of https://juju.ubuntu.com/docs/faq.html are broken,obolete too - https://www.jujucharms.com/ and https://juju.ubuntu.com/kanban/dublin.html | 01:32 |
lifeless | is there an elasticsearch charm ? | 01:33 |
imbrandon | yea | 01:36 |
imbrandon | its not in the store yet | 01:36 |
imbrandon | but its in the review que | 01:37 |
imbrandon | not tried it myself | 01:37 |
imbrandon | but i think jorge may have | 01:37 |
imbrandon | he was talking about it | 01:37 |
imbrandon | there are elb and rds ones as well | 01:37 |
imbrandon | both need a tiny bit of polish but function | 01:37 |
imbrandon | if you can handle rough edges :) | 01:38 |
imbrandon | lifeless: ohh i read that as elasticache | 01:38 |
imbrandon | not elasticsearch | 01:38 |
imbrandon | no i dont think so | 01:38 |
imbrandon | not that i ahve seen tho i havent looked specificly | 01:38 |
lifeless | imbrandon: I would love ones for elasticsearch and opentsdb; it may be time for me to get into it :. | 01:39 |
imbrandon | btw i found whay looks to be all the info i need for the osx stuff ealier , not time to go over it yet but it looked completish | 01:39 |
imbrandon | so ty | 01:40 |
imbrandon | hehe yea, elasticsearch would likely be a good starter charm | 01:40 |
imbrandon | as its probably going to be simple. most external services that are charmed are | 01:40 |
imbrandon | you could see my newrelic charm or clints elb charm as examples | 01:41 |
imbrandon | of two and they are both pretty simple | 01:41 |
imbrandon | well mine is VERY , like 10 lines of code | 01:41 |
imbrandon | but his not much more but still relates etc | 01:41 |
imbrandon | but yea to be your first one, thats actually a good idea, a total beginner i might not say that but you have a solid progmatic head on you and are semi familiar with it already ( via just interaction is all i mean ) | 01:43 |
imbrandon | so i think you should be able to seriosuly pick it upa nd even have it done in an hours time or so | 01:43 |
* imbrandon afk | 01:44 | |
lifeless | btw | 02:19 |
lifeless | https://juju.ubuntu.com/docs/user-tutorial.html | 02:19 |
lifeless | fails to talk the user through environments.yaml config | 02:19 |
lifeless | first time I was looking into this stuff, it was -painful- as a result. Fear and confusion all around. | 02:20 |
lifeless | https://juju.ubuntu.com/Charms is a little incoherent on 'starting a new charm' - it says to push before describing how to init-repo etc :> {I figured it out, others may be more puzzled} | 02:41 |
lifeless | hazmat: your branch generates configs like: Acquire::HTTP::Proxy "{'apt-proxy': 'http://192.168.1.102:8080/'}"; | 03:06 |
lifeless | hazmat: this isn't quite what you intended, I think :) | 03:06 |
lifeless | hazmat: not sure if its the cloud-init support, or the juju code yet, looking into it. | 03:06 |
lifeless | metadata service claims - 'apt_proxy: {apt-proxy: 'http://192.168.1.102:8080/'}' | 03:12 |
lifeless | so I think juju | 03:12 |
lifeless | definitely: (Pdb) print machine_data | 03:14 |
lifeless | {'apt-proxy': {'apt-proxy': 'http://192.168.1.102:8080/'}, 'machine-id': '0', 'constraints': <juju.machine.constraints.Constraints object at 0x348c190>} | 03:14 |
lifeless | hazmat: also, it looks like you override it if not set, which breaks images with that preconfigured. I will see about fixing that too | 03:14 |
lifeless | hazmat: and done - https://code.launchpad.net/~lifeless/juju/apt-proxy-support/+merge/111763 - for merging into your WIP branch | 03:38 |
imbrandon | lifeless: coool ( re: the env.y thing, i'll see if i cant get some clarification there and be a bit more verbose about some of the things that yea it looks to barely touch if at alll even for getting started | 04:24 |
imbrandon | ty | 04:24 |
imbrandon | if not i'm minimally add a trello task for it so someoe will ( hopefully heh ) | 04:25 |
lifeless | :) | 04:26 |
lifeless | imbrandon: I haven't gotten to actually doing the charm yet, more yak shaving happened. | 04:27 |
lifeless | imbrandon: is there a howto for charms? e.g. step-by-step including deploying from local tree ? | 04:29 |
imbrandon | 100% complete i'm not sure or optimistic | 04:36 |
imbrandon | but i do bet that there is a recorded charm schoool | 04:36 |
imbrandon | that does tho | 04:36 |
* imbrandon looks for where they are storeed | 04:36 | |
lifeless | thanks | 04:40 |
imbrandon | yea , there are a few listed there it looks like | 04:43 |
imbrandon | https://juju.ubuntu.com/CharmSchool | 04:43 |
imbrandon | in the webinars | 04:43 |
imbrandon | ( skip to the end of the first one to get a list | 04:43 |
imbrandon | of all the current and upcomming ones ) | 04:43 |
imbrandon | after a once over they look to be fairly complete | 04:44 |
lifeless | this stuff -really- needs to be in the main docs | 04:44 |
lifeless | since we want folk doing it easily. | 04:44 |
imbrandon | yea , i am workingf on that , so is jorge and mims | 04:44 |
lifeless | cool | 04:44 |
lifeless | I realise folk are working hard | 04:44 |
lifeless | I guess I'm ensuring that this is on the radar | 04:44 |
imbrandon | trying to move EVERYTHING to the docs | 04:44 |
imbrandon | yop yup | 04:45 |
imbrandon | many and big docs importvements are on most of ours | 04:45 |
imbrandon | i know mine and jorges for sure, but client and mims and kapil i know all spend copious ammounts of time on them too at times | 04:46 |
imbrandon | heh | 04:46 |
imbrandon | i think this last thursday we had clint me kapil mimms jorge and there was one more , or juan | 04:46 |
imbrandon | all working on them at the same time | 04:46 |
imbrandon | for a good solid 4 to 6 hours | 04:47 |
imbrandon | lots done, still lots to do tho :) | 04:47 |
imbrandon | hehe | 04:47 |
imbrandon | definatly was an intresting time to see that many all at once on a docs setup like we have | 04:47 |
imbrandon | :) | 04:47 |
imbrandon | jujucharms.com/docs is much more updated thanks to that day too compared to the normal docs | 04:48 |
imbrandon | we have a IS ticket in to fix the packages on the one with the main docs but its not completed yet so they are a week or so outdauted but that week saw a metric ton of updates | 04:49 |
imbrandon | so hopefully someone can fix it up tomarrow ( just needs -backports enabled and then packages updated that are already installed ) [ if you had a lil weight to get er done quicker that would be awesome and I could fix more faster :P heheh ] | 04:50 |
lifeless | whats the RT # ? | 04:50 |
imbrandon | umm i need to look in the irc back log, clint filed it | 04:50 |
imbrandon | but he told us in the chan about ~18 hours ago | 04:51 |
imbrandon | or so | 04:51 |
* imbrandon looks quickly | 04:51 | |
imbrandon | lastlog rt | 04:51 |
imbrandon | bah | 04:51 |
lifeless | aiee | 04:52 |
lifeless | imbrandon: http://jujucharms.com/docs/ is a 403 | 04:55 |
lifeless | 'fobidden' | 04:55 |
imbrandon | nice | 04:56 |
* imbrandon pulls the branch to see if its building | 04:56 | |
imbrandon | not sure wherer the output of the cron build for juju.ubuntu.com/docs is either | 04:57 |
imbrandon | if there is one :) | 04:57 |
imbrandon | but i know its supose to build them evert 15 min on cron and did for a long time till we broke 0.6.4 and need 1.0+ 1.1.3 is current | 04:58 |
* imbrandon checks the build currently uploaded | 04:58 | |
imbrandon | thurs day we considered "dogfooding" it like mmims brought up ( and good idea i think personally ) to charm ( and this cleanup too ) juju.ubuntu.com and then just redirect the page to it on ec2/hpcloud etc | 05:00 |
imbrandon | heh | 05:00 |
imbrandon | not sure how far it would "actually" fly if it was attempted | 05:01 |
imbrandon | but its worth a thought later :) | 05:01 |
imbrandon | lifeless: http://jujucharms.com/docs/ fixed up | 05:34 |
imbrandon | made a typo in my last commit :) but that is what the official stite will also look like as soon as builds resume on a newer sphinx than Lucid default ships | 05:36 |
imbrandon | :) | 05:36 |
imbrandon | and hopefully shuffling some of the content to the docs form juju.ubuntu.com its self also so its all central in one place | 05:37 |
imbrandon | ( as well as sexu looking too I think, but just a tad bias, definately easier to read/navigate ) | 05:38 |
imbrandon | :P | 05:38 |
lifeless | did you file the rt #? | 05:38 |
imbrandon | clint did | 05:38 |
lifeless | s/file/find/ | 05:38 |
imbrandon | not yet | 05:38 |
imbrandon | i am actively looking tho | 05:39 |
lifeless | heh | 05:39 |
imbrandon | :) | 05:39 |
lifeless | ok http://jujucharms.com/docs/write-charm.html is *much* better | 05:39 |
lifeless | kudos | 05:39 |
imbrandon | :) ty | 05:40 |
lifeless | http://jujucharms.com/docs/write-charm.html - uses oneiric | 05:40 |
imbrandon | yea there are still many oneiric refs | 05:41 |
imbrandon | i am affraid to sed them out | 05:41 |
imbrandon | bit mostly that can be done | 05:41 |
imbrandon | providers is my next section to hit up hard i think | 05:42 |
imbrandon | lots of pain ares there esp with env.y config options etc | 05:42 |
lifeless | http://jujucharms.com/docs/write-charm.html doesn't know that revision is a separate file yet either. | 05:43 |
imbrandon | nice, so VERY out dated | 05:44 |
bkerensa | SpamapS: I think I borked your merge | 05:44 |
imbrandon | fixes that one now | 05:44 |
imbrandon | bkerensa: nice going slick heh | 05:44 |
imbrandon | bkerensa: j/k , something i can help sort ? | 05:44 |
imbrandon | your golden, once overed the commit and it looks correct, i'll look closer when i make my next docs commit here in a few myself too | 06:01 |
imbrandon | but i'm fairly certain your solid :) | 06:01 |
=== almaisan-away is now known as al-maisan | ||
=== wrtp is now known as rogpeppe | ||
=== zyga_ is now known as zyga | ||
=== al-maisan is now known as almaisan-away | ||
sidnei | SpamapS, added a comment on #920454, seems like precise might be missing some patches on libvirt | 13:25 |
_mup_ | Bug #920454: juju bootstrap hangs for local environment under precise on vmware <local> <juju:Confirmed> < https://launchpad.net/bugs/920454 > | 13:25 |
SpamapS | sidnei: ahh! good sleuthing.. maybe its already reported in precise? | 13:43 |
=== zyga is now known as zyga-food | ||
sidnei | SpamapS, couldn't find a !juju bug for precise no | 13:44 |
imbrandon | SpamapS: do you now what these corospond to for HPcloud | 13:55 |
imbrandon | project-id: | 13:55 |
imbrandon | nova-uri: | 13:55 |
imbrandon | auth-mode: | 13:55 |
imbrandon | ? | 13:55 |
SpamapS | imbrandon: no, I don't even know what that is. :) | 13:55 |
imbrandon | openstack hpcloud ( not openstack_s3 ) gonna try pure first ( i hope ) | 13:55 |
imbrandon | env.y settings | 13:56 |
SpamapS | imbrandon: ohhh | 13:56 |
imbrandon | i pulled them from conf | 13:56 |
imbrandon | but its the only ones i couldent match up | 13:56 |
imbrandon | to something i knew | 13:56 |
SpamapS | sidnei: Well I do think thats a libvirt problem, and it looks like there might be a patch | 13:56 |
SpamapS | imbrandon: I'd suspect auth-mode is sort of an enum | 13:56 |
imbrandon | said proj id is a int, wonder if its the tennant id | 13:57 |
imbrandon | thats and int | 13:57 |
SpamapS | imbrandon: perhaps HPCLOUD will give you those when you get your account? | 13:57 |
SpamapS | I've still never signed up | 13:57 |
imbrandon | yea i got a whole screen of creadentials | 13:57 |
SpamapS | tho with the new ostack provider maybe I will :) | 13:58 |
imbrandon | but not sure what they map to, as the name are slightly off, so i was hoping to at leaste get bootstrap wortking doe i could document them | 13:58 |
imbrandon | heh | 13:58 |
imbrandon | yea i'm trying the ostack provider now | 13:58 |
imbrandon | when you do let me know and i'll pastebin my whole env.y and give ya a head start | 13:59 |
imbrandon | err if | 13:59 |
SpamapS | it shouldn't be this hard | 13:59 |
SpamapS | perhaps the provider needs better docs | 13:59 |
imbrandon | yea | 14:00 |
SpamapS | or HPcloud needs a good slapping if they aren't using the standard terms | 14:00 |
imbrandon | there is -0- now, i;'m reading code | 14:00 |
SpamapS | oh wtf? | 14:00 |
SpamapS | nothing in docs? | 14:00 |
imbrandon | well i think they do, i think the provider is off | 14:00 |
imbrandon | its a bitdiff than ec2 as well | 14:00 |
imbrandon | but yea | 14:00 |
imbrandon | -0- docs | 14:00 |
sidnei | SpamapS, yes, i agree. so mark the bug as affecting libvirt, invalid in juju? :) | 14:00 |
imbrandon | i'm reading source to fig this stuff out | 14:00 |
SpamapS | sidnei: I'm not 100% sure its invalid in juju, so I'm waiting to do that | 14:01 |
SpamapS | sidnei: there may be a workaround we could apply after all | 14:01 |
sidnei | oki | 14:04 |
imbrandon | SpamapS: http://cl.ly/HdVH | 14:09 |
imbrandon | for your ref too, so you have an idea should you need it in a pinch | 14:09 |
SpamapS | imbrandon: I bet tenant-id is project-id | 14:10 |
imbrandon | you'll likely have to click it to zoom to read anything and it should be safe to not have to worry about hoarding ( got bits blured ) | 14:10 |
imbrandon | kk | 14:10 |
SpamapS | I remember hearing that the name in the API was up for consideration to be renamed | 14:10 |
imbrandon | heh | 14:10 |
imbrandon | that page should stay there indefinately too if you wanna bookmark that | 14:11 |
imbrandon | for ref later or something | 14:11 |
SpamapS | No I think I'll just sign up :) | 14:11 |
imbrandon | its actaly my upload account | 14:11 |
imbrandon | kk | 14:11 |
imbrandon | kool, cuz those other two are optional | 14:11 |
imbrandon | so with that i *should* have a full stanza | 14:12 |
imbrandon | of pure OSAPI on hpcloud | 14:12 |
imbrandon | *crosses fingers and prepares to fail* | 14:12 |
SpamapS | Have they at least moved up from Diablo to Essex yet? | 14:13 |
imbrandon | no idea | 14:13 |
imbrandon | i prettu sire its essex | 14:14 |
imbrandon | but dunno how to tell | 14:14 |
imbrandon | i'm prtty ignorant on openstack | 14:14 |
imbrandon | only have the absolute minimum in my head for uit so far | 14:14 |
SpamapS | I don't know how to tell either | 14:14 |
imbrandon | really not even that | 14:14 |
imbrandon | heh | 14:14 |
imbrandon | thats one reason i'm using hp and not rak, zomg their rest api is the suxors | 14:15 |
SpamapS | well, its OSTACK now IIRC | 14:15 |
imbrandon | 8 rest calls to create one mysql db on the new mysql service | 14:15 |
imbrandon | at rak | 14:15 |
imbrandon | nah | 14:15 |
imbrandon | not all of it | 14:15 |
SpamapS | Oh they're not using red dwarf? | 14:16 |
imbrandon | they have joey on half and half , mine only has db access | 14:16 |
imbrandon | then i got a toy acct that only has new access | 14:16 |
imbrandon | yea no, its all helter skelter in their control pannel right now, feels like aws but with 2 competing back ends | 14:16 |
imbrandon | heh | 14:16 |
=== zyga-food is now known as zyga | ||
imbrandon | presonaly i would avaoid rak for like a nother month id say and let them settle some dust | 14:17 |
imbrandon | heh | 14:17 |
imbrandon | btw i do those "full page screenshots + annotations on save" with a BA chrome extension "Awesome Screenshot" heh its perfect to get a whole page when it s like like that | 14:19 |
imbrandon | with one click | 14:20 |
imbrandon | nova_project_id == tennant name | 14:24 |
imbrandon | woot | 14:24 |
imbrandon | bah, these forum are such a pita to use, then o grok some of the source to see why, its d7 heh non-custom d7 at that , heh no wonder | 14:25 |
imbrandon | :) | 14:26 |
imbrandon | these forums == help docs at hpcloud , d7 == drupal 7 | 14:26 |
* imbrandon gets ready to bootstrap env hpc.1-az.1-regon-a.geo1 | 14:27 | |
imbrandon | wow | 14:27 |
* robbiew senses a disturbance in the cloud force...velocity and google IO this week. | 14:30 | |
imbrandon | heheh | 14:30 |
imbrandon | i always love IO but i always like it like 3 weeks late as videos | 14:31 |
imbrandon | :) | 14:31 |
* m_3 sad to be missing out on both events :( | 14:42 | |
twobottux | aujuju: Is juju specific to ubuntu OS on EC2 [closed] <http://askubuntu.com/questions/149952/is-juju-specific-to-ubuntu-os-on-ec2> | 15:27 |
hazmat | morning | 15:36 |
m_3 | o/ | 15:46 |
=== salgado is now known as salgado-lunch | ||
jimbaker` | SpamapS, , i took a look at the failing test on buildd, juju.lib.tests.test_format.TestPythonFormat.test_dump_load. this test normally takes on the order of 0.001s on my laptop and involves no resources other than a json serializer/deserializer. i just ran it for 10000 loops and i didn't see it explode | 16:58 |
jimbaker` | SpamapS, just looking at the test and the code it exercises, i would not expect resource starvation failures. maybe in similar code in test_invoker, which involves processes. but not here | 17:00 |
SpamapS | /wi/win 7 | 17:09 |
SpamapS | arghhh!! | 17:10 |
SpamapS | bursty network.. yargh | 17:11 |
SpamapS | jimbaker`: So perhaps it is a race of some kind? | 17:11 |
jimbaker` | SpamapS, in this specific code, no. twisted can report problems in the wrong place, so that could be the problem | 17:12 |
jimbaker` | in terms of the twisted trial test runner | 17:13 |
jimbaker` | SpamapS, i assume this failing test is stable? | 17:13 |
jimbaker` | as reported on buildd? | 17:14 |
SpamapS | jimbaker`: I'm not sure, I'll retry a few of the builds. | 17:17 |
SpamapS | jimbaker`: it may have been transient. quantal succeeded on retry | 17:18 |
SpamapS | jimbaker`: I'm retrying all the others. If they succeed, we can chalk this up to a buildd problem I think | 17:19 |
jimbaker` | SpamapS, again, this is not a test i would expect to fail transiently, since it's not async. but again, it's completely possible for twisted trial to point the wrong finger from some other transient bug | 17:20 |
jimbaker` | so given that, i think this is the best strategy for now | 17:20 |
sidnei | uhm, is there a 'juju scp' counterpart to 'juju ssh'? if not, it could be handy | 17:20 |
jimbaker` | there was one other error reported, which was a zk closing connection problem | 17:21 |
jimbaker` | sidnei, yes | 17:21 |
jimbaker` | sidnei, just to be clear, juju scp exists, not that would be hypothetically handy ;) | 17:22 |
sidnei | ah, i totally missed it in juju --help | 17:22 |
=== salgado-lunch is now known as salgado | ||
SpamapS | sidnei: its quite useful for pulling down modified hooks actually. :) | 17:33 |
sidnei | SpamapS, im trying to figure out if i can get rsync to work with that, using 'juju ssh' as the remote shell, but service name has '/' so it gets interpreted as a path instead of a machine name | 17:33 |
SpamapS | sidnei: IMO we need a 'juju get-public-hostname' so you can just do rsync `juju get-public-hostname foo/3`:/path/to/file | 17:35 |
SpamapS | sidnei: I think juju-jitsu might have that actually | 17:35 |
sidnei | ah, that could work too | 17:35 |
SpamapS | bkerensa: Hey | 17:44 |
SpamapS | bkerensa: that was a pretty MASSIVE merge proposal you merged into lp:juju/docs | 17:45 |
SpamapS | bkerensa: I'd have liked to hear more than just your +1 ;) | 17:45 |
lifeless | SpamapS: hazmat: morning guys; hope my spam overnight wasn't an issue ;) | 19:08 |
imbrandon | heya | 19:09 |
imbrandon | SpamapS: whats the rt# for that docs ticket? lifeless would like to know :) | 19:09 |
lifeless | hey! | 19:09 |
hazmat | lifeless, no that was awesome | 19:17 |
hazmat | lifeless, i'm running around today at velocity conf with meetings, i'm going to try and circle back to your branches/bugs this evening. | 19:17 |
lifeless | cool, thanks! | 19:17 |
lifeless | hazmat: the main one I'd like info on is the use of ip; that could be contentious for reasons other than code. | 19:17 |
hazmat | lifeless, yeah.. that changes the display in status as well. | 19:18 |
hazmat | lifeless, ideal would be to capture both, and use the appropriate one if the other is missing | 19:18 |
lifeless | hazmat: in a good way IMNSHO :) | 19:18 |
lifeless | hazmat: 10.0.0.3 is much more useful than 'server-2345' | 19:18 |
hazmat | lifeless, so do you have a a valid dns name in your context, or is it just ip | 19:18 |
hazmat | hmm.. so you have invalid dns names | 19:18 |
lifeless | hazmat: I assert that noone running openstack outside of rackspace and perhaps hp has valid public DNS names. | 19:19 |
lifeless | hazmat: its not even part of the deploying openstack guidelines yet. | 19:19 |
lifeless | let alone labs etc | 19:19 |
hazmat | lifeless, but everyone running in a public cloud or maas probably does | 19:19 |
hazmat | and for them i would posit its better to have things displayed by name then ip | 19:20 |
lifeless | hazmat: public cloud will, maas *may* (if they route DNS to the maas controller) | 19:20 |
lifeless | hazmat: maas seed clouds though won't, same issue as openstack. | 19:20 |
lifeless | dns will be usable w/in the cloud, but not by machines outside it. | 19:20 |
hazmat | lifeless, i've seen and assisted with several maas demos where dns does work fwiw | 19:21 |
lifeless | hazmat: from a machine that isn't the cloud controller ? | 19:21 |
lifeless | hazmat: because, that machine will be specially configured to use the dnsmasq instance on the controller. | 19:21 |
hazmat | lifeless, yes, its a machine on the same network though | 19:21 |
hazmat | lifeless, probably | 19:22 |
lifeless | I'll match your probably and raise you an almost certainly ;) | 19:22 |
lifeless | anyhow, we can gather data if we care. | 19:22 |
hazmat | lifeless, no takers :-) | 19:22 |
lifeless | I'm not sure that we -care- about the dns name though; what does it offer the user? | 19:22 |
hazmat | a nicer symbolic name | 19:23 |
hazmat | lifeless, think about configuring a vhost in apache | 19:23 |
lifeless | hazmat: have you seen the public names ec2 uses ? | 19:23 |
hazmat | or browsing to an app.. | 19:23 |
hazmat | lifeless, true | 19:23 |
lifeless | these are not the same as symbolic names | 19:23 |
lifeless | they are bulk loaded static mappings | 19:23 |
lifeless | per region | 19:24 |
lifeless | such as ec2-184-73-10-99.compute-1.amazonaws.com | 19:24 |
hazmat | agreed realistically, true dns management for services is already separate than the provider dns entries | 19:25 |
hazmat | er. a separate concern | 19:26 |
hazmat | lifeless, okay i'm convinced | 19:30 |
lifeless | hazmat: cool, thanks for taking the time; chat more when you're done @velocity | 19:39 |
SpamapS | lifeless: I'm with you on this. The DNS causes more problems than it solves. | 19:48 |
=== Guest73293 is now known as dpb_ | ||
=== dpb_ is now known as dpb__ | ||
=== dpb__ is now known as dpb___ | ||
negronjl | SpamapS: ping | 20:04 |
negronjl | SpamapS: can you give me the commands that you use to run gource with jitsu ? | 20:05 |
negronjl | SpamapS: I am trying to show the "pretty deploying stuff screen" | 20:05 |
SpamapS | jitsu gource --run-gource | 20:12 |
jimbaker` | the gource integration definitely demos well, i used it as part of my demo for usenix two weeks ago. nothing like seeing some good pictures of what's going on | 20:14 |
SpamapS | Yeah I really think a web app showing a network diagram will play even better.. if we can ever get such a thing :) | 20:15 |
jimbaker` | SpamapS, yes, i'm sure that will be quite nice :) | 20:15 |
jimbaker` | it might be nice to also demo the checking off of expectations of jitsu watch, here's how service orchestration happens | 20:17 |
lifeless | does one need to wait for a deploy to complete (via status) before doing add-relation ? | 20:38 |
lifeless | or will it Just Work if you run add-relation immediately that the deploy returns ? | 20:39 |
jimbaker` | lifeless, you can do juju add-relation as soon as you have deployed the two services | 20:42 |
lifeless | jimbaker`: as soon as the juju cli returns, you mean ? | 20:42 |
jimbaker` | it just works because add-relation records this setup in zookeeper; it's the responsibility of the agents to carry out this policy | 20:43 |
lifeless | jimbaker`: right, but deploy returns before the agents are even running | 20:43 |
jimbaker` | yes, as soon as the juju cli returns | 20:43 |
lifeless | jimbaker`: which is why I'm probing for specifics ;) | 20:43 |
jimbaker` | yes, even in that case | 20:43 |
lifeless | cool | 20:43 |
lifeless | thanks | 20:43 |
jimbaker` | because the agents carry out the policy, as recorded as it is in ZK | 20:43 |
lifeless | how can one recover from a wedged node ? | 20:44 |
lifeless | by which I mean | 20:44 |
jimbaker` | lifeless, basically what you see with the juju cli returning is that the update to zk has been made | 20:44 |
lifeless | when I run 5 or 6 deploys in quick succession, openstack is throwing a tanty and gets an unspecified 'error' on one of the nodes - it gets an instance id and never comes up | 20:44 |
lifeless | How can I tell Juju 'btw that machine, it didn't, so toss it away and start clean' | 20:45 |
jimbaker` | lifeless, sounds like a bug with the provider and possibly the provisioning agent | 20:45 |
jimbaker` | lifeless, so if you can get the provision agent log (on machine 0), that would be helpful. or use juju debug-log | 20:46 |
lifeless | jimbaker`: I'm sure its accentuated for me here and now, local openstack install. *but*, it can happen in e.g. ec2, when service disruption happens, that a reservation request goes into limbo or even fails asynchronously. | 20:46 |
lifeless | jimbaker`: the API call to provision the instance fails. | 20:46 |
lifeless | bah | 20:46 |
lifeless | succeeds* | 20:46 |
lifeless | its a cloud backend failure. Async. | 20:47 |
jimbaker` | lifeless, it should eventually succeed | 20:47 |
jimbaker` | it's a bug if it doesn't | 20:47 |
lifeless | To a degree, I agree. Ideally we could say 'its a bug over there, go fix that' | 20:47 |
jimbaker` | so some sort of occasional failure is expected | 20:47 |
lifeless | but, ^^ that. | 20:47 |
lifeless | how do I recover without wiping the whole juju environment of 10 instances | 20:47 |
jimbaker` | we just tried to engineer the provisioning agent so that it does appropriate retries | 20:47 |
lifeless | jimbaker`: but it doesn't retry if the API call succeeds? | 20:48 |
jimbaker` | lifeless, iirc, it does do dead machine detection for cases like that | 20:48 |
lifeless | how long does it wait? Perhaps I wasn't patient enough | 20:49 |
jimbaker` | lifeless, without logs, i'm afraid i can only speculate :) | 20:49 |
lifeless | ok, well I didn't capture any (didn't konw how) | 20:49 |
lifeless | juju debug-log; how do I get the provision agent log ? | 20:49 |
jimbaker` | lifeless, it should just be there, see https://juju.ubuntu.com/docs/user-tutorial.html#starting-debug-log | 20:50 |
lifeless | jimbaker`: how do I get the whole log though?I mean, this is some time back presumably... I don't notice instantly. | 20:51 |
jimbaker` | lifeless, you can also just grab it from the machine 0 box | 20:53 |
lifeless | jimbaker`: thats what I was asking | 20:53 |
lifeless | where on the box :) | 20:54 |
jimbaker` | lifeless, regrettably i don't have it committed to memory, but i am looking right now | 20:54 |
lifeless | thanks; Perhaps a doc patch :> | 20:54 |
jimbaker` | /var/log/juju/provision-agent.log | 20:55 |
lifeless | cool | 20:55 |
lifeless | I will look there next time it happens and see what I can see | 20:55 |
lifeless | thank you | 20:55 |
jimbaker` | lifeless, sounds good | 20:56 |
lifeless | \o/ finally apparently have hdfs + hadoop up, ready to iterate on stuff using it ;> | 20:56 |
lifeless | super not-simple | 20:56 |
jimbaker` | lifeless, i did check juju debug-log; two things, 1) the check against the flag it sets in ZK executes too late for the provisioning agent upon first install; 2) the PA still does not seem to use debug logging even if restarted. so for now, /var/log/juju/provision-agent.log seems to be the way to go | 21:08 |
jimbaker` | also this doesn't mix output, which is probably why i actually use the file when debugging the PA | 21:09 |
lifeless | first install - of a new service, or first install of the environment ? | 21:09 |
jimbaker` | first install of the environment | 21:09 |
jimbaker` | specifically the bootstrap node | 21:09 |
jimbaker` | (aka machine 0) | 21:09 |
lifeless | that seems hard to avoid, as the zk runs on that node | 21:10 |
lifeless | if it fails to come up, .... | 21:11 |
jimbaker` | lifeless, actually i'm going to backout that earlier statement. you are going to miss some changes to agent log, but there is a watcher in place for such global en settings (currently only debug log). so it appears to be just a fault of logging the PA log | 21:16 |
jimbaker` | lifeless, not certain why, the log setup is fairly generic in terms of adding handlers | 21:17 |
=== salgado is now known as salgado-afk | ||
lifeless | oh, lalalalalala | 22:22 |
lifeless | hazmat: SpamapS: you'll love this: | 22:22 |
lifeless | 2012-06-26 10:21:39,616 INFO Connecting to unit hbase-regioncluster-01/0 at server-10.novalocal | 22:22 |
lifeless | ssh: Could not resolve hostname server-10.novalocal: Name or service not known | 22:22 |
lifeless | (juju ssh) | 22:22 |
lifeless | novalocal is the local search domain on the instance, not globally resolvable. | 22:22 |
lifeless | INSTANCE i-0000000a ami-00000004 server-10 server-10 running 0 m1.small 2012-06-25T20:36:32.000Z nova monitoring-disabled 10.0.0.7 10.0.0.7 instance-store | 22:23 |
lifeless | is what the EC2 API returns :> | 22:23 |
lifeless | has anyone here worked with the hbase charm? I want to create a table from a different charm, which requires running the hbase binary (which is only on hbase units...) | 22:28 |
SpamapS | lifeless: double doh on the IP vs hostname | 22:38 |
lifeless | SpamapS: the rabbit hole gets deeper :) To me, this validates my first point :> Also, split view DNS sucks. | 22:41 |
SpamapS | lifeless: split view DNS is just a relic of the way EC2 has done things.. | 22:45 |
SpamapS | lifeless: you know, there's an openstack provider in review right now.. | 22:46 |
SpamapS | lifeless: have you been able to try it out? | 22:46 |
lifeless | SpamapS: nope | 22:46 |
lifeless | SpamapS: I'm not actually trying to hack on juju at all, just use it ;) | 22:47 |
lifeless | SpamapS: so I'm running precise, etc; just keep hitting yak shave events. | 22:47 |
lifeless | SpamapS: what I *want* to do, is try out opentsdb | 22:47 |
lifeless | and logstash (with elasticsearch) | 22:47 |
SpamapS | oh right | 22:47 |
SpamapS | lifeless: but you're also trying it against a local openstack..... | 22:48 |
SpamapS | which.. complicats matters. :) | 22:48 |
lifeless | SpamapS: a little | 22:48 |
lifeless | but shouldn't break charms as inside the stack dns work | 22:48 |
lifeless | s | 22:48 |
lifeless | its just the outside thing, exactly like canonistack | 22:49 |
SpamapS | yeah | 22:49 |
lifeless | e.g. | 22:51 |
lifeless | ubuntu@server-8:~$ host server-8.novalocal | 22:51 |
lifeless | server-8.novalocal has address 10.0.0.3 | 22:51 |
lifeless | and | 22:51 |
lifeless | ubuntu@server-8:~$ host server-9.novalocal | 22:51 |
lifeless | server-9.novalocal has address 10.0.0.4 | 22:51 |
lifeless | so hbase's tools decided they couldn't find the master. | 22:51 |
SpamapS | lifeless: I assume the relationship sets up the necessary configs for that? | 22:53 |
lifeless | http://jujucharms.com/charms/precise/hbase | 22:53 |
lifeless | I followed the bouncing ball there | 22:53 |
lifeless | juju deploy hbase hbase-master | 22:54 |
lifeless | juju deploy hbase hbase-regioncluster-01 | 22:54 |
lifeless | juju deploy zookeeper hbase-zookeeper | 22:54 |
lifeless | juju add-relation hbase-master hbase-zookeeper | 22:54 |
lifeless | juju add-relation hbase-regioncluster-01 hbase-zookeeper | 22:54 |
lifeless | juju deploy --config example-hadoop.yaml hadoop hdfs-namenode | 22:54 |
lifeless | juju deploy --config example-hadoop.yaml hadoop hdfs-datacluster-01 | 22:54 |
lifeless | juju add-relation hdfs-namenode:namenode hdfs-datacluster-01:datanode | 22:54 |
lifeless | juju add-relation hdfs-namenode:namenode hbase-master:namenode | 22:54 |
lifeless | juju add-relation hdfs-namenode:namenode hbase-regioncluster-01:namenode | 22:54 |
lifeless | juju add-relation hbase-master:master hbase-regioncluster-01:regionserver | 22:54 |
lifeless | ^ is exactly what I ran | 22:54 |
lifeless | with a minute or so between deploys | 22:54 |
lifeless | and nothing between relation adding | 22:54 |
lifeless | (sorry for the Spam) | 22:54 |
SpamapS | why minutes between deploys? | 22:56 |
SpamapS | You should have been able to run all of the deploys all at once | 22:56 |
SpamapS | and all the relations | 22:56 |
SpamapS | and then wait for everything to finish. :) | 22:56 |
lifeless | SpamapS: see my chat with jimbaker` above about vagaries of local stack deploys with one compute node | 22:58 |
lifeless | something whinges when I allocate 32GB of ram etc all in one batch | 22:59 |
SpamapS | lifeless: I see. Thats a bit disappointing given how "scalable" openstack is supposed to be... | 23:01 |
lifeless | well | 23:01 |
lifeless | I mean, I have one node here with 16GB ram | 23:02 |
lifeless | but yes, there is some glitch in there, and I've already shaved enough yaks on this exercise. | 23:02 |
lifeless | the local install is v useful considering latency to canonistack, f'instance | 23:02 |
SpamapS | indeed | 23:03 |
SpamapS | lifeless: I think a native OpenStack provider, and more people banging on OpenStack's with juju, will help this go a lot more smoothly. | 23:03 |
lifeless | well, I'm beyond that bit, workarounds R us. | 23:04 |
lifeless | SpamapS: whats a good charm that knows how to create db's on demand ? | 23:06 |
lifeless | I presume the wordpress <-> mysql pair does that ? | 23:07 |
jimbaker` | lifeless, that pairing would be a good choice | 23:09 |
lifeless | jimbaker`: my suspicion/concern is that it does that using the network protocol | 23:11 |
lifeless | hmmm | 23:11 |
* lifeless needs to know more hbase ops | 23:11 | |
jimbaker` | lifeless, i'm not certain what you mean by that re "network protocol". i will say that it does an exchange of relation settings to accomplish the desired service orchestration | 23:13 |
lifeless | hah | 23:15 |
lifeless | so schema details | 23:15 |
lifeless | I'd be surprised if schema migrations were passed via zk | 23:15 |
lifeless | rather than directly. | 23:15 |
lifeless | where directly == connect to the mysql network port. | 23:15 |
jimbaker` | lifeless, the theory of juju, if there is one ;), is that this is outside of the scope of the charm, unless orchestration is required | 23:17 |
jimbaker` | not trying to reinvent how mysql coordinates, just solve certain issues | 23:17 |
lifeless | jimbaker`: sure | 23:17 |
* hazmat pauses from juju talk to catchup | 23:17 | |
lifeless | its orchestration though | 23:17 |
lifeless | schema application + upgrades is a massive orchestration issue | 23:17 |
jimbaker` | lifeless, then i can see that under the scope of the charm | 23:18 |
lifeless | including the trivial-enough first-bringup case. | 23:18 |
lifeless | the charm needs to facilitate it | 23:18 |
lifeless | which is what I'm trying to figure out for hbase : | 23:18 |
lifeless | > | 23:18 |
jimbaker` | so if the services need to coordinate on add-relation/remove-relation or add-unit/remove-unit, then sure, use juju there | 23:19 |
lifeless | I think we're talking past each other | 23:20 |
SpamapS | lifeless: the charm gets you credentials and a place to connect to. But no doubt, if a relationship is more specific, schema details and dependencies across multiple charms could be orchestrated very easily. | 23:21 |
SpamapS | lifeless: such a thing just has yet to surface yet. | 23:21 |
jimbaker` | lifeless, in particular, this can be done through an advanced form of service orchestration, which orchestrates with respect to relations. adam_g has been doing this for the openstack charms fwiw | 23:22 |
jimbaker` | lifeless, you might need to do this. i guess we just need to settle on something concrete to say one way or the other :) | 23:23 |
* hazmat gets back to the audience questions | 23:25 | |
zirpu | velocity talk went well. woot | 23:55 |
m_3 | whoohoo! | 23:57 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!